Chapter  61:  Management of Cancer Symptoms: Pain, Depression, and Fatigue: Evidence Report/Technology Assessment Number 61

Patient Population and Settings

The EPC accepted all studies published in English of patients with a diagnosis of cancer who suffered from pain, depression, or fatigue due to cancer or treatment of cancer. It placed no restrictions on the patients' age, gender, ethnicity, level of advancement of the primary disease (staging), or presence of metastases. The conference planning committee was interested in covering the full trajectory of disease, including but not limited to, periods of active treatment and end of life.

Literature Search

Literature searches were conducted to identify studies published between 1966 and 2001 in MEDLINE^®, CANCERLIT^®, and the Cochrane Controlled Trials Registry. For cancer pain, the EPC applied the same search strategy used in its previously published Management of Cancer Pain evidence report to identify new studies published in the period from December 1998 through June 2001. The National Library of Medicine, as a partner in the NIH Consensus Development Conference process, with input from the EPC staff, performed the literature search for cancer-related depression and cancer-related fatigue. The searches were supplemented with reviews of bibliography of selected references. The EPC also identified published meta-analyses and used their data for selected topics.

Study Selection

Only studies that assessed the prevalence of the symptom as the primary purpose of the study were used for estimating the prevalence of cancer-related symptoms. For assessment, both retrospective and prospective studies were used, as well as randomized and nonrandomized trials, and cross-sectional and longitudinal studies. Randomized controlled trials were used to analyze efficacy of interventions.

Reporting the Results

The nine topics addressed in this evidence report are presented in the order of prevalence, assessment, and treatment. Each of these issues covers the symptoms of pain, depression, and fatigue. Evidence is summarized using three complementary approaches. Evidence tables provide detailed information about the characteristics and outcomes of all the studies examined. Information from the evidence tables is synthesized into summary tables describing the findings of each study. A narrative description of the studies along with an evidence-grading scheme accompanies the summary tables.

Prevalence of Cancer-related Pain

Surveillance data on the incidence and prevalence of cancer and observational and survey data on the incidence of cancer-related pain indicate that a majority of patients experience pain at some point during their course of treatment, and that cancer pain impairs quality of life and functionality. This disturbing finding reflects data from developed countries, where patients are often in tertiary care or specialist consultative settings. The likelihood of pain increases, as does its severity, with advancing cancer stage. (Minorities, women, and the elderly may be at greater risk for undertreatment of cancer pain.) Pain is generally not eliminated, despite analgesic therapy administered according to the World Health Organization method for cancer pain relief, and may continue to be a problem even after eradication of the underlying neoplasia. Multiple processes underlie cancer-related pain, yet survey data for the most part do not distinguish between different etiologies and mechanisms, nor do they provide a comprehensive picture of pain over the continuum of care, nor of the relationship between effectiveness of pain control and quality of life. The number of patients enrolled in methodologically sound trials of cancer pain relief is a small fraction of those receiving care.

Prevalence of Cancer-related Depression

Major depression and depressive symptoms occur frequently in patients with cancer. Despite standardized measures to calculate incidence and prevalence, there is a wide range of reported data. Prevalence rates varied from 10 to 25 percent for major depressive disorders and a similar range exists for clinically significant depressive symptoms. This range is the probable result of several factors that include timing of the assessment, concurrent treatment, medical morbidity, and pain, gender, and age. Cancer patients are a heterogeneous population with different sociodemographics, cancer types, treatments, and responses to treatment. Given that the estimated point prevalence of major depression in the general population is 2.2 percent, the rates in cancer patients may be at least four times greater.

During the time frame of the studies, reports of incidence ranged widely from about 2 to 17 percent. However, these studies like other prevalence studies face the same difficulties of heterogeneous populations, and there are too few naturalistic studies that follow patients from the point of diagnosis and few that serially measure depression.

Prevalence of Cancer-related Fatigue

Estimations of fatigue prevalence have been performed in the setting of many types of cancer treatment, in the palliative setting, and among cancer survivors, but the data is by no means consistent or comprehensive. Many types of cancer were not specifically addressed.

A very broad range of prevalence rates has been reported, from 4 percent in breast cancer prior to starting chemotherapy and 8 percent in prostate cancer prior to radiation therapy, to 91 percent in breast cancer patients after surgery and chemotherapy and before bone marrow transplantation. Findings of significant concern were the prevalence rates of fatigue in cancer survivors: 26 percent in Hodgkin's disease survivors; 35 to 56 percent in breast cancer survivors; and 48 percent in a cohort treated for various cancers. Comparisons of the prevalence rates in these studies are problematic, however, since each study used different criteria for defining the presence or absence of fatigue and its severity.

Assessment of Cancer-related Pain

Many types of instruments are applied to assess pain and related analgesic outcomes. In 218 trials, 125 distinct tools were employed. By far the most frequently employed were unidimensional scales of pain intensity, followed by scales of pain relief, then measures of peak or summed pain intensity differences between experimental and control groups. Other tools applied in the selected studies include global evaluations of efficacy and the McGill-Melzack pain questionnaire. Also applied were measures of analgesic consumption and a four-point side effect scale. Descriptions of the need for detailed assessment conducted within a psychosocial framework are presented in virtually all guidelines or monographs on cancer pain management. A voluminous literature describes the multidimensional, experiential nature of cancer pain and links poor control of cancer pain to impaired quality of life, including functionality. Current expectations for detailed, multidimensional assessment of cancer pain, including quality of life assessment, during cancer care contrast with the minimalist assessments of pain intensity presented during relatively brief observation intervals reported in nearly all of the trials. Side effects limit analgesic dosage and hence impede pain control in many patients, yet only one of the 16 most widely employed outcomes measures is concerned with side effects; that one is a coarse, four-point measure.

Assessment of Cancer-related Depression

Because depression may go undetected and thereby untreated in oncology practice, the importance of appropriate assessment and screening tools has been emphasized. Some assessments, like the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders (SCID), may be useful in research studies, but they are too time consuming in a clinical setting. Briefer self-report assessments are available for clinical use. These assessments range from questionnaires to The Distress Thermometer, a visual analogue scale that the National Comprehensive Cancer Network (NCCN) guidelines suggest for the screening of psychosocial distress.

While the standard of care for diagnosing depression is a clinical interview, available data on the sensitivity, specificity, predictive values, and cross-correlations of assessment instruments are presented in the evidence-based table.

Although these assessment tools may be valid, there is currently no evidence on how widely they are used clinically or whether they affect clinical care and outcomes.

Assessment of Cancer-related Fatigue

A wide array of patient self-assessment instruments has been used to evaluate fatigue. Most studies in the last several years have used instruments that assess multiple dimensions of fatigue and have been tested for validity, consistency, and reliability. Issues still remain in terms of the clinical interpretation of the scores obtained on these instruments, and the comparison of fatigue measurements obtained using different instruments. Methods for evaluating fatigue in practice settings have not been the subject of extensive research. The NCCN has published guidelines on cancer-related fatigue that include a general approach to assessment of fatigue in clinical practice. This approach is based on the experience of a panel of experts rather than on evidence from randomized controlled trials.

Treatment of Cancer-related Pain

Direct inter-class comparisons of efficacy do not differentiate between the relative efficacy of opioids and nonsteroidal antiinflammatory drugs (NSAIDs) administered through various routes to patients with mild, moderate, or severe cancer pain. Opioid dose-sparing is achieved by co-administration of NSAIDs but without a consistently demonstrable reduction in side effects. The heterogeneity of existing trials precludes meta-analyses to address most subquestions. A difference in analgesic efficacy between NSAIDs was only evident in a single retrieved trial. Likewise, the efficacy of NSAIDs versus "weak" opioids could not be discerned in the retrieved trials. However, such trials enroll relatively small numbers of patients and follow them for intervals of hours to days, and only occasionally as long as 2 weeks. Many examine drugs not available in the United States or no longer in general use for cancer pain relief (e.g., pentazocine). Prior efforts described in the previous evidence report to strengthen such evidence by examining nonrandomized trials were not fruitful. One randomized controlled trial evaluated oral transmucosal fentanyl citrate for breakthrough pain (using a study design in which rescue doses of morphine were available) and demonstrated its superiority to placebo. Another randomized study in ambulatory cancer patients provided evidence for greater analgesia and faster onset of relief after oral transmucosal fentanyl citrate than after the usual rescue drugs used by these patients. The EPC found no randomized controlled trials addressing analgesic efficacy and safety of NSAIDs selective for the cyclooxygenase-2 isozyme in treating cancer pain. The use of bisphosphonates and radiation therapy are both supported by the retrieved trials. Unfortunately, studies that point to the optimal sequence of application of the many currently available interventions for pain control were not identified.

Treatment of Cancer-related Depression

Current evidence shows that psychosocial interventions are beneficial for depressive symptoms in cancer patients, but the magnitude of the effect size seems to be in the mild to moderate range. Because there are hundreds of studies on psychosocial interventions in cancer patients, we limited our analysis to published meta-analyses of these studies. Here, the contribution of preventative studies and depression treatment studies were not defined. The effects of these interventions may vary in these two different kinds of studies.

Although not all pharmacologic studies showed benefit for depression in cancer patients, every study that used antidepressants and conformed to usual practices for antidepressant trials did. Since antidepressants typically can take 4 to 6 weeks for their full effect, studies of antidepressants under 6 weeks tended to show less benefit. Currently, there is data that selective serotonin reuptake inhibitors (SSRIs) and tricyclic antidepressants are effective. Although trazodone, an atypical antidepressant, showed some benefit in treating depressive symptoms, it is not commonly used as an antidepressant because of severe sedation at therapeutic doses.

Although there have been reports describing alternative or complementary therapy programs, there have been no controlled trials for their efficacy for depression in people with cancer.

Treatment of Cancer-related Fatigue

A limited number of controlled clinical trials of treatment for cancer-related fatigue have been published. The only treatment supported strongly by the available clinical evidence is the use of epoetin alfa in patients with anemia due to chemotherapy treatment. A few controlled trials evaluated exercise programs, in some cases with promising but preliminary results. Some positive outcomes have also been reported with psychosocial interventions.

Treatment trials for cancer-related fatigue usually have small sample sizes, and there is a possibility that many of these studies were underpowered to detect the outcome of interest.

Cancer-related Pain

Randomized controlled trials establish that many current treatment modalities can individually reduce cancer pain. The scientific evidence on cancer pain relief, however, compares unfavorably with the massive amount of information known about the efficacy and effectiveness of treatments for other high-impact conditions, including cancer itself. Quality of life has not been uniformly assessed in trials of analgesic drugs and non-drug interventions for cancer pain. Limited evidence from the retrieved trials demonstrates that optimal analgesia benefits quality of life. Advances in quality-of-life assessment and insights from research on chronic non-cancer pain into relationships between pain, disability, and impairment offer the opportunity to begin to understand these interactions in the context of cancer pain. Carefully designed trials with cancer pain relief as a primary outcome are required in patients with well-defined disease and pain mechanisms. Such trials must conform to rising expectations for clinical trials in general. High-quality trials of cancer pain relief should enroll greater numbers of patients for longer intervals than has generally been true in the past; apply blinding and active placebos when appropriate, or uniform control treatments otherwise; employ adequate between-arm washout intervals and consider advancing disease state in crossover trials; and assess side effects, pain mechanisms, and rest, incident, or breakthrough pain in a standardized, combinable fashion. Investigations of cancer pain and its control should seek to evaluate the influence of gender, race, age, psychosocial context, ethnicity, and culture on the experience and report of pain. The influence of such factors should also be examined during studies aimed at defining the efficacy of specific treatments and their associated side effects. Drug interactions during long-term cancer pain treatment require clarification. It is unclear whether a mechanism-based approach to diagnosing and relieving each component of pain in an individual is more effective than an empiric regimen in which each patient's treatment is based upon pain intensity alone. Another key unanswered question is how to optimally combine drug with non-drug therapies, given that the latter are safe and inexpensive. Despite the importance of pediatric cancer pain control, practically no analgesic drug trials focus on children.

Data that address individual variations in preferences for, responses to, and costs incurred by these options are a foundation for potential evidence-based approaches to cancer pain control, but are sparse. For example, the spinal route of analgesia is widely employed but much remains to be learned about optimal patient selection, the comparative efficacy of spinal drug infusion versus systemic drug administration, and the selection of initial or secondary agents or combinations. Exploring these fundamental questions will enhance the ability of translational clinical research to clarify the clinical relevance of an increasing number of basic insights into unique mechanisms and mediators of cancer pain.

Cancer-related Depression

There is much variance in the literature on reports of rates of depression in cancer patients. Even when standardized instruments are used, wide variance is still observed. One recommendation would be to conduct more prevalence studies that examine the reasons for such variance and contributing factors for differing rates. The timing of measurements of depressive symptoms does appear to be important and may contribute to the variance. One goal may be to develop a statistical model that could predict the rate of depression given the cancer and treatment demographics of the population. Studies should always include assessment of past histories of depression.

The existing incidence studies of depression in patients with cancer all start at some time after the diagnosis of cancer. It is recommended that more prospective studies start at the time of diagnosis, or even before, in order to arrive at more accurate estimates of the incidence of depression once people are diagnosed with cancer. These studies should also assess past histories of depression.

There are many instruments currently in use for assessing depression in cancer patients. Researchers can select from a variety of instruments based on weighing the ease of use for their study population and the effectiveness of an individual instrument as documented in the evidence tables. There should be further trials to replicate the promising results of a single-item screening, asking, "Are you depressed?"

Although some of these instruments are widely used in clinical practice, further research on their effectiveness is needed. The development of brief instruments that assess all three symptoms (depression, pain, and fatigue) could be one area of future research.

Psychopharmacologic, psychosocial, and alternative interventions offer some benefit on treatment for depressive symptoms with cancer patients. There are great opportunities for research on psychopharmacologic interventions for depression co-morbid with cancer. The newer antidepressants, especially the atypical ones, need to be studied in this population. Although antidepressant trials are more complicated to conduct in cancer patients, they should still adhere to a standard study length of 6 weeks or greater. Common clinical practices, such as the use of psychostimulants for depression, need to be evaluated in controlled trials. There should also be more research on the use of antidepressants for the prevention of depressive symptoms in patients with cancer.

Hundreds of studies exist on psychosocial interventions for cancer patients and depression, but a meta-analysis of psychosocial therapies specifically for the treatment of depression in cancer patients remains to be done. Although many patients may be using complementary and alternative treatments, controlled trials are required to determine their efficacy in depression co-morbid with cancer.

Cancer-related Fatigue

Future research in cancer-related fatigue should also include more comprehensive studies of the prevalence of fatigue in a wider variety of diseases and settings. Longitudinal studies are needed. Useful prevalence data can potentially be extracted from studies of health-related quality of life, general symptom surveys and treatment trials. However, methods to compare results from studies that employ different assessment instruments must be devised. Additional research is needed to elucidate the clinical significance of the fatigue scores obtained using these instruments.

There is sufficient preliminary evidence to support randomized controlled trials of several interventions for cancer-related fatigue, including exercise programs, psychosocial interventions, and stimulant medications. Further laboratory research and observational studies on the physiology of cancer-related fatigue are needed in order to generate rational hypotheses for future intervention trials. Clinical trials for cancer-related fatigue need to utilize appropriate study designs, including the prospective identification of outcomes of interest and sample sizes calculated to provide a reasonable likelihood of detecting those outcomes.

For all of the topics examined in this evidence report, there is a paucity of studies in the pediatric population, and research is urgently needed to address the symptoms of pain, depression, and fatigue in children.

Prevalence of Cancer-related Pain

Cancer has a profound impact around the world because of its prevalence and associated morbidity and mortality (The World Health Organisation Global Programme For Cancer Control, 1993; Wolff, 1997). Poorly controlled pain is key to the impaired quality of life (QOL) that patients with cancer may suffer. Work in the 1970s and 1980s by Bonica (1985), Twycross (1976), Foley (1985), Daut and Cleeland (1982) and Stjernsward (The World Health Organisation Global Programme For Cancer Control, 1993) established the importance of cancer pain control and showed that about three-quarters of patients with advanced cancer experience pain. Surveillance data gathered in developed countries in order to characterize the incidence, prevalence, and in some studies the nature and severity of pain have enrolled about 60,000 patients. This global experience indicates that one-third to one-half of all patients undergoing active cancer treatment experience pain, and that the likelihood of pain is influenced by type of tumor, stage of disease, and extent of metastases (Bonica, 1985; Daut and Cleeland, 1982).

Of the over one million Americans diagnosed annually with cancer, 1500 die daily from this cause (Landis, Bolden, and Wingo, 1999). Nearly 10 million Americans now have or previously had cancer. During the past 25 years, age-adjusted incidence rates pooled across ages, race, and cancer type have increased significantly (about 20 percent) in the United States (Landis, Bolden, and Wingo, 1999; National Cancer Institute, 1999). In parallel with a global trend throughout developed countries, the population of the United States has on average grown older, and numbers of elderly have risen (U.S. Bureau of the Census, 1999). Although cancer is the second leading cause of death (after accidents) in children younger than 14 years in the United States, 5-year survival rates have improved substantially for many childhood cancers since the 1970s (Landis, Bolden, and Wingo, 1999). Unfortunately, despite calls for assigning a high priority to alleviating cancer pain in children (Schechter, Berde, and Yaster, 1993), many children who die of cancer have substantial suffering at the end of life (Wolfe, Grier, Klar, et al., 2000). The incidence and prevalence and hence impact of cancer increases dramatically with age. The cancer incidence rate is less than 50 cases per 100,000 in those under 25, and rises steadily with age to over 200 cases per 100,000 in the 40 to 44 age group. Pooled data for all ages from birth to 54 yield an incidence of just over 100 cases per 100,000; this rate increases 10-fold to over 1000 cases per 100,000 in the 55 to 64 age group, and further doubles in those over 65 to greater than 2000 cases per 100,000 (National Cancer Institute, 1999). Correspondingly, the cumulative percentage of the US population experiencing invasive cancers during their lifetime increases sharply from below 2 percent in the 0 to 39 age group, to just under 10 percent in the 40 to 59 age group, to about 30 percent in those older than 60 (Landis, Bolden, and Wingo, 1999; National Cancer Institute, 1999).

The aging and overall growth of the population plus the increased incidence and prevalence of cancer in those over 60 ensure that the national disease burden of cancer will increase at least in the near term, and that this burden will continue to fall disproportionately on the elderly (Ferrell BR and Ferrell BA, 1996).

Assessment of Cancer-related Pain

Patients with cancer may experience acute or chronic pain related to their primary diagnosis, from treatment, or from unrelated, even pre-existing disorders. Because of the multiple, often changing origins of pain and the importance of its control, the Joint Commission for the Accreditation of Healthcare Organizations requires that pain be assessed initially and serially thereafter during each clinical encounter in patients at risk for undertreated pain (2000). The subjective nature of pain requires patient participation in its assessment. Therefore the mere act of probing this aspect of a patient's personal, internal experience validates that experience (Morris DB, 1998) and demonstrates a patient-centered point of view (Gerteis, Edgman-Levitan, Daley, et al., 1999). A comprehensive approach to pain assessment in cancer care extends beyond nociceptive evaluation to encompass comorbid medical and psychosocial problems, the meaning and impact of pain upon the patient and significant others, and its effect upon quality of life (The World Health Organisation Global Programme For Cancer Control, 1993).

Initial evaluation of any patient with cancer-related pain is expected to include the essentials of any symptom history (location, intensity, quality, temporal characteristics, exacerbating and relieving factors, and responses to prior treatments), together with psychosocial assessment, physical examination, and appropriate diagnostic studies (Jacox, Carr, Payne, et al., 1994). Psychosocial assessment addresses the mood of the patient, his or her coping skills, family support structure, signs and symptoms of anxiety or depression, and expectations regarding pain management. Initial evaluation seeks to establish a pathophysiological mechanism for each pain, if possible as a recognized syndrome with well-described key features, natural history, and treatments of choice (Caraceni, Portenoy, and a Working Group of the IASP Task Force on Cancer Pain, 1999; Portenoy and Lesage, 1999). The determination that pain is primarily neuropathic or nociceptive helps guide initial selection of drug or non-drug therapy such as surgery or radiation therapy. Neuropathic pain can result from many potential disorders such as deafferentation, mono- or polyneuropathies, or a complex regional pain syndrome (Woolf and Mannion, 1999). Nociceptive pain may be due to somatic or visceral pathology (Cervero and Laird JMA, 1999).

Failure to assess cancer pain intensity serially using a standard, validated scale makes it difficult to judge the effectiveness of any analgesic regimen, or to compare one regimen with another (Jadad-Bechara AR, 1994; Max, 1996; McQuay and Moore, 1998). The 0-10 visual or verbal analog scales, or variants such as a thermometer, are validated, easy to administer, and widely used. Unless the patient is asked to rate pain at its lowest, average or highest intensity during a specific time interval, analog scale measurements provide only an instantaneous "snapshot" of pain intensity that is an incomplete picture of pain with activity, across days or weeks, or that prevents sleep at night. The Brief Pain Inventory (Daut and Cleeland, 1982) therefore captures data on pain intensity across tie as well as at the moment of completion of this instrument. Recognition that pain may impair other dimensions of quality of life and functionality has prompted inclusion of health-related quality of life (HRQOL) measures into clinical analgesic trials, and some clinics to monitor HRQOL during routine care.

Treatment of Cancer-related Pain

Undertreatment of cancer pain occurs over a third of the time even in wealthy, industrialized nations (Cleeland, Gonin, Hatfield, et al., 1994). The basis for such undertreatment is multifactorial. Inadequacy of clinicians' knowledge of effective pain assessment and management, negative attitudes of patients and clinicians toward the use of drugs for pain relief, regulatory issues that promote healthcare providers' concerns about regulatory scrutiny and unwarranted investigation (Joranson et al., 1998; Joranson et al., 2002) and problems of cost and reimbursement for effective pain management (Cleeland, 1987) each contribute.

Present day practical therapeutic options overlap substantially with those applied to treat noncancer-related pain. A generic approach to pain management involves beginning an NSAID with or without an adjuvant, adding a weak opioid for persistent or unresponsive pain, then exchanging the weak opioid for a strong opioid such as morphine. In this context, adjuvant refers to medications that treat concurrent symptoms that exacerbate pain, such as nausea or insomnia; augments or treats a side effect of opioid analgesia; or can diminish specific types of pain such as neuropathic pain (e.g., an anticonvulsant or a tricyclic antidepressant). This approach, developed and disseminated by the World Health Organization (WHO) as a three-step staircase or ladder whose therapeutic escalation is driven by persistent or unresponsive pain, is widely applied to treat cancer-related and noncancer pain in primary through tertiary care practice settings (Carr, Jacox, Chapman, et al., 1992). Because cancer pain provokes biopsychosocial responses in addition to purely nociceptive reactions, the effective management of cancer pain normally requires a multidisciplinary approach that includes behavioral interventions (similar to the approach to rehabilitation of patients with chronic pain not due to cancer). The "low-tech" WHO approach to pain control relies primarily upon the oral route of drug administration but requires no essential modification when alternate routes such as rectal, sublingual, subcutaneous, topical or transdermal are used for drug delivery (Bruera, Brenneis, Michaud, et al., 1988). Whether or not they suffer from cancer, patients whose pain fails to respond adequately to the WHO method may be offered, in settings with suitable infrastructures, more invasive interventions such as intravenous or intrathecal drug administration. Additional measures to control cancer pain that are rarely if ever used to treat noncancer pain include the use of biphosphonates, radionuclides, and chemical or surgical neurolysis.

Prevalence of Cancer-related Depression

It is estimated that 20 to 25% of all cancer patients will experience depression during the course of their illness (Bottomley, 1998). People with cancer are three times more likely than the general population and almost twice as likely than other medically hospitalized patients to develop depression (Arolt, Fein, Driessen, et al., 1998; Kessler, McGonagle, Zhao, et al., 1994). The prevalence of depression in cancer patients is even higher in those with the greatest disability and distressing physical symptoms, especially uncontrolled pain.

There is great variation in the reported frequencies of depression in cancer populations. These varying estimates of depression in cancer patients may result from differing definitions of "depression," differing assessment tools, and different cancer populations with different significant variables such as timing of the assessment, physical debilitation, and concurrent treatment. Many of the retrieved studies do not employ the criteria for major depression described in the American Psychiatric Association's Diagnostic and Statistical Manual for Mental Disorders and appear instead to be measuring depressive symptoms (American Psychiatric Association, 1994).

Assessment of Cancer-related Depression

Given the seriousness of its impact, it is important for caregivers to recognize and treat depression. Past studies have shown that oncologists and primary care providers have difficulty recognizing depressive symptoms in cancer patients (Newell, Sanson-Fisher, Girgis, et al., 1998; Passik, Dugan, McDonald, et al., 1998). Major depression is a clinical entity with specific signs, symptoms, and treatments. It is more than just sadness.

"Depression" in comparison to pain or fatigue can be a set of symptoms or clinical syndromes. Depressive symptoms are present in several psychiatric disorders, with the most common ones in cancer patients being major depressive disorder, adjustment disorder, and depression secondary to a medical condition. Depressive symptoms can also be present in the absence of a psychiatric disorder.

Depression is more complicated and difficult to distinguish from symptoms of the underlying disease in cancer patients. The psychiatric diagnosis of major depressive disorder is defined by a set of criteria in the American Psychiatric Association's Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) (American Psychiatric Association, 1994). The DSM defines depression as having at least five of the following symptoms for at least 2 weeks: 1) depressed mood most of the day; 2) loss of interest or pleasure; 3) change in appetite and/or change in weight; 4) insomnia or hypersomnia; 5) psychomotor retardation or agitation; 6) loss of energy; 7) feelings of worthlessness or guilt; 8) poor concentration; and 9) thoughts of death or suicidal ideation. In order to meet the criteria for major depressive disorder, one of the patient's symptoms must be either depressed mood or loss of interest/pleasure, and the individual must also be experiencing distress or impairment in social, occupational, or other areas of functioning. Major depressive disorder is usually distinguished from an adjustment disorder by the degree, duration, or amount of symptoms.

Making the diagnosis of depression can be a challenge with the cancer patient. Many neurovegetative symptoms of depression -- especially loss of energy, loss of appetite, and sleep disturbance -- overlap with common symptoms of cancer or other medical illnesses, and with the side effects of medical treatments used in cancer patients.

Because these symptom-based diagnostic criteria may not be specific for depression in the context of medical illness, a set of psychological criteria has been suggested in their place. Endicott suggested substituting the psychological symptoms of self-pity, brooding, crying spells, and pessimism for the neurovegetative symptoms that overlap with cancer (Endicott, 1984). Some clinicians highlight the importance of the cognitive symptoms of depression, such as depressed thoughts, hopelessness about all aspects of their lives, guilt or worthlessness, and suicidal ideation.

Additionally, many medical factors in patients with cancer can contribute to, exacerbate, or even mimic depressive symptoms. These include uncontrolled pain, hypercalcemia, anemia, endocrinologic abnormalities, cancer involvement of the central nervous system, glucocorticoids, interferon, and some other chemotherapeutic agents (Roth and Holland, 1994).

The diagnosis of major depression is further entangled with patients' reactions to being diagnosed with cancer. Holland and Massie have described the natural history of emotions following the diagnosis of cancer (Massie, Speigel, Lederberg, et al., 1995). When confronted with a diagnosis of cancer, most patients experience a period of shock and disbelief. The second phase that follows is characterized by more visible distress: sadness, depressed mood, insomnia, anxiety, anorexia, and irritability that can last up to 2 weeks. During this time there is a sense of sadness and uncertainty about the future, and patients are often preoccupied with thoughts of illness, death, and losses. All of these feelings are within the normal limits of reacting at the time and do not constitute an episode of depression. However, some patients may have difficulty carrying out daily activities and have trouble concentration and processing information. Within a few weeks, however, most patients have adapted to their diagnosis, their depressive symptoms resolve, and treatment begins with some optimism and hope for the future.

Although clinical evaluation specifically for psychiatric symptoms is generally thought to be the best assessment, several instruments, both self-report and clinician-administered are being used to assess depressive symptoms. These instruments include three main types: standard psychiatric assessments (such as the Beck Depression Inventory or Hamilton Depression Inventory); instruments designed to assess symptoms in a medically ill population (such as the Hospital Anxiety and Depression Scale and the Brief Symptom Inventory); and rapid assessment instruments (such as The Distress Thermometer). Some of these instruments are currently being used in clinical settings to screen for depressive symptoms.

Treatment of Cancer-related Depression

Recognizing the importance of psychological symptoms that accompany cancer, the National Comprehensive Cancer Network (NCCN) published its consensus guidelines for the management of psychosocial distress in cancer patients (National Comprehensive Cancer Network, 1999). Depressive symptoms fall under the umbrella of psychosocial distress and are thought to be common in patients with cancer.

The NCCN guidelines contain three main components for management: screening, assessment, and treatment. Part of the assessment process is determining an appropriate treatment or referral. Depending on the severity of symptoms and the presence of suicidal ideation, the NCCN has algorithms for treatment.

As with depression in non-cancer populations, treatment for depression in cancer patients can take the form of psychosocial interventions and psychotherapy, medications, or alternative treatments. Many times patients receive a combination of these treatments.

Much of the research on treatment for depression in cancer patients has been on psychosocial interventions. Support groups, group therapy, psycho-education, cognitive-behavioral therapy, and individual counseling are all psychosocial interventions that are being used clinically and have been studied. In contrast to non-cancer populations and based on the high prevalence rates of depression in cancer patients, there have been many preventative trials of psychosocial interventions on depressive symptoms.

There has been considerably less research on the effects of psychopharmacologic interventions on depression in cancer patients. This may be due at least in part to some of the special challenges that an oncology sample would pose compared to a standard psychiatric sample in a clinical trial. Holland, Romano, Heiligenstein, et al., (1998) have cited some of these challenges: finding treatment groups that are homogeneous enough to obtain meaningful results, recruitment of large numbers of subjects to achieve adequate power to detect differences, and difficulties with retention and completion, often complicated by medical morbidity and mortality.

Increasing general interest in alternative treatments in both oncology and psychiatry has prompted many clinicians and patients to try these therapies. Although there is less data available on these treatments, research in this area is also increasing.

Prevalence of Cancer-related Fatigue

Reported prevalence rates of fatigue in cancer patients and survivors are extremely variable, reflecting both the heterogeneity of the populations studied, the variety of techniques used to measure fatigue, and the absence of consensus on criteria for the definition of fatigue. There is reason to suspect, however, that many studies systematically underestimate the degree of fatigue experienced by cancer patients, since the most fatigued patients may decline to participate in studies or may be unable to attend academic referral centers where most studies were performed. Thus, there are potential biases involved in the use of small, non-random cohorts in studying the prevalence of cancer-related fatigue. There are only two true population-based studies (Cella, Davis, Breitbart, et al., 2001; Vogelzang, Breitbart, Cella, et al., 1997). They assessed the prevalence of fatigue in telephone surveys of cancer patients and survivors recruited from large, representative samples of households in the United States, rather than in non-random or convenience samples as were used in many other studies.

Each study assessed fatigue over a short time period, for example during a course of radiation therapy and shortly after its completion (Irvine, Vincent, Graydon, et al., 1998), or in survivors at one time point after treatment (Woo, Dibble, Piper, et al., 1998). We were unable to identify longitudinal studies of fatigue reporting its prevalence and patterns over the entire course of illness using uniform methodology.

Assessment of Cancer-related Fatigue

The NCCN has developed practice guidelines for cancer-related fatigue reflecting currently accepted approaches to assessment and treatment (Atkinson, Barsevick, Cella, et al. 2000). That report emphasizes the high prevalence of fatigue at all stages of cancer and the need for screening and assessment of fatigue as an integral part of cancer care. Since fatigue is subjective, patient self-reports are central to its assessment. Information provided by family members can offer a perspective on the behavioral consequences of fatigue. The medical history, physical examination, and laboratory data may also be useful in the assessment of fatigue. The NCCN guideline has recommended the routine use of a quantitative or semi-quantitative screening device, such as a 0-10 numeric scale. For those with moderate or severe fatigue, a focused evaluation is performed to assess the characteristics of the fatigue and to determine whether contributing factors such as pain, emotional distress, anemia, sleep disturbance or endocrine dysfunction are present. If such factors are not identified a more comprehensive evaluation is recommended, including a review of systems, review of medications, assessment of comorbidities, nutritional and metabolic assessment, and assessment of activity.

The assessment of cancer-related fatigue in the context of clinical research has a somewhat different emphasis than the clinical guidelines promulgated by the NCCN. The foci of assessment in fatigue research have been: 1) to quantify the prevalence, pattern and severity of fatigue in a variety of settings 2) to determine factors that predict or correlate with fatigue, including other symptoms, biological parameters (e.g., cytokine or hormone levels), disease and treatment variables, and demographics, and 3) to assess the response of fatigue to interventions. For these purposes, a number of patient self-report instruments have been developed, and an epidemiology of fatigue is beginning to emerge as described in this report. A number of issues remain to be addressed regarding the assessment of cancer-related fatigue, however. These include the clinical significance of measurements of fatigue, the development of objective, measurable physiologic correlates of fatigue, methods for comparing measurements obtained by different fatigue instruments, and the degree to which cancer-related fatigue can be attributed to specific etiologic factors such as the effects of chemotherapy and radiotherapy, anemia, and depression.

Treatment of Cancer-related Fatigue

There is limited evidence to support specific interventions to ameliorate cancer-related fatigue, with the exception of epoetin alfa for the treatment of anemia and its symptoms during chemotherapy (Littlewood, Bajetta, Nortier, et al., 2001). This is perhaps not surprising given the fragmentary understanding of the fundamental physiology of fatigue, and its probable multifactorial basis in most patients. The initial approach advocated by the NCCN panel is therefore to assess regularly and address the common, potentially remediable factors that may contribute to fatigue, such as anemia, sleep disturbances, medication effects, electrolyte abnormalities and hypothyroidism. The degree to which cancer-related fatigue can be attributed to such reversible causes is unknown, but it is fairly clear that there is an enormous burden of fatigue that is not attributable to such specific causes or persists despite appropriate treatment for them. When specific, reversible causes of fatigue cannot be identified, or fatigue fails to respond to appropriate intervention for them, the NCCN panel of experts advocates a more comprehensive assessment, education and counseling, and consideration of a number of interventions. These include non-pharmacologic approaches such as exercise, stress management, and energy conservation, and pharmacologic approaches such as corticosteroids, psychostimulants, and antidepressants. These recommendations were the result of a consensus based on clinical experience, rather than evidence from randomized clinical trials. As reviewed below, a small number of randomized controlled trials have evaluated psychosocial interventions, exercise programs, and other methods of treatment for fatigue. The data on exercise is promising, but in general unless a specific etiology of fatigue (e.g., anemia or depression) can be identified and treated, there is extremely limited evidence from clinical trials to support any interventions for cancer-related fatigue.

Cancer-related Pain

For cancer pain, we applied the same search strategy used in our Management of Cancer Pain evidence report (Goudas, Carr, Bloch et al., 2001) to identify new studies published in the period from December 1998 through June 2001. This methodology is only briefly summarized below since it is already provided in detail in the earlier evidence report. We performed literature search in the MEDLINE^® and CANCERLIT^® databases. Overlapping reports between the MEDLINE^® and CANCERLIT^® databases were excluded from the CANCERLIT^® search. We also searched the Cochrane Controlled Trials Registry and consulted technical experts and examined references of published meta-analyses and selected review articles for additional studies.

Separate literature searches were conducted for oral mucositis and postherpetic neuraligia. We identified published reports of randomized clinical trials on prevention and treatment of oral mucositis by using a search strategy in MEDLINE^® between 1966 and November 2001. The search strategy consisted of the keywords "Stomatitis," "Mouth Mucosa," "Radiation Injuries," "Neoplasms," and "cancer," and was limited to "human and English language" and to "prospective studies" or "randomized" or "random allocation" or "clinical trials" or "double blind method." This search strategy yielded 660 reports from which we selected 114 RCTs and 2 systematic reviews pertinent to the question at hand. We performed supplemental hand searches based on the literature cited in these articles. The supplemental searches added no qualified randomized trial to this report. The abstracts of these reports were screened to select appropriate articles for inclusion in the present synthesis.

We performed a systematic review of the literature aiming to address the questions of prevalence of acute zoster related pain and postherpetic neuralgia and of the relative efficacy of available treatments for herpes for acute zoster pain and postherpetic neuralgia in cancer patients. We identified published randomized clinical trials reporting on acute zoster pain and postherpetic neuralgia following treatment of acute herpes infection. We searched MEDLINE^® between 1966 and November 2001. The search strategy consisted of combinations of the keywords "Neuralgia," "Herpes Zoster," "Pain," "neoplasia," and "Neoplasms," and was limited by the keywords "human," "English language," and "controlled clinical trial." We performed supplemental hand searches based on the literature cited in these articles. The supplemental searches added no qualified randomized trial to this report. The abstracts of these reports were screened to select appropriate articles for inclusion in the present synthesis.

Cancer-related Depression

NLM staff performed the literature search for cancer-related depression articles in November 2001. Several search strategies were evaluated. The initial search was conducted in PUBMED^® and used broad medical subject headings including: neoplasms combined with depression OR depressive disorder OR antidepressive agents. Letters, news, editorials, and non-English citations were eliminated. PsycInfo, CINAHL^®, and BIOSIS^® were also searched using depression and neoplasms as major headings. The initial search strategy yielded over 3,000 citations. The final search strategy that we used limited the retrieval to those citations that had the term depression as a descriptor or in the title. This strategy yielded about 1,000 articles, and the domain expert of this evidence report screened them. Additionally, bibliographies of review articles or chapters were used to identify relevant studies.

Cancer-related Fatigue

NLM staff performed two separate but linked searches in September 2001, one from MEDLINE^® and another from several databases (EMBASE, PsychInfo, BIOSIS^®, NTIS, CINAHL^®, and Allied and Complementary Medicine) to identify English-language articles that dealt with assessment, prevalence, and treatment of fatigue in cancer patients. The searches yielded 1,137 abstracts, and they were screened for relevance to the specific topics. One hundred seventy-six abstracts were selected for retrieval. Screening of these articles resulted in the elimination of almost half, and ultimately 56 papers were judged to be relevant.

Data were subsequently and systematically extracted, and their elements were the following: population and setting of the cancer patients, size of trial, age, range and percentage of male/female, types of cancer studies, scales used to assess the symptoms of fatigue, time points of measurement, the results and conclusions of the authors.

Patient Population Studied

We accepted all studies of patients with a diagnosis of cancer who suffered from pain, depression, or fatigue due to cancer or cancer treatment. We placed no restrictions on the patients' age, gender, ethnicity, level of advancement of the primary disease (staging) or presence of metastases. The conference planning committee was interested in covering the full trajectory of disease, including but not limited to, periods of active treatment and at the end of life.

Cancer-related Pain

In this report, we retrieved studies presenting data on three broad categories of patients:

• Patients with pain resulting from direct tumor involvement, from either local disease or distant metastases, and involving sites such as bone, soft tissue, or neural structures.

• Patients with pain resulting from a therapeutic, diagnostic, or palliative intervention (procedural pain), such as chronic post-mastectomy or lumbar puncture pain.

• Patients with pain resulting from the side effects of anti-tumor treatment, such as acute herpes zoster or postherpetic neuralgia or oral mucositis pain.

We did not review and summarize randomized controlled trials already included in the previously published evidence report on management of cancer pain, or included in published systematic reviews that were retrieved during the present search process. Instead, we summarized results of those systematic reviews we deemed comprehensive. Studies on acute postsurgical pain in patients were excluded. We placed no restriction on article inclusion according to etiology, nature, or mechanism of pain as classified according to any cancer-related pain classification system.

Cancer-related Depression

Because the search for depressive symptoms in cancer produced several thousand citations, a second more selective search was performed. In order to focus the scope, we limited our literature review to studies that specifically assessed depressive symptoms rather than including more general quality of life (QOL) data from every cancer clinical trial. Not only would reviewing all of the QOL studies be beyond the scope of this report, but also there is some evidence that the mental health domains of QOL scales may not be sensitive for clinically significant depressive symptoms in oncology populations. We also did not include studies that compared QOL outcomes between different cancer treatments.

Because "depression" was not limited to major depressive disorder, choices were made regarding the definition of "depression" and the scope of the review. Unlike pain and fatigue, depression can be both a set of symptoms and clinical syndromes. Depressive symptoms are present in several psychiatric disorders with the most common in cancer patient being major depressive disorder, adjustment disorder, and depression secondary to a general medical condition. Because major depressive disorder is the most described in this population, we focused the review on studies of major depressive disorder.

However, limiting the review to major depressive disorder does not capture the prevalence of depressive symptoms in cancer patients, regardless of diagnosis. We also chose to review studies that assessed the presence of depressive symptoms. Because of the numerous instruments used to assess depressive symptoms and psychological distress in people with cancer, the most frequently used instrument was chosen to review in order to allow some comparability of data.

Similarly, the assessment of depression would include the assessment of major depressive disorder as well as depressive symptoms. However, again because of the numerous instruments used to assess psychological distress in people with cancer, we chose to only review papers that directly compared instruments.

The bulk of treatment studies for depression in cancer patients has used psychosocial interventions. Because there have been hundreds of studies and published meta-analyses were identified, we limited our review of these interventions to the meta-analyses. The reviews of treatment studies using psychopharmacologic and alternative interventions were limited to controlled trials.

Cancer-related Fatigue

This report summarized the scientific evidence on the assessment, prevalence, and treatment of fatigue in cancer patients. It was not limited to certain types of malignancy or treatment modalities, but occurred in patients with all types of cancer, from early to advanced stages, receiving chemotherapy, radiation therapy, hormonal therapy, biologic therapy, bone marrow or stem cell transplantation, or combined modality treatments.

There were many studies that included the assessment of one or more cancer-related symptoms, more than we can evaluate in this evidence report. To address the topic of the prevalence of cancer-related fatigue, we accepted studies that assessed fatigue as the primary purpose of the study. We accepted both retrospective and prospective studies. We excluded studies that used general health QOL measurements and also clinical trials that measured fatigue as part of the outcomes. We accepted only randomized controlled trials for the topic of treatment of cancer-related fatigue.

Updates of the Management of Cancer Pain Evidence Report Key Questions

For the updates to the previously published Management of Cancer Pain evidence report, we followed the same format and used the same key question numbers in reporting the new evidence. In general, we grouped studies that met the inclusion criteria according to six broad treatment categories derived from the earlier Management of Cancer Pain Clinical Practice Guideline (Jacox, Carr, Payne, et al., 1994):

• primary pharmacological interventions (opioids, acetaminophen, and NSAIDs, local anesthetics)

• secondary pharmacological interventions or adjuvant analgesics (psychostimulants, alpha-2 agonists, tricyclic antidepressants, etc.)

• nonpharmacological interventions (physical, psychosocial, and educational interventions, e.g., hypnosis, massage, TENS, music, relaxation, and acupuncture)

• nonpharmacological invasive interventions (neuroaugmentation, neurolytic block)

• antineoplastic interventions (radiotherapy, chemotherapy, biphosphonates)

• other various treatments interventions (not under previous categories)

Data from studies addressing the same question were included in the same category of evidence table. Variables that generally apply to any clinical trial (e.g., study design) as well as more specific variables (e.g., therapy for breakthrough pain) that apply only to studies on cancer pain management were considered in selecting variables to be included in the evidence tables (see the Evidence Tables for these variables).

Internal Validity

Internal validity addresses the design, conduct, and reporting of the clinical trial. Some of the items belonging to this entity have been widely used in various "quality" scales and usually include items such as concealment of random allocation, treatment blinding, and handling of dropouts. In this evidence report, we define a four category internal validity scale: A (least bias), B (susceptible to some bias), C (likely to have large bias), I (unable to assess due to lack of reported information).

• A

  Double-blinded, well-concealed randomization, few drop outs, and no (or only minor) reporting problem of the trial that is likely to cause significant bias.

• B

  Single-blinded only, unclear concealment of randomization, or has some inconsistency in the reporting of the trial but is unlikely to result in major bias.

• C

  Unblinded study, inadequate concealment of random allocation, high dropout rate, or has substantial inconsistencies in the reporting of the trial such that it may result in large bias.

• I

  Inadequately reported (very often trials do not report certain data; this may occur by intent or due to oversight.)

Applicability

Applicability, also known as generalizability or external validity, addresses the issue of whether the evidence from the study population is sufficiently broad as to be able to generalize to the population at large. Individual studies are often unable to achieve broad applicability due to restricted study population characteristics and a small number of study subjects (Lau, Ioannidis, Schmid, 1997). We define the applicability grade as below:

• A

  Patients enrolled in the trial represent a broad spectrum of the population (high degree of applicability). Typically this would be a large study, although a large study in itself does not guarantee a high degree of generalizability.

• B

  The study included only a narrow/restricted study population, but the result is relevant to similar types of patient population (restricted applicability). Typically this would be a small study, but may also be a large study of a very homogeneous population.

• C

  Studied outlier population that is not immediately relevant to the study question (very limited direct applicability or not applicable), or where the study reported only limited information.

• I

  Not reported or insufficient information to assess external validity issues (uncertain applicability).

Because the efficacy of pain treatments may depend on the baseline level of pain, we also extracted data on baseline pain intensity of the study population to assist in the interpretation of results. We report in the evidence grading tables, along with the applicability rating, the baseline pain intensity expressed as VAS (visual analog scale) of 0-10cm (or 0-100mm) when this data is reported in the study. Studies that did not provide 0-10cm VAS data but reported qualitative descriptions or other scale are so noted in the tables.

Magnitude of the Treatment Effect of Cancer-related Pain Studies

In each of the result tables, "effect size" reflects the difference between outcomes in the treatment arms of the study, not pre- versus post comparisons in the experimental group. For example if an experimental opioid were compared with morphine, and both treatments were found to have a large effect upon pain scores, then the effect size assigned to this study would be a "±".

The following effect size scale is employed for studies that provide consistent reporting of a pain-related outcome:

It should be noted that large difference in effect does not necessarily imply a statistically significant difference.

The outcomes reported by available studies on some of the questions were heterogeneous and were not amenable to categorizing the effect size on the same scale. This group of heterogeneous outcomes includes drug consumption, pain relief, and QOL-related indices. These studies were evaluated by pain management experts and assigned a qualitative score for the effect size.

Prevalence of Cancer-related Pain

Prevalence of Cancer-related Depression

Because "depression" was not limited to major depressive disorder, choices were made regarding the definition of "depression" and the scope of the review. Unlike pain and fatigue, depression can be both a set of symptoms and clinical syndromes. Depressive symptoms are present in several psychiatric disorders with the most common in cancer patient being major depressive disorder, adjustment disorder, and depression secondary to a general medical condition. Depressive symptoms may also be present in the absence of a psychiatric diagnosis. Because major depressive disorder is the most described in this population, we first focused the review on studies of major depressive disorder.

However, limiting the review to major depressive disorder does not capture the prevalence of depressive symptoms in cancer patients, regardless of diagnosis. We also reviewed studies that assessed the presence of clinically significant depressive symptoms.

Prevalence of Cancer-related Fatigue

Our search strategy identified 27 studies in which a defined endpoint of the research was a quantitative estimation of the prevalence of cancer-related fatigue in a specified target population. Thirteen studies included patients with a variety of cancers. Five specifically focused on breast cancer, four on lung cancer, two on prostate cancer, and one each on Hodgkin's disease and rectal cancer. We did not include fatigue prevalence rates from studies of general health-related quality of life, symptom surveys, or treatment trials unless fatigue assessment was specified as an endpoint of the study. In addition to the tabulated studies generated by our search, we have reviewed a number of studies that, although they do not report a specific fatigue prevalence, focus on the pattern of fatigue in cancer and the various disease, treatment, and patient-related factors that correlate with it.

Patterns and Correlates of Fatigue

Table 5. Prevalence of Fatigue in Cancer Patients

Table 5. Prevalence of Fatigue in Cancer Patients

Author Year UI	N	Population/Setting	Mean Age (Range)/% Male	Cancer Type	Prevalence
King 1985 85242295 USA	96	During and post-XRT	(26-83) 52% M	Chest, head and neck, GU, GYN,	65-93% during XRT, 14-46% @ 3 months (% reported for each anatomic site)
Hurny 1993 94207627 Switzerland	127	Chemo trial	ND Gender ND	SCLC	43% moderate or severe at baseline, 30-37% during chemo
Donnelly 1995 95271387 USA	743	palliative care service	(61-70) 53% M	Various cancers	48% "clinically important" (moderate or severe)
Hickok 1996 97089233 USA	50	Radiation therapy	63 (37-78) avg. 68% M	Lung cancer patients	78% experienced fatigue at some point during XRT
Longman 1996 97158314 USA	307	Patients on chemo, hormonal therapy or XRT	55 (25-82) 0% M	Breast cancer, stage I-IV,	83%; 60.2% "problematic"
Richardson 1997 98155331 UK	129	During chemo	58 (26-82) 44% M	Various	89% at some point during chemo
Sarna 1997 97165457 USA	60		58.3 (33-80) 0% M	Advanced lung cancer	56.7% had "serious" fatigue (>3 on 1-5 scale)
Vogelzang 1997 97397931 USA	419	Patients who had received chemo or XRT	65 33% M	Various cancers	78% reported fatigue during their disease and treatment, 32% on daily basis
Smets 1998 98435611 Netherlands	250	Ambulatory patients receiving XRT with curative intent	64±13 59% M	Various cancers	During XRT 40% were tired most of the time, 33% sometimes, 27% hardly ever. 44% were more fatigued after than before XRT, 26% were less fatigued, 30% no change
Smets 1998 98435610 Netherlands	154	Patients in remission after XRT	65±12 57% M	Various cancers	51% recalled fatigue in first 3 months after XRT (19% very much, 32% moderate). No significant differences in fatigue scores between cases and controls at 9 months
Gaston-Johansson 1999 20152209 USA	127	Patients after surgery and chemotherapy, before autologous stem cell or bone marrow transplant	45±7.6 0% M	Stage II, III & IV breast cancer	91% had fatigue on VAS
Jacobsen 1999 20004863 USA	54 cases 54 controls	Patients receiving adjuvant chemotherapy	51±10 0% M	Breast cancer	4% of patients had severe fatigue before cycle 1, 28% before cycle 4 (MSAS). Patients had significantly more fatigue than controls at all time points
Loge 1999 99385422 Norway	459 cases 2214 controls	Patients after curative treatment: 38% XRT, 14% chemo, 47% XRT+chemo	44±12 55% M	Hodgkin's Disease	26% of Hodgkin's survivors were fatigue cases (total dichotomized score > 4 and symptom duration of >6 months) vs. 9% of male and 12% of female controls
Miaskowski 1999 99283638 USA	24	Outpatient XRT for bone metastases	56.6±13 50% M	Various cancers	79% had moderate or severe fatigue at bedtime and 48% on awakening
Monga 1999 99334561 USA	36	XRT	66.9 (55-79) 100% M	Localized prostate cancer	8% were fatigued (>6 on PFS) prior to XRT, 25% at completion of XRT
Stone 1999 99202777 UK	95 cases 98 controls	Palliative care units, no chemo or XRT in > 4 weeks	67 (30-89) 43% M	Patients with advanced cancer	75% had severe fatigue (> 95th percentile of controls on FSS)
Bower 2000 20139478 USA	1957	Breast Cancer Survivors 1-5 years after diagnosis	55 0% M	Breast cancer	35% classified as fatigued (scores in disability/limit-ation range on RAND survey)
Curt 2000 20497163 USA (same as Cella 2001)	379	Patients post- chemo or XRT	53 21% M	Breast cancer (62% of patients) and various other	76% had fatigue at least a few days per month during most recent chemo, 30 % had daily fatigue
Okuyama 2000 21408236 Japan	134	Post-surgery patients (77% mastectomy, 23% breast-conserving) 28.1% had had chemo, 8.9% XRT	55.1±10.3 0% M	Breast cancer patients stage 0-III,	56% perceived themselves as fatigued per the CFS.
Servaes 2000 21023870 Netherlands	85 comparison group 16 chronic fatigue	Patients disease-free at a mean of 2.9 years after treatment	47.5±14 56% M	Various cancers and treatments,	29% had heightened and 19% severe fatigue (>27 or >35 on CIS)
Stone 2000 20314191 UK	62	Patients receiving hormonal therapy	69 (55-80) 100% M	Prostate cancer, various stages	14% had "severe fatigue" at baseline, 17% at 3 months (NS). (severe fatigue defined as > 95th percentile on FSS in controls without cancer)
Stone 2000 20363241 UK	98	Patients receiving inpatient palliative care	66 (30-89) 56% M	Early breast or prostate cancer, inoperable lung cancer, or advanced cancer	48% of cases had "severe fatigue" (defined as >95th percentile of control group scores)
Stone 2000 20489733 UK	576	Patients attending three regional cancer centers over a 30 day period	59 (18-89) 37% M	Various cancers and stages	58% reported being "somewhat" or "very much" fatigued
Cella 2001 21348064 USA (same as Curt 2000)	379	Patients post- chemo or chemo + XRT	53 21% M	Various cancers (50% breast)	17% met proposed criteria for cancer-related fatigue; 37% reported > 2 weeks of fatigue in preceding month
Given 2001 21291233 USA	841		(>65) 55% M	Breast, colon, lung, prostate	26-33% had fatigue at 4 time points over 1 year
Okuyama 2001 21408236 Japan	157	ambulatory patients with advanced lung cancer, no surgery, chemo or XRT in past 4 weeks	63.1 (27-80) 71% M	Advanced lung cancer	51.3% had clinical fatigue, defined as interfering with at least one domain of daily life
Wang 2001 21481486 USA	72	Patients receiving pre-op chemo & XRT	56±11 50% M	Locally-advanced rectal cancer	At baseline 26% had moderate and 18% severe fatigue; at end of treatment 28% had moderate & 31% severe fatigue

A number of additional studies (not included in Table 5) have examined the pattern or correlates of fatigue. Some of the key findings from this literature are reviewed.

Not surprisingly, fatigue has been found to correlate with impairments in HQL in patients receiving radiation therapy (Irvine, Vincent, Graydon, et al., 1998; Lovely, Miaskowski, and Dodd, 1999), chemotherapy (Redeker, Lev, and Ruggiero, 2000) and in long-term survivors (Bower, Ganz, Desmond, et al., 2000).

Several studies examined putative biological correlates of fatigue, with generally disappointing results. In patients with lung cancer undergoing radiation therapy, significant weight loss was observed, but neither weight loss nor a laboratory marker of impaired nutritional status (prealbumin) correlated significantly with fatigue (Beach, Siebeneck, Buderer, et al., 2001). Another group with high fatigue levels, patients who have undergone autologous bone marrow transplants for lymphomas, were assessed for endocrine and immunologic abnormalities. Although gonadal dysfunction was common, it was not associated with greater fatigue; likewise, there was no correlation between fatigue and serum levels of inflammatory cytokines (interleukin-6, tumor necrosis factor, soluble tumor necrosis factor receptor) (Knobel, Loge, Nordoy, et al., 2000). There was also an absence of correlation between fatigue and mild leydig cell dysfunction in survivors of various hematologic malignancies (Howell, Radford, Smets, et al., 2000). Greenberg, Gray, Mannix, et al. (1993) reported that serum interleukin-1 levels rose between weeks one and four in men receiving radiation therapy for prostate cancer, but no statistical correlation was possible.

Cancer-related fatigue has been associated with psychosocial and demographic factors, other symptoms, and disease and treatment variables. A common theme in many studies is an association between psychological distress, in particular, depression and fatigue.

In breast cancer survivors, the variance in fatigue has been examined as a function of disease and treatment variables, symptoms and demographics. Bower, Ganz, Desmond, et al. (2000) examined correlates of fatigue in a large case-control study and found that the type of adjuvant treatment (chemotherapy, radiation therapy, or both) did not predict fatigue levels. The only significant predictors were the current symptoms of depression and pain. Similarly, Broeckel, Jacobsen, Horton, et al. (1998) found that demographic, disease and treatment variables were not significantly correlated with fatigue after adjuvant chemotherapy. Again, current symptoms and conditions were correlated with fatigue, in this case poor sleep, menopausal symptoms, catastrophizing as a coping mechanism, and psychiatric disorders. A third study (Okuyama, Akechi, Kugaya, et al., 2000) found the same pattern in breast cancer survivors: fatigue was significantly correlated with current symptoms of dyspnea, insufficient sleep and depression, but not disease or treatment variables. On the other hand, Mast (1998) and Woo, Dibble, Piper, et al. (1998) found associations between prior chemotherapy and fatigue.

Evidence that fatigue may be related to psychological problems and other symptoms was presented in numerous other contexts. Irvine, Vincent, Graydon, et al. (1998) found correlations between symptom distress, psychological distress, and fatigue but not between disease variables and fatigue in 121 women receiving radiation therapy for breast cancer. Redecker, Lev and Ruggiero (2000) studied fatigue, psychological variables and HQL in 263 patients undergoing chemotherapy. Their findings suggest that fatigue is closely tied to psychological factors, particularly depression. Akechi, Kugaya, Okamura, et al. (1999) examined the correlates of fatigue in 455 ambulatory cancer patients. Cancer site and performance status did not predict fatigue. Aside from demographic variables, depression was the only factor correlating with fatigue. Hann, Garovoy, Finkelstein, et al. (1999) analyzed fatigue in 31 patients undergoing autologous stem cell transplants for breast cancer. There were no associations between fatigue and demographics, disease variables, or the transplant regimen. The factors that were associated with fatigue were time to engraftment, length of hospitalization, depressive symptoms and anxiety. In a study of 457 Hodgkin's disease survivors, Loge, Abrahamsen, Ekeberg et al. (2000) found that anxiety was predictive of chronic fatigue. Bruera, Brenneis, Michaud et al. (1989) found that asthenia correlated with depression but not with nutritional status, lean body mass, tumor mass, anemia, or type of treatment in 64 patients with advanced breast cancer.

A common theme in these studies is the correlation of cancer-related fatigue with depression or, more generally, psychological symptoms. A second observation is that, in some contexts, current physical and psychological symptoms are more significant than disease or treatments variables as predictors of fatigue. This appears to be the case particularly in studies of cancer survivors.

Assessment of Pain

The field of pain assessment is a well-developed one. Originating in analgesic trails before the middle of the last century, it was brought into focus by Beecher's 1957 monograph on the measurement of subjective phenomena in man. Melzack, Turk and many other colleagues from the behavioral sciences contributed to the subsequent refinement of this field, that in recent decades has had an interface with the equally large and thriving discipline of quality of life assessment (Spilker, 1996). Every monograph on cancer pain and all general texts on pain assessment and management describe a comprehensive approach to pain assessment as integral to cancer pain control. Such assessment is uniformly depicted as following the same approach as the general medical assessment of any symptom: taking a detailed history, that includes biopsychosocial dimensions; asking about pain location, quality, frequency, severity, and relieving or exacerbating factors; inquiring as to prior treatments and their effectiveness; and physical examination targeted towards defining the etiology and mechanism of pain. For example, the Brief Pain Inventory is a multidimensional assessment instrument widely applied in cancer pain research (Daut 1983). On the other hand, in the retrieved treatment trials, the instruments employed were extremely diverse and the most frequently applied ones were narrowly focussed upon pain intensity alone. Of 21 assessment tools that were employed a minimum of 5 times each, the four most often used were single-point pain intensity scales. The diverse mechanisms of, and quality of patients' pain were largely not reported in the retrieved clinical trials, and the information that was captured was gathered in a group of instruments sufficiently heterogeneous to preclude merging of results. Of 218 retrieved trials, there were 125 distinct outcomes assessed.

Assessment of Depression

The clinical standard of diagnosing major depressive disorder in patients with cancer is a clinical interview. Although DSM criteria for major depressive disorder contain symptoms that overlap with cancer and cancer treatments, the rate of diagnosing major depressive disorder in patients with cancer using substitute criteria is highly correlated to that with DSM Criteria (Kathol, Mutgi, Williams, et al., 1990).

We chose to focus this review on screening instruments for depression.

Assessment of Fatigue

Over the years, many disciplines have been involved in the study of fatigue, including ergonomics, nursing, physical therapy, medicine, psychology, physiology, and biochemistry. Fatigue is the most frequently reported symptom of patients with cancer (Glaus, Crow, and Hammond, 1996). It is the symptom that is reported as the most distressing and causes the greatest amount of interference with daily life (Richardson, 1995). To date, there remains no consensus regarding standard definitions of fatigue, especially cancer-related fatigue. But there has been progress--fatigue is beginning to earn recognition as a valid clinical diagnosis, and in 1998, the International Classification of Disease included criteria for fatigue; in 2000, the National Comprehensive Cancer Network published guidelines for its management (Atkinson, Barsevick, Cella, et al., 2000).

We conducted a search of the medical literature to ascertain clearer definitions of fatigue and to look more carefully at the development and use of instruments that assess fatigue. To this end, NLM staff embarked on two separate but linked searches, one from MEDLINE^® and another from EMBASE and PsychInfo to identify English-language papers, which dealt with assessment, prevalence, and treatment of fatigue in cancer patients. The searches yielded 930 abstracts that were screened for relevance to the specific topics, and there were 176 abstracts related to the assessment of fatigue. More than 100 papers were retrieved and read thoroughly. Preliminary screening of these articles resulted in the elimination of almost half, and ultimately 56 papers were judged to be relevant. Data were subsequently and systematically extracted: population and setting of the cancer patients, size of trial, age, range and percentage of male/female, types of cancer studies, scales used to assess the symptoms of fatigue, timepoints of measurement, the results and conclusions of the authors.

A majority of the papers emanated from the United States and Canada, but many European countries were represented including those in Scandinavia, Holland, Switzerland, England, and France; there are publications from Hong Kong, Japan, Greece, and Australia. The occurrence of fatigue crosses all diagnostic and treatment categories, at all phases of disease, in all segments of populations that contain a wide array of cancer patients. Almost all studies were adult, and one dealt with young children, aged 10-18.

Table 9. Frequency of use, fatigue assessment scales (more...)

Table 9. Frequency of use, fatigue assessment scales

Cancers	1985-89	1990-94	1995	1996	1997	1998	1999	2000	2001*	Total
Breast	1	3	1		1	6	5	5	3	25
Lymphoma	1	2			1	1	2	2	1	10
Gynecological	1	1	1			1		2	3	9
Lung		2			1	2		3	1	9
Colorectal					1	2	2	3		8
GI						1	1	3	1	6
Leukemia					1	1	1	2	1	6
Hodgkin's				1			2	1	1	5
Melanoma		1	1				2	1		5
Prostate			1			2		2		5
Testicular			1				2	1		4
Head and Neck						1	1		1	3
Liver							1	1		2
Bladder	1									1
Oral								1		1
Brain										0
Myeloma										0
Total	4	9	5	1	5	17	19	27	12

Some of the studies compare two scales or multiple instruments; some newly created scales attempt to validate performance or reliability in prospective clinical settings; others correlate their results to tools previously validated by experience. Many publications prospectively measured fatigue across the trajectory of cancer from diagnosis, through treatment, following treatment, and in the palliative setting. Other studies present correlations between severe fatigue and markers such as pulmonary dysfunction, between severe fatigue and depression or dyspnea, severe fatigue and endocrinological status and various tumor necrosis factors, or low hemoglobin levels.

Some of these tools are very specific to cancer-related fatigue, such as Piper Fatigue Scale and the Brief Fatigue Inventory. Other tools are difficult to use in the clinical setting because of their complexity or the length of time required to administer the tool. Reliable, clinically valid tools for measuring fatigue, such as numeric severity of fatigue scales, may be better suited to the clinical setting.

Treatment of Cancer-related Pain

The material that follows updates the evidence published in the evidence report, Management of Cancer Pain (Goudas, Carr, Bloch et al., 2001). The numbering of the questions corresponds to those in the original report.

Target population

Studies with patients undergoing radiation treatment to the head and neck area were eligible. Reports where head and neck cancer patients received neoadjuvant or concurrent chemotherapy were included, whereas trials that evaluated patients with other malignancies receiving systemic therapy were considered ineligible. Similarly, studies where patients were treated with radiation therapy alone, but which included patients with disease at sites other than the head and neck region (for example, head and neck plus lung or esophagus cancers), were also deemed ineligible.

Intervention

Studies were eligible if they compared any intervention to a control group that included no active treatment for oral mucositis, where the intent was to prevent the development or progression of oral mucositis. Studies of agents that were clearly used for palliation, such as analgesics and anesthetic agents, were excluded.

Outcome measures

Studies that reported clinician-assessed oral mucositis scores, proxy measures of oral mucositis such as radiotherapy interruptions or G-tube placements, or patient-assessed ratings of oral mucositis or other symptoms were included.

Type of study

All studies, including Phase II and observational studies, that met the eligibility criteria were included for the purpose of developing the classification scheme, assessing trends in, and possible future directions for, research. However, only randomized trials were included in the analysis from which inferences on effectiveness were drawn.

The authors selected the validated assessment tool developed by Jadad, Moore, Carroll et al. (1996) to assess the quality of the selected studies.

Data processing and transformation

In studies where dichotomous outcome measures were not reported, the data were derived in one of two ways. If individual patient information was available in the report, this was simply abstracted in a dichotomous format. For continuous data, when means were provided but where different scales were used by different researchers, the data were transformed to a common percentage scale, using the method described by Eisenberg, Berkey, Carr et al. (1994). The data were then dichotomized using the technique of Moore, McQuay and Gavaghan (1996), that was tested and found to be robust in the oral mucositis model.

The outcomes of interest were severe oral mucositis as assessed by clinicians and as rated by patients

These are summarized for all studies for which they were available, using individual odds ratio (OR) of response to treatment (test vs. placebo) and associated 95% confidence intervals, for each trial. For this analysis, an OR > 1 favors control, while an OR = 1 indicates exact equivalence between the two groups.

A pooled interval estimate of the population OR was calculated. A test for heterogeneity was done, using the Chi-square test. Significance for this test was set liberally at p 0.1, since in practice the test often lacks the power to detect inter study differences of the treatment effect (Lau, Ioannidis, and Schmid, 1997). The DerSimonian and Laird Random Effects Model of pooling (DerSimonian and Laird, 1986) was used, based on the assumption of the presence of interstudy variability, to provide a more conservative estimate of the true effect.

Subgroup analysis

Several sources of heterogeneity were anticipated. To explore the relationship between treatment effect and study features, several a priori hypotheses regarding heterogeneity were developed and subgroup analyses planned. For each of the outcome measures, an analysis was done for each of the classifications: direct cytoprotectants, indirect cytoprotectants, and antibacterials. Based on the a priori hypothesis that sucralfate, which forms a mechanical barrier, in addition to its other properties as a direct cytoprotectant, might be more effective than the other direct cytoprotectants, a subgroup analysis was planned to assess this. Similarly, it was postulated that antibiotic lozenges, which are thought to selectively eliminate the aerobic Gram-negative flora associated with acute oral mucositis, might be more effective than the broad-spectrum rinses, and so a further analysis was undertaken of these two groups. The impact of radiotherapy dose (50 Gy vs. <50 Gy) and the influence of methodologic quality (score 3 vs. score <3) were analyzed.

Trial characteristics

Fifteen randomized, controlled trials were included in the analysis. One study in progress, a multicenter trial of an antibiotic lozenge currently under way in Canada (NCIC, 1997) was identified.

A total of 1022 patients were included in the 15 studies. Nine studies assessed direct cytoprotectants. Of these, five evaluated the barrier sucralfate, and four evaluated protectants that are thought to stimulate epithelial response (one each of prostaglandin, beta-carotene, hydrogen peroxide, and laser therapy). One assessed indirect cytoprotectants (benzydamine), and five trials considered antibacterials. Of those, three studied broad-spectrum antibacterials, while two evaluated narrow-spectrum antibiotic lozenges.

In nine studies, it was clear that the radiation field covered the oral cavity/oropharynx in all patients, whereas in six reports other head and neck sites such as larynx were also included, or the field was unspecified. Seven of the studies excluded patients who had received prior chemotherapy or radiotherapy.

The radiotherapy dose was at least 50 Gy in 10 trials, while five trials included some patients who had received a total dose of 45 Gy or less. In two studies (Foote, Loprinzi, Frank, et al., 1994; Okuno, Foote, Loprinzi, et al., 1997) with sample sizes of 52 and 112, respectively, the radiotherapy dose ranged from 30-45 Gy. In 13 of the studies, radiotherapy was delivered in a conventional fractionation scheme, while two trials (Carter, Hebert, Smink, et al., 1999; Symonds, McIlroy, Khorrami, et al., 1996) included a hyperfractionation schedule in one stratum. Three trials (Carter, Hebert, Smink, et al., 1999; Mills, 1988; Rahn, Adamietz, Boettcher, et al., 1997) used concurrent chemotherapy as part of the planned treatment.

Methodologic quality

The authors found a median quality score of 3 of a maximum possible score of 5 (range 1-5); the scores for each study can be obtained from the original report. Although all studies stated that they were randomized, only three described the method of randomization. Twelve of the studies were described as double-blind, and the method of double-blinding was clearly appropriate in 11 of these. Eight of the 15 provided a statement on withdrawals and dropouts. Agreement for the quality of studies was modest (Kappa = 0.43).

Outcome: severe oral mucositis, clinician assessments

Thirteen of the 15 trials supplied information on the proportion of patients who developed severe oral mucositis, as assessed by clinicians (Bensadoun, Franquin, Ciais, et al., 1999; Carter, Hebert, Smink, et al., 1999; Cengiz, Ozyar, Ozturk, et al., 1999; Feber, 1996; Foote, Loprinzi, Frank, et al., 1994; Hanson, Marks, Reddy, et al., 1997; Lievens, Haustermans, Van den, et al., 1998; Makkonen, Bostrom, Vilja, et al., 1994; Mills, 1988; Okuno, Foote, Loprinzi, et al., 1997; Rahn, Adamietz, Boettcher, et al., 1997; Spijkervet, van Saene, Panders, et al., 1989; Symonds, McIlroy, Khorrami, et al., 1996). Two trials did not report the data in a manner that could be abstracted (Epstein, Stevenson-Moore, Jackson, et al., 1989; Epstein and Wong, 1994) for this outcome. The use of all interventions together had a significant impact on mucositis, reducing the odds of developing severe oral mucositis by 36% (OR 0.64; 95% CI: 0.46, 0.88).

The baseline risk for developing severe oral mucositis in the control groups of the combined studies was 43%, while the overall absolute risk reduction when prophylactic interventions were used was 9%.

The odds ratio favors antibacterial agents over placebo (OR 0.47; 95% CI: 0.25, 0.92), and within this grouping, the only significant effect is for the narrow-spectrum antibacterials (OR 0.45; 95% CI: 0.23, 0.86). Studies of higher quality (validity score 3) tended to support the use of treatment across all studies (OR 0.68; 95% CI: 0.48, 0.96). An overall treatment effect of the interventions was seen when all patients received a radiotherapy dose of at least 50 Gy; no difference was demonstrated when studies included some patients receiving less than 50 Gy.

Future research

The authors suggest that in other promising areas, where the research has been of an exploratory nature, randomized controlled trials are needed. If such trials demonstrate a benefit from individual agents, comparison studies and studies of combinations of agents from different classes will be beneficial. In addition, the difference between clinician-rated and patient-rated measures needs to be acknowledged. In studying patient-assessed outcomes in head and neck radiotherapy, it seems that symptom-specific scales need to be complemented or replaced with the use of multidimensional quality of life measures. In planning future research strategies, the choice of clinically relevant primary outcome measures using validated measurement tools, in larger, methodologically sound trials, is essential.

What are the effects of medications on depression in cancer patients?

Only eleven controlled studies of the effects of medications on depressive symptoms in cancer patients exist. Nine of them are primarily treatment studies on depressive symptoms. One is a pain study that also assessed depressive symptoms and one is a depression prevention study.

These studies reflect the history of psychopharmacology. Antipsychotics became available first and then tricyclic antidepressants. Later selective serotonin reuptake inhibitors appeared.

The first study took place in 1972 and was a 6-week placebo controlled trial of thioridazine 25 mg tid for depression in a hetergeneous sample of 50 cancer patients. Thioridazine is an antipsychotic medication that is now not usually used for the clinical treatment of depression. The study included inpatient, outpatient, and terminal patients with various cancers. Depression was assessed by physician ratings of depressive symptoms. Although it appeared better than placebo for depressive symptoms at the end of the first week, this difference was not statistically significant at weeks three and six. However, at all time points, it was significantly better than placebo for insomnia and crying spells. The authors reported that no side effects were observed. Because this study did not clearly chose participants with a diagnosis of major depressive disorder and did not use standardized rating instruments, it is difficult to fully interpret this data (Johnston, 1972).

Purohit and colleagues conducted a 4-week placebo controlled trial of imipramine in 39 hospitalized cancer patients receiving radiation therapy (Purohit, Navlakha, Modi, et al., 1978). All patients staerted with a physician diagnosis of major depressive disorder. The imipramine was dosed between 25 and 50 mg a day. The doses were adjusted for tolerability. Although they demonstrated that 80% of the imipramine group improved compared to 42% of the controls using the Hamilton Depression Scale, they did not analyze their data for statistical significance of these differences.

The effects of a glucocorticoid were studied by Bruera and colleagues (Bruera et al., 1985). This 33 day randomized, placebo-controlled trial used methylprednisolone 16 mg bid in 40 terminal oncology patients with various cancers. Although at day 13, the methylprednisolone showed greater improvement in the Hamilton Depression Scale than placebo (p<0.05), there was no significant difference at day 33. Side effects included 5% of patients reporting increased anxiety and 5% of patients developed Cushingoid features.

Costa and colleagues conducted a 4-week randomized controlled trial of mianserin in 73 women with cancer (Costa, Mogos, and Toma, 1985). Women required significant depressive symptoms as measured by the Hamilton depression Scale to enter the study. Mianserin was dosed between 30-60 mg per day. At four weeks, there were more responders in the miasnserin group assessed by changes in CGI (p<0.25) and the mianserin group had a greater improvement in the Hamilton Depression Scale (p<0.01). There were no significant differences in side effects.

Bruera and colleagues studied the effects of another medication, mazindol, in cancer patients (Bruera, Carraro, Roca, et al., 1986). Twenty-six terminal patients with various cancers participated in this 12-day randomized placebo-controlled trial. The mazindol was dosed at 1 mg tid. At 12 days, there was no significant difference in changes in the Hamilton Depression Scale between groups, but the mazindol group experienced "serious toxicity." These side effects included nervousness, sweating, delirium, and weakness.

Holland and colleagues, comparing alprazolam with progressive muscle relaxation (Holland, Morrow, Schmale, et al., 1991), did the only non-blind trial included in this series. One hundred forty-seven people with various cancers participated in this 10-day trial. Patients were included if they met criteria for depression or anxiety. Both groups showed improvement in the Hamilton Depression Scale and the Affects Balance Scale. However, the alprazolam group had greater improvement in the Affects Balance Scale (p<0.04) and an improvement trend with the Hamilton Depression Scale (p<0.08). There were more drop outs in the alprazolam arm and side effects reported were drowsiness, sedation, and lightheadedness.

Mianserin was also studied by Van Heeringen and colleagues (van Heeringen and Zivkov, 1996). In this 6-week randomized, placebo-controlled trial in fifty-five women with breast cancer receiving radiation, depression was diagnosed with a DSM-III interview. Mianserin was dosed at 60 mg per day. Hamilton Depression scale scores were significantly lower in the mianserin group compared to placebo at week 2 (p<0.056), week 4 (p<0.004), and week 6 (p<0.004). The number of responders was greater in the mianserin group compared to placebo at weeks 4 and 6 (p<0.05). Although placebo participants tended to terminate the study earlier, the some of the mianserin group reported postural symptoms and sedation.

Although depression was not the primary endpoint, Eija and colleagues assessed depression in their 4-week, randomized placebo-controlled trial of amitriptyline for neuropathic pain in fifteen breast cancer patients (Eija, Tiina, and Pertti, 1996). Amitriptyline was dosed at 25-100 mg per day. No significant differences were noted in depression between groups, but depression was not assessed in a standardized way. Patients were asked two questions regarding depression with a four-point scale. Side effects reported from amitriptyline included sedation, dry mouth, constipation, and sweating.

Razavi and colleagues conducted a 4-week randomized, placebo-controlled study of fluoxetine after one week of placebo in both groups (Razavi, Allilaire, Smith, et al., 1996). Ninety-one people with various cancers and DSM-III diagnoses of major depressive disorder participated. Fluoxetine was dosed at 20 mg per day. No significant differences in the numbers of responders as defined by HADS < 8 or changes in depression scores with the HADS and MADRS were found between groups. There were also no significant differences in side effects between groups.

Fluoxetine was also studied by Holland and colleagues in a 6-week comparison study with desipramine, after one week of both groups receiving placebo (Holland et al., 1998). Thirty-seven women with cancer diagnosed with depression by DSM-II-R interview participated. Although doses were variable with responses, fluoxetine was dose at 20 mg per day and desipramine was dose at 100 mg per day. Both groups improved significantly by both the Hamilton Depression scale and the CGI. However, there were no significant differences between groups. Both groups reported side effects that included nausea, dry mouth, insomnia, and dyspepsia.

Although major depressive disorder was not the focus, Razavi and colleagues studied the effects of trazodone and clorazepate on depressive symptoms in a 4-week comparison trial (Razavi, Kormoss, Collard, et al., 1999). Twenty-seven breast cancer patients diagnosed with adjustment disorder with depressed mood by DSM-II-R criteria participated. Trazodone was dosed between 50-150 mg per day and clorazepate was dosed between 10-30 mg per day. Although there were a greater number of responders by CGI in the trazodone group, this difference was not statistically significant. There were also no significant differences in change in depression scores with the HADS.

The last study was not a depression treatment study. Musselman and colleagues studied paroxetine in preventing depression in melanoma patients receiving interferon (Musselman, Lawson, Gumnick, et al., 2001). This 14-week trial had 40 participants. Paroxetine was initially dosed at 10 mg per day and increased up to 40 mg per day. Paroxetine significantly reduced the incidence of depression (p=0.04) and the severity of depressive symptoms (p<0.001) as measured by the Hamilton depression Scale. Adverse events did not differ significantly between groups.

With the exception of two studies, all medications classified as antidepressants showed benefit. The two studies that did not show benefit studied the medications for not more than four weeks. This is not surprising because antidepressants can typically take between four to six weeks to take effect. Medications other than antidepressants did not appear to be effective.

Are psychosocial interventions effective in treating depressive symptoms in cancer patients?

There have been hundreds of studies on the effectiveness of psychosocial interventions on depressive symptoms in cancer patients. Because of the large number of studies, we limited our review to published meta-analyses of these studies.

Although these meta-analyses were not done necessarily on patients with significant depressive symptoms, there does appear to be a small to moderate effect size from these treatments. Though one of these meta-analyses did not note a significant difference in effect size among different types of treatments, the limitations of that study make interpretations of that observation difficult.

Despite its title as a meta-analysis of psychoeducational care, Devine and Westlake, (1995) is actually a meta-analysis of psychosocial interventions in adult cancer patients. It included 98 studies with 5,326 subjects published from 1976 to 1993; 47% were published in a journal or book, 45% doctoral dissertations, and 6% theses published in a journal. Inclusion criteria were provision of a psychosocial intervention to adults with cancer; use of an experimental, quasiexperimental, or pre-post single test design; and outcome measures of physical and emotional well being. Exclusion criteria were studies that had comparison arms to other treatments (such as medications); studies with less than five subjects, and all treatment groups not being from the same setting. Interventions included educational, behavioral/cognitive counseling, non-behavioral/cognitive counseling. The most prevalent intervention was behavioral/cognitive counseling. It is not noted whether both individual and group interventions were included. Although the studies were not necessarily on patients with depression, a positive effect was present in 92% of the studies with the average effect size being medium.

Meyer and Mark (1995) is a meta-analysis of psychosocial interventions in adult cancer patients. It included 45 studies with 2,840 subjects. Its inclusion criteria were published randomized experiments; psychosocial intervention compared to control or minimal intervention; and the inclusion of behavioral and emotional outcome measures. The only exclusion criterion was hospice or terminal care studies. Interventions included educational, behavioral, non-behavioral counseling, social support or other interventions (music therapy, for example). It is not noted if both individual and group interventions were noted. Although this meta-analysis showed a small effect size, it was not as stringent in evaluating depressive symptoms. Measures of emotional adjustment were included rather than measures of depression. This meta-analysis also did not show a significant difference in effect size according to type of intervention.

Sheard and Maguire (1999) is a meta-analsis of psychological interventions for anxiety and depression in cancer patients. It included 20 studies with 1,101 subjects. Inclusion criteria were studies of psychosocial interventions for psychodistress in cancer patients; a control condition; and published in English in a journal or indexed as a dissertation. The one exclusion criterion was a single group design without a control. In the evidence-based table, only the effects of studies that assessed for depressive symptoms were included. Both individual and group data are included in the analysis. The interventions included individual therapy, relaxation, group therapy, group therapy excluding psychoeducation, and group psychoeducation. Although these studies were not specifically done on patients who were depressed, a small to medium effect size was seen, but the effect size decreased with the authors' assessment of the quality of the study.

Are alternative treatments effective for the treatment of depressive symptoms in cancer patients?

Although there have been descriptive reports of alternative or complementary treatments for depression in people with cancer, there have been no controlled trials.

Cancer-related Pain

Pain is an important component of the already considerable disease burden of cancer. Minorities, women, and the elderly are particularly at risk for cancer-related pain, and this observation possibly reflects both underassessment and undertreatment. Nearly all epidemiological studies that contain population-based estimates of the prevalence and severity of cancer pain report few details concerning specific mechanisms of cancer pain, nor do they track pain and other symptoms longitudinally across time; however, credible survey information is emerging that suggests the prevalence of pain and other symptoms is related to the type of cancer (though not related to information such as grade or stage). The total number of patients surveyed in published reports of the prevalence and severity of cancer-related pain, as well as in methodologically sound trials of cancer pain relief, is a minor fraction of those receiving care -- much lower than for nearly all other high-impact, costly conditions. There remains an incomplete picture of variations in cancer pain prevalence, severity, and course with respect to: patient factors (age, gender, ethnicity, race, culture); disease characteristics (type, grade, stage as well as other features such as genotypic or phenotypic classification); the setting in which care is provided (provider attributes and professional qualifications, location and nature of healthcare practice); and specific treatments directed toward the underlying disease and its associated pain.

Cancer-related Depression

Major depressive disorder and depressive symptoms occur frequently in patients with cancer. Despite using standardized measures, there is a wide range of reported incidence and prevalence. From our review of the literature, the prevalence rates appear to be between 10 to 25 percent for major depressive disorder and a similar range for clinically significant depressive symptoms regardless of psychiatric diagnosis. Given that the estimated point prevalence of major depressive disorder is 2.2 percent in the general population, these rates in cancer patients may be at least four times greater.

Despite using standardized measures, there is a wide range of reported incidence and prevalence. This range may be the result of several factors that include the timing of the assessment, concurrent treatment, medical morbidity and pain, and age. Cancer patients are a heterogeneous population with different sociodemographics, cancer types, treatments, and responses to treatment. More accurate estimates might be obtained in studying the rates in more homogeneous sub-groups.

Reports on the incidence of depressive symptoms are limited to prospective studies of patients receiving certain cancer treatments. Reports of incidence range widely from about 2 to 17 percent during the time frame of the studies. Although it is difficult to generalize these results to all cancer patients, the incidence appears to be somewhere between 2 and 7 percent per year. However, these studies, like the prevalence studies, face the same difficulties of heterogeneous populations, and there are too few naturalistic studies that follow patients from the time of diagnosis conducting serial depressive assessments. Despite the lack of high quality data, it appears that the incidence of depression in cancer patients is much higher than the general population's yearly aggregate incidences: 0.25 to 0.6 percent for women and 0.08 to 0.2 percent for men.

Cancer-related Fatigue

Estimations of fatigue prevalence have been performed in many settings but the data is by no means comprehensive. Our search identified 27 reports on fatigue prevalence. Thirteen studies included patients with a variety of cancers, five studies addressed breast cancer at various stages and during various treatments, and four focused on lung cancer and two on prostate cancer. Many types of cancer were not specifically addressed. A more complete picture of cancer-related fatigue could be obtained by extracting fatigue prevalence data from HQL and general symptom studies, and from clinical trials in which fatigue is reported as a side effect of treatment.

A very broad range of prevalence rates has been reported. Prevalence rates ranged from 4 percent in breast cancer prior to starting chemotherapy and 8 percent in prostate cancer prior to radiation therapy (Jacobsen, Hann, Azzarello, et al., 1999; Monga, Kerrigan, Thornby, et al., 1999), to 91 percent in breast cancer patients after surgery and chemotherapy and before bone marrow transplantation (Gaston-Johansson, Fall-Dickson, Bakos, et al., 1999). Findings of significant concern were the prevalence rates of fatigue in cancer survivors: 26 percent in Hodgkin's disease survivors (Loge, Abrahamsen, Ekeberg, et al., 1999); 35 to 56 percent in breast cancer survivors (Bower, Ganz, Desmond, et al., 2000; Okuyama, Akechi, Kugaya, et al, 2000); and 48 percent in a cohort treated for cancers (Servaes, van der Werf, Prins, et al., 2000).

Comparisons of the prevalence rates in these studies are problematic, however, since each study used different criteria for defining the presence or absence of fatigue and its severity. A few studies used criteria for fatigue that were based on normative data from control populations (Stone, Hardy, Broadley, et al., 1999; Bower, Ganz, Desmond, et al., 2000). In many studies, however, the criteria for fatigue were arbitrary.

Population-based surveys of fatigue prevalence (Vogelzang, Breitbart, Cella, et al., 1997; Cella, Davis, Breitbart, et al., 2000) represent an advance since they potentially avoid the selection bias inherent in small cohort studies, although other bias may be introduced by the fact that only a minority of potential subjects were contacted. The primary purpose of the study by Cella, Davis, Breitbart, et al. (2001) was to evaluate proposed ICD-10 criteria for cancer-related fatigue, which, if adopted, would allow comparison of data from a wide range of sources.

The period over which the prevalence of fatigue was assessed in any group of patients was short, generally confined to the period of treatment and immediately after, or one time point in studies of survivors. No studies used a uniform methodology to track the time course of fatigue longitudinally. Few studies on the prevalence of fatigue included data on factors that might contribute to fatigue (such as anemia, infections, etc.), or attempted to determine to what extent fatigue was due to treatment, disease, or other factors.

Cancer-related Pain

One hundred eighty-four unique trials were retrieved for the prior evidence report, plus an additional 34 in the updated search, for a total of 218 trials. In these trials, a total of 125 different outcome measures were employed. Many were similar to each other but not directly combinable, e.g., four-point and five-point pain intensity scales. The most frequently employed scales were of pain intensity: the 0-100 VAS (58 uses), a 10 cm VAS (44 uses), and five-point (26 uses) or four-point (24 uses) pain intensity scales. Of the 21 instruments that were employed five or more times in the pooled retrieved studies, four were numerical measurements of pain intensity and one was an integrated pain intensity score, three were pain relief scales, two were peak or summed pain intensity differences between treatments, two were global evaluations or global efficacy of intervention scales, and one was the McGill-Melzack pain questionnaire. Only one of the 21 most frequently applied instruments (a four-point scale) appraised side effects (5 uses); analgesic consumption was assessed 21 times.

The need for detailed assessment conducted within a psychosocial framework is addressed in virtually all guidelines or monographs on cancer pain management. A voluminous literature exists that describes the multidimensional, experiential nature of cancer pain and links poor control of cancer pain to impaired quality of life, including functionality. Expectations for detailed, multidimensional assessment of cancer pain, including quality of life assessment, during cancer care contrast with minimalist assessments of pain intensity presented during relatively brief observation intervals reported in nearly all of the retrieved trials. Side effects limit analgesic dosage and hence impede pain control in many patients, yet only one of the 16 most widely employed outcomes measures addresses the issue of side effects; that one is a coarse, four-point measure.

Cancer-related Depression

The clinical interview, using DSM criteria, is the standard of care for diagnosing major depressive disorder and other depressive syndromes in people with cancer.

Because depression may go undetected and untreated in oncology practices, the importance of screening tools and rapid assessments has been emphasized. Many self-report assessments are available that could be completed by patients before visits in clinic waiting rooms. These assessments range from circling answers to series of questions to The Distress Thermometer, a modified visual analogue scale that is presented in the NCCN guidelines for the treatment of psychosocial distress. Available data on the sensitivity, specificity, predictive values, and cross-correlations are presented in the evidence-based table.

Although these assessment tools may have been validated in studies, there is currently no evidence on how widely they are used clinically or to suggest that they affect clinical care and outcomes.

Cancer-related Fatigue

The literature on fatigue assessment focuses on tools that are used in research studies, and to a much lesser extent on methods of assessment for clinical use. The NCCN practice guidelines on cancer-related fatigue recommended the use of simple 0-10 numerical self-report scales or verbal scales (e.g., mild, moderate or severe) to assess the severity of fatigue in practice settings. This recommendation is based on a study by Piper, Dodd, Ream, et al. (1999) correlating self-report scores on a 0-10 scale of fatigue with the Physical Functioning Subscale of the Medical Outcomes Study Short Form 36 (SF-36). They found a significant impairment in physical function on the SF-36 when fatigue scores were >7. If moderate or severe fatigue is reported, the NCCN panel then recommended a focused history and physical examination and evaluation of the pattern of fatigue, associated symptoms, and interference with normal functioning. Five potential causes should be addressed because they are common and potentially reversible: pain, emotional distress, sleep disturbance, anemia, and hypothyroidism. A review of medications was not included in this initial evaluation although in practice this is certainly done. If no etiology of fatigue is identified, the NCCN panel recommended a more comprehensive evaluation, education and counseling, and consideration of a variety of strategies to reduce fatigue. Although this approach is not based on controlled clinical trials validating its effectiveness, it is intuitively reasonable. It would be useful to know the extent to which causes of fatigue can be identified and reversed using this algorithm.

In the context of clinical research, assessment of fatigue involves the use of patient self-assessment tools of varying levels of complexity. Most studies in the last several years have used instruments that assess multiple dimensions of fatigue and have been tested for validity, consistency, and reliability. Issues still remain in terms of the clinical interpretation of the scores obtained on these instruments, and the comparison of fatigue measurements obtained using different instruments.

Cancer-related Pain

As reported in the prior evidence report, the number of randomized controlled trials to evaluate analgesic interventions for cancer pain is approximately one percent of the total of initial titles retrieved. The heterogeneity of existing trials precludes meta-analysis to answer most clinically relevant questions. For example, the trials do not disclose differences between the relative efficacy of opioids and nonsteroidal antiinflammatory drugs (NSAIDs) administered by various routes to patients with mild, moderate or severe cancer pain. There is evidence of an opioid dose-sparing effect from co-administration of an NSAID but no consistent reduction in side effects from doing so. There is little evidence for significant differences in analgesic efficacy between NSAIDs (only one trial out of 18 reported this outcome). The studies comparing different NSAIDs could not be combined in a meta-analysis due to between-study heterogeneity in the outcomes assessed, drug doses and schedules compared, and study duration. Trials that compare the efficacy of NSAIDs versus "weak" opioids (i.e., opioids commonly prescribed for mild to moderate pain) reveal no difference in analgesic efficacy between these two classes of agents, even when the latter are co-administered with the same NSAID tested in the other study arm. These trials enroll relatively small numbers of patients and are of relatively short duration. A single randomized controlled trial of breakthrough pain treatment demonstrated the superiority of oral transmucosal fentanyl citrate to placebo. We found no randomized controlled trials addressing analgesic efficacy and safety of NSAIDs selective for the cyclooxygenase-2 isoenzyme in treating cancer pain.

Published trials exploring different routes of administration for NSAID or opioid drug classes show no difference in analgesic efficacy between oral or rectal routes of delivery of each class of drug. Limited data from trials show no benefit from a purely analgesic standpoint of parenteral (intramuscular or intravenous) administration over enteral administration. Failure to prove overall superiority of one route over another does not diminish the value of a particular route in a specific clinical situation (for example, the use of suppositories or transdermal administration in the presence of dysphagia). Insufficient information exists to permit comment upon differences in patient preference for specific routes and administration, or acceptability of side effects. There was no evidence to indicate improved analgesia with controlled-release oral formulations versus immediate-release formulations or transdermal delivery. The benefit of less frequent doses upon patient compliance is a possible advantage of controlled-release formulations.

Trials of biphosphonate therapy are heterogeneous with respect to inclusion criteria, concomitant medical and radiotherapeutic treatments, disease categories, dosage regimens, choice of agent, and duration of follow-up. Differences in pain assessment methods are also considerable, ranging from analgesic consumption to a "requirement" for palliative radiation therapy. However, many studies showed a positive effect, some showed no effect, and no study showed a detrimental effect of biphosphonate therapy on skeletal symptoms of metastatic disease or myeloma. Positive effects were less evident in patients who received concurrent hormones or chemotherapy that might by themselves have a favorable effect on bone symptoms. The evidence in aggregate suggests that biphosphonates are effective in reducing bone pain in patients with cancer, although this benefit may be less marked when such therapy is combined with other tumor-directed therapy. The radionuclide strontium-89 was more effective than placebo (inactive strontium), and equally effective as external radiation.

The literature on the effects of various chemotherapeutic and hormone therapy regimens on pain is also quite heterogeneous, with differing inclusion criteria and therapeutic regimens. Analgesic consumption is reported in a minority of these studies. In only one chemotherapy trial, and in no hormonal therapy trial, did pain outcomes differ significantly between treatment arms.

Eighteen trials compared fractional dosing schedules of external radiotherapy for pain from bony metastases. No trial found more than a transient difference in pain between fractionation schedules, although external radiation as a whole is effective in decreasing pain. Meta-analysis was precluded by heterogeneity of the dosing schedules, variability in the anatomic sites and fields treated, and outcomes assessed. Short courses (even single doses) of palliative treatment with higher doses appear to produce results similar to those of longer courses that deliver a lower dose per treatment. The minimal total dose of radiation to provide pain relief has not yet been determined. A possible detrimental effect upon quality of life due to transient skin irritation and discoloration at the site of post-mastectomy chest radiation was reported in one study (Whelan, Levine, Julian, et al., 2000).

The number of studies of physical or psychological treatments to decrease cancer-related pain is small, and variability of the specific intervention and type of pain addressed precludes any broad conclusions. Studies of education evaluated different interventions in patients, medical staff, and the community at large. Studies of hypnosis in the pediatric and adult age groups indicate benefit for procedural and mucositis-related pain. Cognitive behavioral treatments may also be helpful. More studies are needed, with larger numbers of patients and control groups.

Sufficient randomized controlled trials on neurolytic celiac plexus block (NCPB) for pain relief in pancreatic and other visceral cancers were identified to indicate the efficacy of this modality. NCPB lowered pain scores or produced a prolonged dosage-sparing effect upon analgesic drug requirement. The scarcity of randomized or controlled trials on the efficacy of spinally administered opioids or other agents led us to retrieve nonrandomized reports in an effort to estimate the efficacy of this modality. These additional reports, although positive, were case series without control groups and hence yielded no data on relative efficacy of the spinal versus systemic route of drug administration. Similarly, the efficacies of ablative neurosurgical interventions such as cordotomy or rhizotomy were addressed only in case series. There were no trials that addressed the efficacy of acupuncture.

Cancer-related Depression

The current evidence shows that interventions are beneficial for depressive symptoms in cancer patients. There appears to be a clear benefit of psychosocial interventions, although the magnitude of the effect size seems to be in the mild to moderate range. Because of the hundreds of studies on psychosocial interventions in cancer patients, we limited our review to published meta-analyses of these studies. However, in limiting our review in this way, the contribution of effects from preventive studies and depression treatment studies could not be separated. The effects of psychosocial interventions may vary in these two different kinds of studies.

Although not all pharmacologic studies showed benefit for depression in cancer patients, all studies that used antidepressants and conformed to usual practices for antidepressant trials did show benefit. Since antidepressants typically can take four to six weeks for their full effect, studies of antidepressants under five weeks tended to show less benefit. Currently, there is efficacy data for selective serotonin reuptake inhibitors (SSRIs) and tricyclic antidepressants. Another antidepressant that showed benefit, mianserin, is an atypical antidepressant not available in the United States. Although trazodone, an atypical antidepressant, showed some benefit in treating depressive symptoms, it is not commonly used as an antidepressant because it often causes sedation at therapeutic doses.

No controlled studies on alternative treatments for depression in cancer patients were identified.

Cancer-related Fatigue

Ten randomized, controlled trials were identified that assess interventions for cancer-related fatigue. Four involved psychosocial interventions (support groups, psychotherapy, relaxation therapy, and a comprehensive coping strategy program). Three clinical trials evaluated the impact of exercise on fatigue. One trial involved massage therapy and one evaluated a "bedside wellness system" using virtual reality technology. There was only one trial of pharmacotherapy (epoetin alfa for fatigue related to anemia in patients receiving chemotherapy). Some of the concerns regarding the methodology of these studies are addressed in the results section. Endpoints were often poorly defined and sample size calculations absent. It is probable that many of these studies were underpowered to detect the outcome of interest. Although several studies reported statistically significant associations between the intervention being tested and various outcomes, the absence of prospectively defined endpoints renders these results difficult to interpret.

The studies by Mock, Hassey Dow, Meares, et al. (1997) provides evidence that exercise may be helpful in reducing or preventing fatigue in patients receiving radiation therapy for early stage breast cancer. The beneficial effects of exercise are less clear in other contexts such as peripheral blood stem cell transplantation (Dimeo, Stieglitz, Novelli-Fischer, et al., 1999). Nonetheless, this is certainly an approach that warrants further investigation.

The randomized, placebo-controlled, double-blind trial by Littlewood, Bajetta, Nortier, et al. (2001) suggests a substantial benefit associated with epoetin alfa in terms of quality of life, fatigue, and hematologic parameters in anemic patients undergoing chemotherapy. The findings of this study are supported by large, non-randomized trials.

Cancer-related Pain

Comprehensive, credible surveillance data describe the incidence and prevalence of cancer in the United States and survival rates for each major neoplastic disorder. The evidence that undertreated pain adds substantially to the disease burden of cancer is equally credible but less extensive and detailed. Limited cross-sectional data link tumor type and stage with pain quality or intensity (Caraceni, Portenoy, and the IASP Working Group, 1999), and there is no corresponding longitudinal, tumor-specific data on individual pain trajectories during treatment of cancer and cancer-related pain. Tumor- and population-specific data of this nature are needed if the natural history of cancer pain and its relief are to be understood with sufficient precision to advise individual patients and their families on the selection of pain control options. Data of this sort is available, for example, to facilitate informed choices by patients, families, and their care providers as to the likely progression, appearance of complications of, and risks and benefits of therapies for many prevalent conditions such as hypertension, heart disease, diabetes mellitus, infectious disease, and (as mentioned above) cancer itself.

Cancer-related Depression

There is much variance in the literature on reports of rates of both major depressive disorder and depressive symptoms in cancer patients. Even when standardized instruments are used, wide variance is seen. One recommendation would be to conduct prevalence studies that examine reasons for such variance and contributing factors for such differing rates. The timing of measurements of depressive symptoms appears to be important and may contribute to the variance. Since depression is more common in women in the general population, the effect of gender on depression in cancer patients needs to be more carefully studies. One goal may be to develop a statistical model that could predict the rate of depressive symptoms given the cancer, treatment, and demographics of the population.

The existing incidence studies of depression in patients with cancer all start at some time after the diagnosis of cancer. It is recommended that more studies start at the time of diagnosis, or even before if possible, to get better estimates of the incidence of depression once people are diagnosed with cancer. Both studies of incidence and prevalence should assess past histories of depression.

Research on the prevalence and incidence of depression in children with cancer is scarce and must be enlarged.

Cancer-related Fatigue

Studies of the prevalence of cancer-related fatigue are by no means comprehensive. Additional studies of fatigue prevalence in many types of cancer and many different clinical contexts are needed.

Data on fatigue prevalence are reported in studies on health-related quality of life, general studies of symptom prevalence, and from treatment trials in which fatigue is reported as a side effect. Extracting fatigue prevalence data from these sources could potentially provide a much more detailed and complete picture of the scope and impact of cancer-related fatigue. Methods for comparing prevalence rates obtained using different assessment tools would need to be devised.

We were unable to identify studies that provided a longitudinal assessment of fatigue prevalence. All existing studies were confined to a brief period associated with treatment or a limited number of time points during the palliative phase or in cancer survivors. Thus, while data exists, for example, on the prevalence of fatigue in women with breast cancer during chemotherapy, radiation therapy, and in long-term follow-up, these studies use heterogeneous assessment techniques and do not provide insight into the time course of fatigue. There is a need for studies of the prevalence of fatigue over the entire course of treatment and follow up using uniform methods of assessment.

Cancer-related Pain

The present sophistication of quality-of-life assessment, and documentation of relationships between pain, disability, and impairment, mandate that clinical trials of cancer pain control incorporate outcomes measures beyond mere pain intensity. Textbooks, guidelines and review articles on cancer pain all describe the need to perform a detailed history so as to assess the psychosocial and cultural frameworks and individual factors underlying pain experience and complaint. Such monographs further recount the importance of formulating whenever possible a mechanism-based pain diagnosis so as to guide therapy, the quantification of pain intensity as well as distress and impairment of quality of life and function associated with pain, and the reassessment of pain and distress across the continuum of care. Indeed, assessment of pain intensity and appropriate treatment are already clinical standards of the Joint Commission on Accreditation of Healthcare Organizations. To accomplish pain assessment and treatment related to clinical trials as well as clinical care in an increasingly diverse society creates a need for developmentally, culturally, and gender-appropriate instruments that are sensitive, reliable, and easy to administer. Instruments to assess health-related quality of life, particularly functional status, have been applied in recent years during cancer treatment trials. Analgesic trials for the most part have omitted such instruments, and those that incorporated them did so in a heterogeneous fashion. Because breakthrough pain contributes to overall pain morbidity in many patients, pain intensity itself is normally assessed in clinical practice not only at a single time but also across the course of illness and most importantly, during activity that includes movement. Uni-dimensional, "dip-stick" measures of pain intensity at rest, repeated a few times during a short observation interval and often after only a single dose of medication, must be supplanted by more clinically relevant and comprehensive pain assessment during clinical trials.

Important information to strengthen interpretation of the results of clinical trials is at present mostly lacking. Such information, to permit responder or non-responder subgroup analyses, includes genetic and genders characteristics, medical comorbidity and concurrent treatments, as well as other individual features such as ethnicity and culture.

Because side effects of therapy often limit doses of analgesic agents employed, and contribute to global morbidity associated with cancer pain, the prospective assessment of side effects should be accomplished with the same care as has been bestowed upon assessment of analgesia. To do so will require development and validation of age and culture-appropriate scales for individual side effects that may appear during therapy.

Cancer-related Depression

There are many screening and assessment instruments currently available for depression in cancer patients. Researchers can choose instruments based on ease of use versus the instrument performance. The performance of brief instruments that appear to have promising predictive value, such as the one-item, "Are you depressed?," require further study and replication. The development of short instruments that assess all three symptoms (depression, pain, and fatigue) could be one possible area of future research.

Although some of these instruments are currently being used in clinical practices, there are currently no published studies of their effect on outcome. Outcome research, both psychological and medial, needs to be done using the instruments as "lab tests."

More research needs to be done on the assessment of depression in children with cancer.

Cancer-related Fatigue

Further research is needed on methods for the clinical evaluation of fatigue. A useful strategy would involve studies to validate the screening and assessment algorithms recommended in the NCCN practice guidelines on cancer-related fatigue. These would include studies on the utility of brief fatigue screening tools such as visual or verbal analogue scales in different clinical settings, and on the clinical application of more elaborate assessment instruments. Further research is also needed on the prevalence of treatable factors contributing to fatigue, and the reversibility of fatigue based on the treatment of these underlying factors.

In the research context, there should be a continued reliance on assessment instruments that are well characterized in terms of their validity, consistency and reliability. Clinical interpretation of the outcomes of these measures still remains problematic. Further comparisons are needed between data on fatigue in cancer patients and normative data from control groups.

Given the large number of instruments used to assess fatigue, methods to compare results from studies using different instruments are needed. Researchers in this field should consider developing consensus criteria for the assessment of fatigue for use in future studies.

Cancer-related Pain

This is an exciting time in translational research on cancer-related pain, because preclinical research has generated novel insights into distinctive mechanisms through which tumors produce pain. Increased understanding of the pathogenesis of cancer-related pain (as well as other symptoms such as depression and fatigue) may foster therapies that are increasingly targeted, i.e., mechanism-based. At present, however, in nearly every respect (number of trials, sample size, representative study populations, and study design), the quality of the scientific evidence on cancer pain treatment compares unfavorably with that for cancer treatment. Cancer treatment trials for the most part exemplify mechanism-based clinical research. Leading investigators in the area of cancer pain relief trials have repeatedly called for improving the quality of trials in this area. This goal cannot be achieved merely by incorporating standardized pain assessment and health-related quality-of-life measurements into cancer treatment trials. Although such a strategy is laudable, data so gathered cannot be generalized to the treatment of pain during intervals of stable disease, or to patients who are in remission but who continue to experience residual pain. Carefully designed trials with pain or pain relief as a primary outcome are required in diverse populations with well-defined disease. These groups include patients with stable disease, others with treatment-induced, incident, or breakthrough pain, and those with specific pain syndromes (such as postmastectomy syndrome) during disease remission.

Standards for cancer pain treatment trials must adhere to those for clinical trials in general, as expected by editors of most leading medical journals. Trials of cancer pain relief should enroll more patients and follow them longer than has generally been customary in the past; apply blinding and active placebos, when appropriate, or uniform control treatments otherwise; employ adequate between-arm washout intervals and consider advancing disease state in crossover trials; and assess side effects, pain mechanisms, and rest, incident, or breakthrough pain in a standardized, combinable fashion. To these criteria must be added the need to study gender, race, age, ethnicity and culture with greater precision than in the past, to avoid overgeneralization of results. Categorizing patients by tumor type and stage, and by mechanism of pain, with inclusion criteria that yield homogeneous groups within individual studies, appears to offer the best chance of translating preclinical advances into improved clinical analgesia. Pilot studies that indicate gender and ethnic differences in analgesic pharmacokinetics and pharmacodynamics merit larger scale follow-up. Small-scale, short-term randomized controlled trials that establish treatment efficacy for purposes of Food and Drug Administration approval are not designed to prove effectiveness in larger scale, long-term applications in the treatment of cancer pain relief. To meet this need, outcomes research can provide valuable data that are not feasible to acquire through controlled trials.

Just as combination chemotherapy is employed to treat many forms of malignancy, the practice of analgesia commonly involves drug co-administration. Methods are needed to synthesize published evidence on drug interactions and to apply and extend existing methods (now employed in acute pain studies) to characterize such interactions during long-term cancer pain treatment. Related to this area is the important issue of developing clinical evidence on optimal strategies for the sequence of drug therapies employed for cancer pain control (that is, the WHO model or other treatment algorithms), and the optimal means to combine drug and non-drug therapies. Trials to address these issues, like those to evaluate one component of a multidrug antitumor regimen, are effort intensive and may require large numbers of subjects per treatment arm.

Data on individual variation in preferences for, responses to, and costs of drug and non-drug interventions are limited. For example, the spinal route of analgesia is widely employed but much remains to be learned about optimal patient selection, the comparative efficacy of spinal drug infusion versus systemic drug administration, and the selection of initial or secondary agents or combinations. Drug interactions during long-term cancer pain treatment require clarification. A host of complementary therapies are now employed, but with little rigorous testing of their efficacy. It is unclear whether a mechanism-based approach to diagnosing and relieving each component of pain in an individual is more effective than an empiric regimen in which each patient's treatment is based upon pain intensity alone. Another key unanswered question is how to optimally combine drug with non-drug therapies given that the latter are safe and inexpensive. Despite the importance of pediatric cancer pain control practically no analgesic drug trials focus on children.

Cancer-related Depression

Psychopharmacologic, psychosocial, and alternative interventions offer some benefit on treatment for depressive symptoms with cancer patients. Research needs to be done to support current clinical practices in the prescribing of medications for depression in cancer patients. Because antidepressant medications have been shown to be effective, the newer antidepressants, especially the atypical ones, should be studied in this population. There should also be more trials of the use of antidepressants for the prevention of depressive symptoms in patients with cancer.

Hundreds of studies exist on psychosocial interventions for cancer patients and depression, but a meta-analysis specifically of treatment studies on depressed patients remains to be done. This will probably change the effect sizes estimated in the meta-analyses that included large numbers of prevention trials and may better differentiate between the effectiveness of types of psychosocial interventions.

To evaluate their efficacy, controlled trials on alternative therapies for depression in cancer patients should be done. Studies on the treatment of depression in children with cancer are needed.

Cancer-related Fatigue

The absence of treatment options for cancer-related fatigue reflects a lack of understanding of its fundamental physiology and the very small number of controlled clinical trials that have addressed this problem. Further basic research is needed, including the development of animal models to study the role of cytokines, nutritional factors, muscle wasting, and other potential contributors to fatigue. Studies correlating such factors with fatigue in cancer patients are also needed in order to develop rational hypotheses for treatment trials.

Several promising strategies for treatment of fatigue have been identified based on preliminary clinical trials or clinical experience. These strategies require further investigation in randomized controlled trials. Among the more promising treatment approaches are exercise programs, psychosocial interventions (with a particular focus on the detection and treatment of depressive symptoms), and stimulant medications. Research on fatigue in other contexts may provide leads for effective therapies in cancer patients. For example, Breitbart, Rosenfeld, Kaim, et al. (2001) recently reported improvement in fatigue associated with the psychostimulants methylphenidate and pemoline in a randomized, placebo-controlled trial in patients with HIV disease.

Other potential treatment approaches that warrant preliminary laboratory investigations and/or pilot trials include hormonal treatments such as human growth hormone, melatonin, and androgens, as well as anti-inflammatory medications, and dietary interventions.

Future clinical trials for cancer-related fatigue should use appropriate study designs, including prospectively defined endpoints. They should be appropriately powered to detect differences in the endpoints of interest. Strategies are needed to identify and eliminate barriers to effective diagnosis and treatment of fatigue and other cancer symptoms, and to enhance accrual to studies of these symptoms.