Chapter  64:  S-Adenosyl-L-Methionine for Treatment of Depression, Osteoarthritis, and Liver Disease

Search Strategy

Twenty-five biomedical databases were searched through year 2000: MEDLINE®, HealthSTAR, EMBASE, BIOSIS Previews®, MANTIS™, Allied and Complementary Medicine, Cochrane™ Library, CAB HEALTH, BIOBASE, SciSearch®, PsychINFO, Mental Health Abstracts, Health News Daily, PASCAL, TGG Health & Wellness DB, and several pharmaceutical databases. The researchers searched using the term SAMe and its many pharmacological synonyms, the three focus disease states, study design, and article type. They also searched the bibliographies of review and meta-analysis articles and questioned experts to identify additional citations. An additional 62 articles were identified from these sources, particularly from review articles and from citations suggested by the advisors.

Selection Criteria

Reports were included in the synthesis of evidence if they focused on SAMe for one of the selected diseases and presented the results of randomized clinical trials on human subjects. Language of publication was not a barrier to inclusion. About 25 percent of the selected studies were in foreign languages, mainly Italian.

Data Collection and Analysis

All selected titles, abstracts, and articles, in all languages, were reviewed independently by two reviewers who were fluent in the appropriate language, and all disagreements were resolved by consensus. Information was collected about patient demographics, disease state, intervention, study design, and outcomes. Sufficient numbers of homogeneous studies existed to permit a meta-analysis of the efficacy of SAMe for treatment of four conditions: depression versus placebo and active (pharmacological) therapy, osteoarthritis versus placebo and active (pharmacological) therapy, cholestasis of pregnancy versus placebo and active therapy, and intrahepatic cholestasis associated with liver disease versus placebo. The remainder of the liver disease studies were too heterogeneous for pooled analysis and were assessed qualitatively.

Pharmacology And Pharmacokinetics Of SAMe

SAMe functions primarily as a methyl donor. A wide variety of molecules accept SAMe's methyl groups; naturally occurring recipients include proteins, phospholipids, DNA, RNA, hormones, and the energy-transporting amino acid creatine (Lieber, Casini, DeCarli, et al., 1990). Methyl transfer is essential to the development of the phospholipid bilayer that makes up the outer membranes of normal cells and contributes to membrane fluidity. SAMe-mediated transmethylation is also critical for the formation of neurotransmitters in the central nervous system (Bottiglieri, Hyland, and Reynolds, 1994). Hepatic transmethylation is also important: 85 percent of total body transmethylation occurs in the liver (Lieber, Casini, DeCarli, et al., 1990).

After donating its methyl group, SAMe is converted to s-adenosylhomocysteine, which initiates the important trans-sulfuration pathway in the liver, resulting in the generation of glutathione, the most potent antioxidant in the liver (Stramentinoli, 1987). The generation of hepatic glutathione is an important step in hepatic detoxification (Friedel, Goa, and Benfield, 1989), and SAMe increases hepatic glutathione in alcoholic and non-alcoholic liver disease patients (Vendemiale, Altomare, Trizio, et al., 1986; Vendemiale, Altomare, Trizio, et al., 1989). Both folate and vitamin B12 are essential co-nutrients in the metabolism and replenishment of SAMe (Mato, Alvarez, Ortiz, et al., 1997).

Pharmacokinetic studies have established that SAMe is able to cross the intestinal wall, where it is rapidly metabolized, and its methyl group is incorporated into stable pools including those of plasma proteins and phospholipids (Chawla, Bonkovsky, and Galambos, 1990; Friedel, Goa, and Benfield, 1989; Stramentinoli, 1987). SAMe was found to be 80–90 percent bioavailable following an intramuscular injection (Stramentinoli and Catto, 1976; Osman, Ownes, and Burroughs, 1992). The half-life of an injected 100 mg dose was 81 ± 8 minutes,and that of a 500 mg dose was 101 ± 7 minutes, with volumes of distribution of 0.41 and 0.44 L/kg for doses of SAMe (100 mg and 500 mg), respectively (Giuldori, Cortellaro, Moreo, et al., 1984).

Because of the low bioavailability of orally administered SAMe in healthy volunteers, a significant first-pass effect (degradation in the gastrointestinal tract) and rapid metabolism in the liver are presumed (Stramentinoli, Pezzoli, and Galli-Kienle, 1979). Peak plasma concentrations obtained with an enteric-coated tablet formulation are dose related, with peak plasma concentrations of 0.5 to 1 mg/L achieved 3 to 5 hours after single doses ranging from 400 mg to 1000 mg (Giulidori, Cortellaro, Moreo, et al., 1990; Stramentinoli, 1987). Peak levels decline to baseline within 24 hours.

Sex differences in bioavailability have been reported in a single study (Stramentinoli, 1987). Women showed a 3- to 6-fold greater peak plasma value than men.

Plasma-protein binding of SAMe is less than or equal to 5 percent (Kaye, Blake, and Burroughs, 1990). SAMe crosses the blood-brain barrier, with slow accumulation in the cerebrospinal fluid (Friedel, Goa, and Benfield, 1989). Excretion of unmetabolized SAMe in humans is split between urinary excretion (15.5 ± 1.5 percent) and fecal excretion (23.5 ± 3.5 percent) (Kaye, Blake, and Burroughs, 1990).

No significant adverse events have been reported in the clinical-trial literature for the use of either parenteral or enteral SAMe, and it is generally considered to be safe (Jellin, Batz, and Hitchens, 2001). However, the involvement of SAMe in the re-methylation of methionine from homocysteine (an amino acid intermediate) raises a theoretical concern that disruption of the regulation of transmethylation could result in increased homocysteine, which has been identified as an independent risk factor for heart disease (Selub and Miller, 1992; Finklestein, 2000). However, a pharmacological study found that oral doses of SAMe decreased the levels of homocysteine in the blood and did not adversely affect the activity of 5-methyltetrahydrofolate, a major co-factor involved in the metabolism of homocysteine (Loehrer, Schwab, Angst, et al., 1997).

Epidemiology

Depression is a risk for people of all ages, ethnicities, and education and income levels, regardless of marital status. An estimated 10 percent to 25 percent of women and 5 percent to 12 percent of men in the United States will experience a major depressive disorder in their lifetime. The point prevalence rates have been estimated at 5 percent to 9 percent for women and 2 percent to 3 percent for men (Pincus and Pettit, 2001). With few exceptions, this increased prevalence in women is observed worldwide. Approximately 10 to14 million people are depressed in any given year, with the largest group comprising women ages 18 to 45. Risk factors for depression include female gender, a history of depressive illness, and prior episodes of major depression (Hyman and Rudorfer, 2000; Pincus and Pettit, 2001). Most individuals with depression suffer from recurrent or chronic illness. Without prophylactic treatment, recurrence rates after two or three episodes may range from 60 to 85 percent (Thase and Sullivan, 1995).

Economic

The annual cost of depression in the United States has been estimated at $43.7 billion to $52.9 billion, including costs due to health care, suicide, and losses in productivity. Workplace costs related to absenteeism and reduced productivity have been estimated at $23.8 billion (Greenberg, Stiglin, Finkelstein, et al., 1993). The tremendous negative impact of depression and the lack of effective prevention strategies make the identification of effective treatment of primary importance (Depression in Primary Care Clinical Practice Guideline, 1993).

Conventional Treatment

The mainstays of treatment for depression are antidepressant drugs and short-term psychotherapies. Other treatments such as electroconvulsive therapy (ECT) are typically reserved for specific types of clinical presentation or failure to respond to other interventions.

Antidepressants are a heterogeneous group of compounds, nearly all of which have been effective in the treatment of unipolar and bipolar depression (Hyman and Rudorfer, 2000; Nierenberg, 2001). Antidepressants work acutely for depression in general and for chronic depression in particular. After acute remission, long-term continuation and long-term maintenance protect patients from depressive relapses and recurrences (Dunner, 2001).

Complementary and Alternative Medicine (CAM) Therapies For Depression

A survey of psychiatric outpatients by Knaudt and colleagues (Knaudt, Conner, Weisler, et al., 1999) revealed that 47 percent of all psychiatric patients in the study and 34 percent of patients with major depression were using alternative therapies. In a recent overview, Ernst notes that 20 percent of depressed patients used a CAM therapy in the previous year (Ernst, Rand, and Stevenson, 1998). Depression was one of the 10 most common reasons indicated for CAM use, and a wide variety of therapies including herbs, supplements, exercise, and traditional Chinese medicine have been reported in the literature. Patient preferences drive the choice of CAM therapy for depression, as was demonstrated in a recent article focusing on use of St. John's wort (Wagner, Jester, LeClaire, et al., 1999). Patients cited the desire for personal control of their medical care, the perception of high risk associated with prescription medication, and the ease of access to St. John's wort as reasons for their use of this alternative therapy. All these factors make it quite likely that patients who are depressed will try over-the-counter alternative therapies.

CAM therapies most often cited for use in depression include St. John's wort, which has been the subject of systematic reviews (Linde, Ramirez, Mulrow, et al., 1996; Gaster and Holroyd, 2000) as well as a Cochrane Review (Linde and Mulrow, 1999). Other herbs have been studied for effectiveness in depression or mood alteration but not subjected to systematic review (Werbach and Murray, 2000).

SAMe For Depression

SAMe is the primary methyl donor in the central nervous system. A number of neurotransmitters, fatty acids, phospholipids, and other chemicals must accept methyl groups from SAMe for activation. Thus, it is postulated that the observed antidepressant effect of SAMe is due to its influence on neurotransmitter metabolism and on membrane fluidity and receptor activity (Bottiglieri and Hyland, 1994). Both serum and cerebrospinal fluid levels of SAMe are reported to be low in depressed patients (Bottiglieri, Chary, Laundy, et al., 1988; Bottiglieri, Godfrey, Flynn, et al., 1990), although reports of depressed serum SAMe levels were not consistent (Bottiglieri and Hyland, 1994). However, an increase in SAMe serum levels has been shown to be positively correlated with responses to both SAMe and imipramine therapy (Bell, Potkin, Carreon, et al., 1994), as defined by a 50 percent lower score on the Hamilton Depression Rating Scale.

Etiology

Although the causes of OA are not completely understood, biomechanical stresses affecting the articular cartilage and subchondral bone and biochemical changes in both the articular cartilage and synovial membrane are all important in its pathogenesis. Evidence has been cited on the role of systemic factors such as genetics, diet, estrogen use, and bone density; and local biomechanical factors such as muscle weakness, obesity, and joint laxity (Moskowitz, 1987).

Conventional Treatment

OA has no known cure. Goals of therapy include control of pain and improvement in function and health-related quality of life. Available treatment modalities include joint protection and muscle strengthening; pharmacological treatments; intra-articular injections; biomechanical interventions such as exercise and bracing; and surgical treatments (Felson, Lawrence, Hochberg, et al., 2000; Moskowitz, 1987). Surgical management, including total joint arthroplasty, is generally reserved for failed medical management that affects a patient's quality of life and functioning. Increasingly, appropriate treatment of osteoarthritis combines one or more oral agents with exercise and other biomechanical techniques.

CAM Therapies For Osteoarthritis

Musculoskeletal pain is one of the most common complaints brought to alternative providers. In a survey of family practice patients who use CAM therapies, back pain and other musculoskeletal pain accounted for 36 percent and 17 percent of the visits respectively (Drivdahl and Miser, 1998). Eisenberg (1998) reported that arthritis was the fourth most common condition for which patients sought alternative care. Of the patients surveyed, 26.7 percent had used CAM therapies in the previous 12 months; 38.5 percent had used CAM therapies while also seeing a physician; and only 10 percent had actually seen a CAM practitioner. Thus, a significant portion of their alternative care was self-directed. Forty-seven percent of older adults reported using at least one alternative therapy in a 20-week period and out-of-pocket expenses were very similar for both CAM and conventional therapies at $1,127 versus $1,148 per year respectively (Ramsey, Spencer,Topolski, et al., 2001).

A wide variety of CAM therapies have been used to treat osteoarthritis, but have generally not been the subject of systematic reviews. A recent systematic review (Ezzo, Hadhazy, Birch, et al., 2001) of acupuncture treatment for osteoarthritis of the knee suggested benefit from this therapy. Long and colleagues (Long, Soeken, and Ernst, 2001) have systematically reviewed the literature on herbal medicines for osteoarthritis and found supportive evidence for a variety of herbs and some proprietary blends.

SAMe For Osteoarthritis

SAMe has been shown to have anti-inflammatory and analgesic properties in animal models when given both orally and parenterally (Stramentinoli, 1987). The mechanism by which this effect is achieved is not clear, but it does not seem to be mediated via the prostaglandin system, as inferred from the finding that oral administration of a single dose or repeated doses of up to 1200 mg/kg for 30 days in rats did not affect the integrity of the gastric mucosa, which suggests that SAMe does not interfere with the cytoprotective function of prostaglandins in gastrointestinal tissues. Furthermore, SAMe does not influence platelet aggregation (Stramentinoli, 1987) and has not been demonstrated to affect the eicosanoid system (Di Padova, 1987).

A stimulatory effect of SAMe on chondrocytes has been suggested from studies using human chondrocyte cultures (Harmand, Vilamitjana, Maloche, et al., 1987), and SAMe has also been shown to increase the incorporation of sulfate into proteoglycans (Harnand, Vilamitjana, Maoche et al., 1987). In addition, SAMe does enter the synovial fluid following oral doses of 400 mg (Giulidori, Cortellaro, Moreo et al., 1984), which suggests a reasonable mechanism of action by which SAMe may affect osteoarthritis.

Cholestasis Of Pregnancy

Intrahepatic cholestasis of pregnancy (ICP) is a disease that is diagnosed most frequently in women in their third trimester of pregnancy and is primarily characterized by pruritus, liver enzyme elevations, and occasionally jaundice (Palmer and Eads, 2000). The incidence of ICP is probably about one in 500 to 1000 pregnancies (Williams Obstetrics, 21^st Ed., 2001), but the etiology is not clear, and the condition's occurrence is not predictable. It has been suggested that ICP may arise from an inherited increase in sensitivity to estrogens and progestogens. In addition, altered membrane composition and differences in the canalicular transport systems are implicated in the pathogenesis of this condition. As a result, the normal elevations in maternal levels of bile acids are exaggerated in ICP patients. Also, because the normal fetal-to-maternal transfer of bile acids across the trophoblast is impaired, the excess bile acids accumulate and are toxic to the fetus.

Although maternal pruritus can be severe, overall maternal morbidity and mortality associated with ICP are low (Fagan, 1999). However, fetal morbidity and mortality are significant, with associated risks for meconium-stained amniotic fluid, acute onset of fetal compromise, spontaneous preterm labor, and intrauterine fetal demise (Palmer and Eads, 2000).

The need to manage ICP is dictated by the increased risks of fetal distress, spontaneous preterm delivery, and sudden death, as well as by the desire to alleviate pruritus in the mother. Fetal risks increase progressively until delivery, regardless of serum levels of bile acids and alanine transaminase (a marker for liver function).

Recommended obstetric management includes frequent fetal surveillance and careful intra-uterine monitoring, with prompt delivery when fetal lung maturity has been established. As for conventional therapies, only ursodeoxycholic has been effective in treating maternal pruritus and normalizing the levels of bile acids and sulfated steroids in serum and other body fluids; however, the studies that demonstrated these effects were small (Fagan, 1999). Nicastri and co-workers (Nicastri, Diaferia, Tartagni, et al., 1998) reaffirmed these findings and further reported that S-adenosylmethionine increased the effects of ursodeoxycholic acid (Williams Obstetrics, 21^st Ed., 2001).

Cholestasis In Chronic Liver Disease

Intrahepatic cholestasis (IHC) is defined as the impedance of bile formation and flow without the presence of an extrahepatic obstruction. It is a relatively common complication of a wide variety of chronic liver diseases, both those that result in parenchymal damage and those that do not (McGill and Kwiatkowsi, 1998). In a series of 867 Italian inpatients with cirrhosis (83 percent), chronic active hepatitis (15 percent) and primary biliary cirrohsis (2 percent), 35 percent were found to have intrahepatitic cholestasis (Cimino, Dattilo, Topa, et al., 1989). These results were confirmed in a second survey of 2520 newly diagnosed chronic liver patients (cirrhosis, 56 percent; chronic hepatitis, 27 percent; primary biliary cirrhosis, 6 percent), which demonstrated that 35 percent of these patients had IHC as well (Bortolini, Almasio, and Bray, 1992). Although it is difficult to assign an economic cost to this condition, it is the cause of significant morbidity for the affected patients when it leads to pruritus.

The etiology of IHC appears to vary and may include medications; conditions associated with excessive estrogen (such as the use of oral contraceptive agents or pregnancy); infectious hepatitis (especially viral); autoimmune liver diseases; and toxic liver damage, such as alcoholic liver disease (Sherlock and Dooley, 1997). The clinical presentation is characterized by pruritus, usually with jaundice, and malabsorption due to inadequate bile acids in the intestine. Long-term sequelae may include steatorrhea, hepatic osteodystrophy, and fat-soluble vitamin deficiencies, especially of vitamin K with concomitant coagulopathy and bleeding (Sherlock and Dooley, 1997).

Conventional therapy for this illness includes cholestyramine, ursodeoxycholic acid, steroids, antihistamines, opiate antagonists and rifampicin. However, none of the conventional therapies is markedly effective, and none is without side effects. Some authorities recommend SAMe as an accepted therapy, while others regard it as no more than an experimental therapy (Raidford, 1995; Sherlock and Dooley, 1997).

CAM Therapies And Liver Disease

Flora and his group at the Oregon Health Sciences University report on the widespread use of herbal medications and supplements by their chronic liver disease patients (Flora, Rosen, and Benner, 1996). In a survey of 103 patients attending a hepatology clinic, they found that 31 percent were using such remedies. A wide variety of substances were cited by patients, and 19 patients were using multiple remedies. Milk thistle was the most frequently used remedy. The majority of patients obtained information from sources other than their medical doctors, relying instead on books, magazines, friends, or family.

CAM therapies for use in hepatitis C have been the subject of two recent reviews (Milliman, Lamson, and Brignall, 2000; Patrick, 1999). Milk thistle and its anti-oxidant complex, silymarin, are the herbal remedies most often cited for use in chronic liver disease and have been the subject of a systematic review by the Agency for Healthcare Research and Quality (Mulrow, Lawrence, Jacobs, et al., 2000). Despite the widespread use of herbal therapies for chronic liver disease, some experts call for caution, citing fear of hepatotoxicity (Bass, 1999).

SAMe And Liver Disease

SAMe has been used to treat various types of acute and chronic liver diseases, including viral hepatitis (both B and C), alcoholic liver disease (including cirrhosis), Gilbert's syndrome and other disorders of bilirubin and porphyrin metabolism, as well as drug-induced cholestasis. Additionally, SAMe has been used preventively to protect the liver from the effects of a number of hepatotoxic drugs (Friedel, Goa, and Benefield, 1989). Thus, the focus of clinical trials in this area has been diffuse. However, a number of clinical trials have focused on the effect of SAMe on cholestasis from a variety of causes, including pregnancy.

SAMe may exert beneficial effects on the liver through a variety of mechanisms. Glutathione, the major anti-oxidant in the liver, plays a key role in detoxification and limiting oxidative damage. At customary therapeutic doses, SAMe has been shown to increase hepatic glutathione concentrations in patients with chronic liver disease (Chawla, Bonkovsky, and Galambos, 1990). Results of clinical trials have suggested that SAMe may decrease fatty deposition in the liver (Micali, Chiti, and Balestra, 1983), and SAMe's ability to protect against ethanol damage to the liver was demonstrated in baboons fed a diet high in alcohol (Lieber, Casini, De Carli, et al., 1990).

Effects on membrane fluidity and on the activity of the Na+, K+ ATPase-dependent pump are postulated to account in part for the beneficial effect of SAMe on intrahepatic cholestasis (Almasio, Bortolini, Pagliaro, et al., 1990). Hormones, estrogen in particular, are known to contribute to cholestasis, especially of pregnancy (Bacq and Sapey, 1998; Davidson, 1998; Reyes, 1992; Salen and Batta, 1999). SAMe has been demonstrated to decrease the negative effects of ethnylestradiol on bile flow and bile composition (Coltori, Bortolini, and Di Padova, 1990). SAMe's ability to increase the bioavailability of sulfates through transulfuration has been postulated to improve bile flow by favoring production of more soluble sulfurated bile salts (Yousef, Branswell, Tuchweber, et al., 1987). In fact, SAMe seems to exert its anticholestatic effect in a nonspecific manner, regardless of the etiology of the cholestatic liver injury (Coltori, Bortolini, and Di Padova, 1990).

Technical Expert Panel

The Evidence-Based Practice Center was advised on CAM topics by a group of technical experts regarding the search and inclusion criteria and appropriate analyses. The technical experts represented diverse disciplines including acupuncture, Ayurvedic medicine, chiropractic, dentistry, general internal medicine, gastroenterology, rheumatology, integrative medicine (the practice of combining alternative and conventional medicine), neurophysiology, pharmacology, psychiatry, psychoneuroimmunology, psychology, sociology, botanical medicine, and traditional Chinese medicine. The technical experts assisted the project in several ways. They aided us in identifying potential topics for review, appropriate sources of relevant literature, and technical experts for peer review; assessing our search strategies; and addressing specific questions in their areas of expertise. In addition, one of our expert panel members assisted in reviewing literature in languages other than English, and his contribution is listed in the Acknowledgments appendix. The Acknowledgment appendix also lists members of the expert panel along with their affiliations.

Identification Of Literature Sources

Table 1. Biomedical databases searched

Table 1. Biomedical databases searched

Database	Years
Allied and Complementary Medicine	1984-2000 Oct
Biosis Previews	1969-2000 Dec
Cinahl	1982-2001 Jun
Cochrane Library	1922- 2000 Dec
Embase	1974-2000 Nov
HealthSTAR	1975-2000 Dec
MANTIS	1880-2000 Apr
Medline	1966-2000 Dec

Table 2. Pharmaceutical and other biomedical databases (more...)

Table 2. Pharmaceutical and other biomedical databases searched

Database	Years
Adis LMS Drug Alerts	1983-2000 Nov
CAB HEALTH	1983-2000 Oct
Conference Papers Index	1973-2000 Sep
Derwent Drug File	1983-2000 Dec
Drug News & Perspectives	1992-2000 Nov
ELSEVIER BIOBASE	1994-2000 Nov
Gale Group Newsletter DB	1987-2000 Dec
Health News Daily	1990-2000 Dec
IMSWorld R&D Focus	1991-2000 Dec
Inside Conferences	1993-2000 Dec
International Pharmaceutical Abstracts	1970-2000 Oct
Mental Health Abstracts	1969-2000 Jun
Pascal	1973-2000 Dec
Pharmaceutical News Index	1974-2000 Nov
Pharm-Line	1978-2000 Oct
PsycINFO	1887-2000 Oct
SciSearch (Current)	1990-2000 Dec
TGG Health&Wellness DB	1976-2000 Nov

Table 3. Additional search terms for SAMe

Table 3. Additional search terms for SAMe

ADEMETIONINE

ALDOMET

GUMBARAL

S-ADENOSYL-L-METHIONINE

SADENOSYLMETHIONINE

S-ADENOSYLMETHIONINE

SADENOSYL-METHIONINE

S-ADENOSYL-METHIONINE

Two reviewers independently evaluated the list of 1562 titles that the on-line database searches generated. The reviewers read the lists of titles and accepted articles that satisfied the following criteria:

• Focused on SAMe for depression, osteoarthritis, or chronic liver disease;

• Presented any historical or descriptive background information about SAMe and its use;

• Presented a meta-analysis or systematic review of SAMe and one of the three disease states identified.

Articles that fit the following criteria were rejected:

• Focused on a disease state that was not one of the three selected;

• Contained animal or in vitro data unless human clinical trial information or significant background information was also included.

Language was not considered a barrier to inclusion.

We obtained 62 articles that we identified from the reference lists of the articles previously ordered and from the professional libraries of our project staff and their colleagues.

From this stage of the screening process, the reviewers requested a total of 294 unique articles, of which we were able to obtain 285. Appendix B includes a list of articles that we could not obtain. Selected articles were referred to clinical research reviewers for further evaluation and for possible inclusion in the evidence synthesis.

Table 4. Summary of search strategy

Table 4. Summary of search strategy

Name	Description	Purpose	Number of Titles
Search 1	Focused search on intervention, disease states, and study designs.	Convince reviewers that the quality and amount of literature warranted literature review.	68
Search 2	Biomedical databases searched on intervention only.	Search for potential evidence.	162
Search 3	Biomedical databases searched on intervention, by disease states, and limited to human subjects.	Search for potential evidence.	720
Search 4	Pharmaceutical and complementary and alternative medicine databases searched on intervention, by disease states, and limited to human subjects.	Search for potential evidence.	601
Search 5	All databases searched for SAMe synonyms in Table 3.	Search for potential evidence.	11
Total titles generated from database searches			1562

Extraction Of Data

Detailed information from each of the 102 studies was collected on a specialized data collection instrument (the Quality Review Form) designed for this purpose. This Quality Review Form (Appendix D) was developed in consultation with our technical experts. We included questions about the study design; the technical quality of the study; the number and characteristics of the patients; patient recruitment information; details on the intervention, such as the dose, route of administration, frequency, and duration; the types of outcome measures; and the time between intervention and outcome measurement. Two trained reviewers, working independently, extracted data in duplicate and resolved disagreements by consensus. A senior physician researcher on the project staff resolved any disagreements not resolved by consensus.

To evaluate the quality of the studies, we collected information on the study design, appropriateness of randomization, blinding, description of withdrawals and dropouts (Jadad, Moore, Carroll, et al., 1996), and concealment of allocation (Schulz, 1995). A score for quality was calculated for each study using a system developed by Jadad (Jadad, Moore, Carroll, et al., 1996). See Appendix E for further details (Moher, Pham, Jones, et al., 1998).

Figure 1. SAMe Literature Search and Review Strategy (more...)

   Figure 1. SAMe Literature Search and Review Strategy

Figure 1

Study Inclusion

In selecting studies for the meta-analysis, we considered all 102 studies for which a Quality Review Form (QRF) was prepared. For both depression and osteoarthritis, the available studies were judged to be sufficiently clinically homogeneous to support a pooled analysis. On the other hand, the studies on liver disease encompassed a much wider variety of clinical conditions. The largest group of studies identified within the area of liver disease focused on intrahepatic cholestasis. Within that group, we made the decision to stratify the studies into those on cholestasis of pregnancy and those for which intrahepatic cholestasis may represent a “final common pathway” for a variety of mainly chronic liver disease conditions. In addition, a number of the pregnancy studies had an active therapy comparison arm (ursodeoxycholic acid), which was not the case for the chronic liver disease studies.

For each of the four conditions (depression, osteoarthritis, cholestasis of pregnancy, and intrahepatic cholestasis associated with chronic liver disease), we chose the clinically relevant outcomes and clinically comparable follow-up times that were reported most often. Some of the outcomes were continuous measures, and some were dichotomous variables. For studies with a crossover design, we extracted data only from the first treatment phase, prior to the crossover. If such data were not available, the study was excluded from the meta-analysis.

We synthesized effect sizes for continuous outcomes and risk ratios for dichotomous outcomes. In order for a study to be included in the analysis, the original report had to contain sufficient statistical information for the calculation of an effect size or risk ratio as appropriate for the relevant outcome, and the report could not contain duplicate data. The definitions of duplicate data and what constituted sufficient statistical information follow.

Duplicate Publications, Studies And Data

We distinguished three different types of data duplication. First, multiple citations of the same article were removed at the title screening stage of the project, so all publications that reached the synthesis stage contained reports of unique studies. Second, some publications were based on the same study population and experiment as others, yet each reported different outcome data. If both outcomes were relevant for the meta-analysis, we included both articles in our evidence table and note the link between the studies. Third, in some instances, several publications reported on the same study population and intervention and presented the same outcome data. In these cases, we picked the most informative of the duplicates; for example, if one publication was a conference abstract with preliminary data, and the second was a full journal article, we chose the latter. The publications dropped for duplicate data do not appear in the evidence table but are discussed in the text of the results section within each condition.

Stratification Of Interventions

Potentially significant variability was noted in the interventions studied. Routes of administration for SAMe included both parenteral and oral. Dosages of both SAMe and other comparison drugs varied as well. In an effort to assess the comparability of the interventions tested, we stratified interventions by route of administration and level of dose. For SAMe, the dosage stratifications (i.e. what constituted a high or low dose) were based on expert opinion and clinical experience of the research team members. Non-steroidal anti-inflammatory drugs (NSAIDs) were categorized using a strategy previously developed by the SCEPC (Ofman, in press 2001). This stratification system was considered in the subsequent analysis of the selected studies.

Effect Size Calculations For Continuous Outcomes

For each study, we calculated effect sizes for the SAMe arm of interest and the other arms in each study that were considered relevant. Generally, each study included one comparison between a single SAMe arm and a placebo or treatment (e.g., NSAIDS) arm. Some studies contained more than one SAMe arm (representing different doses) or contained more than one other treatment arm and thus contributed more than one effect size to be considered for analysis. Double-counting patients became a concern if a study contributed more than one effect size and patients were included more than once in those effect sizes. For example, if a study had one placebo arm and two SAMe arms, it contributed two effect sizes, both based on the same placebo patients. In such cases, we used clinical relevance as our criterion, and included in the meta-analysis the effect size for the SAMe arm that was most comparable in terms of dose to the SAMe arms in the other studies in the analysis.

For each study, the means and standard deviations of each outcome at the designated follow-up time for each relevant arm were extracted if available. If studies did not report a follow-up mean, or if a follow-up mean could not be calculated from the given data, the study was excluded from the meta-analysis. If a study did not report a standard deviation, or if a standard deviation could not be calculated from the given data, we imputed the standard deviation by using those studies and arms that did report a standard deviation and weighting all arms equally in the imputed value calculation, or we assumed that the standard deviation was 0.25 of the theoretical range for the specific measure in the study. For example, if a study measured pain on a 0–100 scale, we assumed the standard deviation was 25. For each pair of arms, an unbiased estimate (Hedges and Olkin, 1985) of Hedges' g effect size (Rosenthal, 1991) and a 95 percent confidence interval were calculated. A negative effect size indicates that SAMe is associated with a decrease in the outcome at follow-up as compared with the comparison arm, e.g., placebo. For example, in the OA meta-analysis, the outcome was pain, so a negative effect size indicated that SAMe was associated with a decrease in pain at follow-up as compared with placebo.

Risk Ratio Calculations For Dichotomous Outcomes

We estimated log risk ratios and constructed 95 percent confidence intervals in the logarithmic scale, for variance-stabilization reasons (Ioannidis, Cappelleri, Lau, et al., 1995). We then back-transformed to the risk ratio scale. For studies that had zero outcomes in either the SAMe or comparison arms or had all patients in either arm with the outcome, we performed a continuity correction by adding 0.5 to all cells in the two-by-two table of arm by outcome.

A risk ratio of less than 1.0 indicates that the chance of the outcome in the comparison arm is smaller than that in the SAMe arm. For example, in the depression meta-analysis, one of the dichotomous outcomes was “greater than 25 percent improvement on the Hamilton Depression Scale or not.” A risk ratio of 0.5 indicates that half as many people improved in the comparison arm as in the SAMe arm, or, analogously, twice as many people improved in the SAMe arm as in the comparison arm.

Meta-Analysis

When appropriate, we estimated a pooled random-effects estimate (DerSimonian and Laird, 1986) by combining either effect sizes or risk ratios, depending on the outcome. We also report the Chi-squared test of heterogeneity p-value (Hedges and Olkin, 1985). When relevant, we conducted sensitivity analyses on subgroups of studies to determine the robustness of our conclusions.

We assessed the possibility of publication bias by evaluating a funnel plot of effect sizes or log risk ratios for asymmetry, which results from the nonpublication of small, negative studies. If no publication bias exists, the full symmetric distribution of study effect sizes will be observed in the funnel plot. If bias exists, the distribution will be asymmetric or skewed due to the fact that some studies are missing because they have not been published. Because graphical evaluation can be subjective, we also conducted an adjusted rank correlation test (Begg and Mazumdar, 1994) and a regression asymmetry test (Egger, Davey Smith, Schneider, et al., 1997) as formal statistical tests for publication bias. The correlation approach tests whether the correlation between the effect sizes and their variances is significant, and the regression approach tests whether the intercept of a regression of the effects sizes on their precision differs from zero, i.e., both formally test for asymmetry in the funnel plot. We conducted all analyses and drew all graphs using the statistical package Stata (1999). The results of our analysis are presented in the following section.

Study Inclusion/Exclusion For The Depression Outcomes

We considered 47 studies for inclusion in the depression meta-analysis. All studies had two arms except for one that had three. Studies could be excluded if they represented data duplicated by another included study, could not be obtained, did not have an appropriate outcome measure, or were not clinically similar enough to be included in the analysis.

Five studies (Agnoli, Fazio, Andreoli, et al., 1975; Caruso, Fumagalli, Boccassini, et al., 1984; Salmaggi, Bressa, Nicchia, et al., 1991; Rosenlicht, Kagan, Sultzer, et al., 1989; Kagan, Sultzer, Rosenlicht, et al., 1990) were excluded from the meta-analysis because they contained data also contained in other studies (Agnoli, Andreoli, and Casacchia, 1976; Caruso, Fumagalli, Boccassini, et al., 1987; Salmaggi, Bressa, Nicchia, et al., 1993; Kagan, Sultzer, Rosenlicht, et al., 1990 respectively). Two studies (Bell and Potkin, 1988; Potkin, Bell, Plan, et al., 1988) contained the same data as a single other study (Bell, Plon, Burney, et al., 1988) that was included in the analysis. Another study (Bell, 1990) could not be obtained and so could not be included in the analysis.

Eleven additional studies that were considered for inclusion in the meta-analysis were also excluded. Two studies (Ceruti, Sichel, Perin, et al., 1993; Pons-Villegas, 1983) did not contain data on a relevant outcome. Six studies (Agnoli, Andreoli, and Casacchia, 1976; Barberi, and Pusateri, 1978; Schifano and Garoli, 1993; Fazio, Andreoli, and Agnoli, 1973; Delle Chiaie, Pancheri and Scapicchio, 2000a; and Delle Chiaie, Pancheri and Scapicchio, 2000b) were excluded because they did not include sufficient statistics for the calculations to be performed. The remaining studies involved clinical situations that we did not judge to be appropriate for a pooled analysis. Two studies (Chinchilla, Moreno, Piñero, et al., 1996; Berlanga, Ortega-Solo, Onitveros, et al., 1992) were excluded because they studied the effects of SAMe on the onset of action of a conventional antidepressant medication and thus were not considered clinically comparable to the other studies. A third study (Lo Russo, Monaco, Pani, et al., 1994) involved the treatment of depressive symptoms in patients undergoing opiate detoxification.

Studies excluded from the meta-analysis but presented in the evidence table will be discussed following the presentation of the meta-analysis results.

Meta-Analysis For Depression

Twenty-eight studies were included in the meta-analysis. Ten studies were included in all three analyses, that is, the risk ratio and the effect size analyses. Fourteen studies were included in the effect size analysis only. Two were included in the risk ratio analyses only. One study was included in the effect size and a single risk ratio (greater than 25 percent improvement) analyses; and the final one was included only in the risk ratio greater than 50 percent improvement analysis.

Meta-Analysis Of SAMe Versus Placebo

Risk Ratio Analysis: SAMe Versus Placebo

Table 5. SAMe versus placebo studies risk ratios

Table 5. SAMe versus placebo studies risk ratios

First Author & Year	SAMe Dose	Total N	RR 25%	95% CI for RR 25%	RR 50%	95% CI for RR 50%
Janicak (1988)	HIGH/ PO	12	0.12	(0.01, 1.79)	0.44	(0.02, 9.11)
Kagan (1990)	HIGH/ PO	15	0.43	(0.13, 1.40)	0.25	(0.04, 1.59)
Muscettola (1982)	MED/ IM	18	0.40	(0.10, 1.55)	NR	NA

CI: confidence interval; IM: intramuscular, NR: Not reported; NA: Not applicable; PO: oral administration; RR: risk ratio.

Three studies contained placebo arms and sufficient data to calculate risk ratios for either a 25 percent improvement in the HRSD, or a 50 percent improvement in the HRSD, or both. The results of this analysis are summarized in Table 5. In this and subsequent tables, the total n is the sum of the sample sizes in the SAMe and comparison (in this case, placebo) arms.

Since we could calculate the risk ratios for improvement from only three studies, we do not pool the results but discuss them only descriptively. All three of these studies report improvement, but none of the results reaches conventional statistical significance. Each of the studies is limited by the small number of subjects included. The studies ranged in size from about six patients in the SAMe and placebo groups respectively, to nine patients in each group. The risk ranged from about 25 percent to 75 percent across the two outcomes and across studies in the SAMe group. If we assume that the risk ratio is 0.4, which is about the average risk ratio observed, then the risk in the placebo group ranges from 10 percent to 30 percent. The power to detect differences in risk between the two groups for this range (25 percent versus 10 percent; 50 percent versus 20 percent; and 75 percent versus 30 percent) is extremely low for comparison groups of sizes six and nine. In fact, the power does not exceed 30 percent in any case. Even if the risk ratio is as low as the minimum observed (0.12) and assuming the sample sizes in that study (seven and five in each group respectively), the power is only 34 percent. Therefore, these studies are so limited in statistical power that no definitive conclusion can be drawn.

Both the intervention and the quality of the studies also vary somewhat. Kagan's study (Kagan, Sultzer, Rosenlicht, et al., 1990 Kagan, Sultzer, Rosenlicht, et al., 1990) was well designed and received a Jadad score of 5 (see Methods). However, the study design could not be fully implemented due to regulatory problems. Both the Janicak (Janicak, Lipinski, Davis, et al., 1988) and the Kagan (Kagan, Sultzer, Rosenlicht, et al., 1990) studies reported large numbers of dropouts (Biggs, Wylie, and Ziegler, 1978). Neither study included these patients in their final analyses; therefore, the results reported may be skewed towards a positive effect. Thus, the studies present evidence that supports an effect of SAMe relative to placebo (i.e., all of the risk ratios favor SAMe), but these findings are not conclusive (i.e., the differences between groups are not statistically significant and the studies have methods problems).

Effect Size Analysis: SAMe Versus Placebo

Table 6. SAMe versus placebo: depression effect size, (more...)

Table 6. SAMe versus placebo: depression effect size, pooled analysis

First Author & Year	SAMe Dose & Route of Administration	Total N	Effect Size	95% CI
Ancarni (1993)	HIGH/IV	51	-0.13	(-0.82, 0.56)
Carney (1986)	MED/ IV	32	-0.25	(-0.94, 0.45)
Carrieri (1990)	HIGH/PO	17	-1.28	(-2.34, 0.56)
Caruso (1987)	MED/ IM	59	-0.99	(-1.54, -0.45)
Delle Chiaie (1997)	HIGH/IV	75	-0.43	(-0.89, 0.03)
De Leo (1987)	Med/IM	36	-0.82	(-1.50, -0.13)
Fava (1992)	HIGH/ PO	39	0.07	(-0.56, 0.71)
Kagan (1990)	HIGH/PO	15	-1.20	(-2.32, -0.09)
Muscettela (1982)	MED/ IM	18	-4.29	(-5.96, -2.61)
Salmaggi (1993)	HIGH/ PO	60	-0.35	(-0.86, 0.16)
Thomas (1987)	HIGH/IV	20	0.05	(-0.84, 0.93)
Pooled Random Effects Estimate			-0.65 (1)	(-1.05, -0.25)

(1)

Chi-squared test of heterogeneity p-value < 0.001.

CI: confidence interval; IM: intramuscular; IV: intravenous; PO: oral administration.

Figure 2. Depression Outcome: Continuous Hamilton Score (more...)

   Figure 2. Depression Outcome: Continuous Hamilton Score SAMe Versus Placebo

Eleven studies provided sufficient data to calculate an effect size for the HRSD measured continuously. The results of this analysis are summarized in Table 6 and Figure 2.

The pooled estimate demonstrates a significant decrease of 65 percent of a standard deviation in the HRSD score, associated with SAMe treatment. However, the Chi-squared test of heterogeneity is also significant, with a p-value of less than 0.001.

Given the observed heterogeneity and the variability of dosage level and route of administration of SAMe, a further analysis was done to attempt to ascertain the effect of this variability on the reported effect sizes. When we stratified by SAMe dose and pooled the four medium-dose studies together and the seven high-dose studies together respectively, we obtained the following results. For the medium-dose studies, the pooled effect size is -1.28 (95% CI [-2.26, -0.31]) with a Chi-squared test of heterogeneity of p < 0.001. For the high-dose studies, the pooled effect size is -0.36 (95% CI [-0.66, 0.05]) with a Chi-squared test of heterogeneity of p = 0.22. Thus, stratification by total dose alone does not account for all of the heterogeneity observed. Further, no escalating dose-response relationship was demonstrated in the trials, based on total dose. Route of administration may be a potential confounding factor in assessing response based on dose. Four of the seven high-dose studies involve oral preparations. Thus, it is possible that the lower effect size is the result of the route of administration rather than the total dose.

Results of the pooled analysis summarized in Figure 2 show an effect size of -0.65 in favor of treatment with SAMe over the placebo. In order to assess the clinical importance of such an effect, we transformed this effect size back into a change in the score of the HRSD, assuming the average from all arms in the analysis as the standard deviation. Given this assumption, the pooled effect size is calculated to be equivalent to a drop of 5.6 points on the HRSD.

The clinical significance of a change in the HRSD can be assessed in two ways. First, if the score falls below a particular number (that number varies depending on which version of the score is used), then that patient is no longer considered to be significantly depressed. Second, for a change in the HRSD to represent a “response” to treatment, the score must change by 50 percent or more. A change of greater than 25 percent but less than 50 percent is taken to represent an improvement in symptoms but not a remission or complete response to treatment. However, since the actual number of possible points depends on which version of the test is being administered, the percentage of change represented by 5.6 points would vary as well. Versions of the HRSD used in studies included in our analysis varied from 11 items to 21 items. The versions used most often were the 17-question and the 21-question versions. In a recent study of a new antidepressant, the minimal initial HRSD score (using a 24-item version) for inclusion in the study was 20, and a change in score of 50 percent was defined as a significant clinical response. Thus, in this study, a raw score of 10 would be the minimal acceptable change for a clinically significant response (Keller, McCullough, Klein, et al., 2000). On this basis, it is unlikely that a change of 5.6 points would correspond to a highly significant clinical result such as complete remission of symptoms (greater than or equal to a 50 percent change). However, such a change is in the range of clinically significant improvement greater than or equal to a 25 percent change. Thus, the results of these studies support a clinically and statistically significant benefit of SAMe relative to placebo.

Meta-Analysis Of SAMe Versus Other Drugs

Risk Ratio Analysis: SAMe Versus Other Drugs

Table 7. SAMe versus other drugs risk ratios pooled (more...)

Table 7. SAMe versus other drugs risk ratios pooled analyses

First Author & Year	Other Drug	Drug Dose	SAMe Dose & Route of Admin.	Total N	RR 25%	95% CI for RR 25%	RR 50%	95% CI for RR 50%
Barberi (1978)	Amitriptyline	Usual	MED / IV	20	1.00	(0.83, 1.20)	1.00	(0.83, 1.20)
Bell (1988)	Imipramine	Usual	HIGH /IV	18	0.26	(0.09, 0.78)	0.33	(0.09, 1.23)
Bell (1994)	Desipramine	Usual	HIGH / PO	17	NR	NA	0.61	(0.17, 2.14)
Calandra (1979)	Chlorimipramine	Usual	MED / IV	24	1.00	(0.86, 1.17)	1.00	(0.07, 14.21)
Del Vecchio (1978)	Chlorimipramine	Usual	MED / IM	28	1.25	0.42, 3.70)	1.00	(0.02, 47.18)
Janicak (1988)	Imipramine	Usual	HIGH / IV	10	0.93	(0.37, 2.36)	2.33	(0.21, 26.23)
Mantero (1975)	Imipramine	Usual	LOW / IM	31	0.90	(0.61, 1.33)	0.87	(0.51, 1.48)
Miccoli (1978)	Chlorimipramine & Amitriptyline	Usual	MED / IV	86	1.00	(0.96, 1.04)	0.96	(0.73, 1.27)
Monaco (1979)	Amitriptyline	Usual	LOW / IV	20	0.78	(0.54, 1.14)	0.73	(0.24, 2.27)
Scaggion (1982)	Nomifensine	–	MED / PO	40	0.87	(0.70, 1.09)	0.17	(0.02, 1.29)
Scarzella (1978)	Chlorimipramine	Usual	MED / IV	20	1.00	(0.75, 1.34)	0.78	(0.49, 1.23)
	Pooled Random Effects Estimate				0.99(1)	(0.95, 1.03)	0.93(2)	(0.82, 1.07)

CI: confidence intervals; IM: intramuscular; IV: intravenous; NR: Not reported; NA: Not applicable; PO: orally administered; RR: relative risk.

(1)

Chi-squared test of heterogeneity p-value = 0.43 for RR 25% Change.

(2)

Chi-squared test of heterogeneity p-value = 0.68 for RR 50% Change.

Figure 3. Depression Outcome: Number Greater than 50% (more...)

   Figure 3. Depression Outcome: Number Greater than 50% SMAe Versus Other Drugs

Figure 4. Depression Outcome: Number Greater THan 25% (more...)

   Figure 4. Depression Outcome: Number Greater THan 25% SAMe Versus Other Drugs

We identified 11 studies that contained other drug arms and sufficient data to calculate risk ratios for either a 25 percent improvement in the HRSD, or a 50 percent improvement in the HRSD, or both. The results of this analysis are summarized in Table 7 and Figures 3 and 4.

The data reported in this group of studies do not demonstrate significant statistical heterogeneity. However, with risk ratios of approximately 1 (0.99 and 0. 93), these data do not support an effect of SAMe that is either superior or inferior to the comparison antidepressant drugs.

Effect Size Analysis: SAMe Versus Other Drugs

Table 8. SAMe versus other drugs depression effect (more...)

Table 8. SAMe versus other drugs depression effect size pooled analysis

First Author & Year	Other Drug	Drug Dose	SAMe Dose & Route of Admin.	Total N	Effect Size	95% CI
Barberi (1978)	Amitriptyline	Usual	MED / IV	20	0.25	(-0.63, 1.13)
Bell (1988)	Imipramine	Usual	HIGH / IV	18	-1.59	(-2.64, -0.53)
Calandra (1979)	Chlorimipramine	Usual	MED / IV	24	1.31	( 0.43, 2.19)
Cerutti (1989)	Minaprine	N/A	MED / PO	20	0.05	(-0.82, 0.93)
Del Vecchio (1978)	Chlorimipramine	Usual	MED / IV	28	0.63	(-0.13, 1.39)
Delle Chiaie (1997)	Chlorimipramine	Usual	HIGH / PO	122	0.50	(0.14, 0.86)
Di Padova (2000)	Imipramine	Usual	HIGH / IM	295	0.13	(-0.10, 0.36)
Di Padova (2000)	Imipramine	Usual	HIGH / PO	281	0.13	(-0.10, 0.36)
Kufferle (1982)	Chlorimipramine	Usual	MED / IV	18	0.09	(-0.83, 1.02)
Miccoli (1978)	Chlorimipramine & Amitriptyline	Usual	MED / IV	86	-0.02	(-0.44, 0.40)
Monaco (1979)	Amitriptyline	Usual	LOW / IV	20	-0.19	(-1.07, 0.69)
Scaggion (1982)	Nomifensine	N/A	MED / IV	40	-0.93	(-1.58, -0.27)
Scarzella (1978)	Chlorimipramine	Usual	MED / IV	20	-0.19	(-1.07, 0.69)
Vanna (1992)	Imipramine	Usual	MED / PO	23	0.21	(-0.61, 1.03)
	Pooled Random Effects Estimate				0.08 (1)	(-0.17, 0.32)

(1) Chi-squared test of heterogeneity p-value 0.001.

Figure 5. Depression Outcome: Continuous Hamilton Score (more...)

   Figure 5. Depression Outcome: Continuous Hamilton Score SAMe Versus Other Drugs

Fourteen studies that compared the efficacy of SAMe to that of other antidepressant drugs contained sufficient data to calculate an effect size in the HRSD measured continuously. The results of this analysis are summarized in Table 8 and in Figure 5.

Studies are heterogeneous as demonstrated by the Chi-squared test. Again, the pooled effect is not significant, indicating that the effects of SAMe and the other medications are approximately equivalent. The 95 percent confidence intervals are such that we cannot exclude a 20 percent of standard deviation difference in either direction.

Publication Bias

Table 9. Publication bias test results

Table 9. Publication bias test results

Study/Analysis	Number of Studies	Adjusted Rank Correlation Test	Regression Asymmetry Test
SAMe versus other drugs – 25% RR	10	p=0.15	p=0.09
SAMe versus other drugs – 50% RR	11	p=0.44	p=0.08
SAMe versus placebo – effect size	11	p=0.44	p=0.07
SAMe versus other drugs – effect size	14	p=0.44	p=0.54

We assessed publication bias for four subgroups of studies included in the meta-analysis. The results are shown in Table 9.

Figure 6. Depression Outcome: Continuous Hamilton Begg's (more...)

   Figure 6. Depression Outcome: Continuous Hamilton Begg's Funnel Plot with Pseudo 95% Confidence Limits

Figure 7. Depression Outcome: Continuous Hamilton Begg's (more...)

   Figure 7. Depression Outcome: Continuous Hamilton Begg's Funnel Plot with Pseudo 95% Confidence Limits

Figure 8. Depression Outcome: Number Greater Than 50% (more...)

   Figure 8. Depression Outcome: Number Greater Than 50% Begg's Funnel Plot with Pseudo 95% Confidence Limits

Figure 9. Depression Outcome: Number Greater Than 25% (more...)

   Figure 9. Depression Outcome: Number Greater Than 25% Begg's Funnel Plot with Pseudo 95% Confidence Limits

The results of the regression asymmetry test approached statistical significance, raising the possibility of publication bias (i.e., unpublished studies that do not report favorable results). Visual inspection of the funnel plots was not conclusive (see Figures 6–9).

Summary Of The Results Of SAMe Trials For Treatment Of Depression

All measures of the effect of SAMe versus placebo favored active treatment, although only the effect size analysis reached the level of conventional statistical significance. The clinical significance of the pooled estimate of effect (a 5.6 point improvement in the HSRD) is equivalent to an improvement, but not a total resolution, of symptoms.

Risk ratios and effect size calculations for SAMe compared to other antidepressants did not demonstrate a statistically significant difference. These findings suggest that both therapies were approximately equal in their effects. However, the possibility of publication bias requires tempering these conclusions.

Study Inclusion/Exclusion

We considered 14 studies for inclusion in the osteoarthritis meta-analysis. One study (Bradley, Flusser, Brandt, et al., 1991) was removed from the analysis and the evidence table because it was an abstract that reported the same data as another full-length publication (Bradley, Flusser, Katz, et al., 1994). Of the 13 remaining studies, 12 had two arms: a SAMe arm and either a placebo or a nonsteroidal anti-inflammatory drug (NSAID) arm. The remaining study (Caruso and Pietrogrande, 1987) had three arms: SAMe, placebo, and NSAID.

Two studies (Glorioso, Todesco, Mazzi, et al., 1985; Montrone, Fumagalli, Sarzi Puttini, et al., 1985) were excluded because they did not contain sufficient statistics for a calculation of effect size. We excluded another study (Cucinotta, Mancini, Ceccato, et al., 1980) because it did not report on a pain outcome.

A difficulty arose with the Bradley study (Bradley, Flusser, Katz, et al., 1994), which reported on two study sites (Sites A and B). The samples enrolled at the two sites were not equivalent at baseline and did not behave similarly during the trial, despite the fact that the interventions were identical. Because of these persistent differences, the authors analyzed and reported the results of the two groups separately. Therefore, we treated these sites separately in the meta-analysis as well; in the remaining discussion, we consider them separate studies. However, the results from the two sites appear in the evidence table as a single entry.

As a result of these exclusions and splitting of studies, we included 11 studies in the meta-analysis. These studies included 12 placebo or NSAID arms and 11 SAMe arms.

Meta-Analysis Of SAMe Versus Placebo

Table 10. SAMe versus placebo: OA, pooled analysis

Table 10. SAMe versus placebo: OA, pooled analysis

First Author & Year	SAMe Dose & Route of Administration	Total N	Effect Size	95% CI
Bradley (1994) Site A	MED/ IV & PO	48	-0.40	(-0.97, 0.17)
Bradley (1994) Site B	MED/ IV & PO	33	0.60	(-0.09, 1.30)
Caruso (1987)	HIGH/ PO	458	-0.20	(-0.39, -0.02)
Pooled Random Effects Estimate			-0.07(1)	(-0.52, 0.38)

(1)

Chi-squared test of heterogeneity p-value = 0.064.

CI: confidence intervals; IV: intravenous; PO: oral administration.

We note that the fixed effects pooled estimate of -0.17 (95% CI [-0.34, 0.00]; p=0.044) is just barely significant. The Chi-squared test of heterogeneity, with a p-value of less than 0.10, indicates significant heterogeneity among the studies. In light of the considerable heterogeneity that exists across these studies, a pooled estimate may not be appropriate. Therefore, we discuss the studies separately. Two of the studies show a decrease in pain associated with placebo (a decrease of 20 percent and 40 percent of a standard deviation in the Caruso and colleagues study and at site A of the Bradley and colleagues study, respectively). Site B of the Bradley study does not show a placebo effect (an increase of 60 percent of a standard deviation).

The two sites of the Bradley study (Bradley, Flusser, Katz, et al., 1994) show opposite and almost equal effects. The populations at the two sites differed, with less severely ill patients at site A. In addition, site B had significant problems with randomization. Four of the eight measures of clinical status were significantly different between the sites at baseline, with the SAMe patients generally being more severely affected. Thus, at site B, the results may be affected by adverse assignment. It is likely that the results of the much larger study by Caruso and colleagues more closely represent the effect of SAMe relative to placebo (i.e., a modest but statistically significant effect of SAMe) than do the results of the Bradley study.

Meta-Analysis Of SAMe Versus NSAID

Nine studies compared SAMe to NSAID. One of these was a large study that also included a SAMe versus placebo arm comparison. To identify appropriate comparisons to include in the analysis, we stratified the interventions by dose.

Table 11. Study distribution by NSAID dose and SAMe (more...)

Table 11. Study distribution by NSAID dose and SAMe dose

	SAMe Dose
NSAID Dose	LOW	MED	HIGH	Total
LOW	1	1	2	4
MED	0	0	4	4
HIGH	0	0	1	1
TOTAL	1	1	6	9

For clinical reasons, we dropped the study that compared the effect of a low SAMe dose with that of a low NSAID dose (Polli, Cortellaro, Parrini, et al., 1975) from our pooled analysis. This study compared the effect of SAMe to that of Ibuprofen and had an effect size favoring SAMe of -0.61 (95% CI [-1.34, 0.12]), with a total of 30 patients for whom pain was measured at 14 days. However, we judged that the doses were subtherapeutic (indomethacin [Indocin], 50 mg and SAMe, 30 mg) and that, as a result, these findings were not applicable to clinical use.

Table 12. SAMe versus NSAID pooled analysis

Table 12. SAMe versus NSAID pooled analysis

First Author & Year	NSAID	NSAID Dose	SAMe Dose & Route of Administration	Total N	Effect Size	95% CI
Capretto (1985)	Ibuprofen	LOW	MED/ PO	21	-0.58	( 1.46, 0.29)
Caroli (1980)	Aspirin	MED	HIGH/ PO	54	-1.90	(-2.54, -1.25)
Caruso (1987)	Naproxen	MED	HIGH/ PO	453	0.01	(-0.17, 0.19)
Domljan (1989)	Naproxen	MED	HIGH/ PO	37	-0.15	(-0.80, 0.49)
Maccagno (1987)	Piroxicam	HIGH	HIGH/ PO	45	1.07	( 0.44, 1.69)
Marcolongo (1985)	Ibuprofen	LOW	HIGH/PO	150	-0.06	(-0.38, 0.26)
Muller-Fassbender (1987)	Ibuprofen	LOW	HIGH/ PO	36	-0.11	(-0.77, 0.54)
Vetter (1987)	Indocin	MED	HIGH/ PO	35	0.78	( 0.09, 1.47)
		Pooled Random Effects Estimate			-0.11(1)	(-0.56, 0.35)

(1)

Chi-squared test of heterogeneity p-value < 0.001.

CI: confidence intervals; PO: oral administration.

Figure 10. OA Outcome: Pain at Four Weeks Follow-up (more...)

   Figure 10. OA Outcome: Pain at Four Weeks Follow-up SAMe (Medium/High Dose) Versus NSAIDS

Figure 10

The heterogeneity among these eight studies is large, and the pooled estimate is not significant (p=0.65). A sensitivity analysis was undertaken to attempt to account for the observed variance.

Publication Bias

Figure 11. OA Outcome: Pain at Four Weeks Follow-up (more...)

   Figure 11. OA Outcome: Pain at Four Weeks Follow-up Begg's Funnel Plot with Pseudo 95% Confidence Limits

The number of SAMe versus placebo studies (three) was too small to allow a meaningful assessment of publication bias. Thus, we assessed publication bias only for the eight SAMe versus NSAID studies. Neither the adjusted rank correlation test (p=0.54), the regression asymmetry test (p=0.77), nor a visual inspection of the funnel plot (see Figure 11) indicated the presence of publication bias. The small number of available studies limited our ability to determine if publication bias did occur.

Analysis Of Studies Of Cholestasis Of Pregnancy

The Pruritus Outcome

Study inclusion/exclusion. We considered nine studies for inclusion in the pruritus meta-analysis. One study had four arms (placebo, SAMe, ursodeoxycholic acid [UDCA], and SAMe plus UDCA). A second study had three arms (placebo and two SAMe arms of different doses). The remaining six studies had two arms (SAMe and placebo).

One study (Floreani, Paternoster, Melis, et al., 1996b) was removed from the analysis and the evidence table, because it was an abstract that contained the same data as another full-length study that was considered for the analysis (Floreani, Paternoster, Melis, et al., 1996a). However, the study itself was eventually excluded from the meta-analysis, because it did not contain sufficient statistics for an effect-size calculation. Two studies (Frezza, Centini, Cammareri, et al., 1990; Frezza, Cammareri, Di Padova, et al., 1987) reported the results of studies on the same study population, but each reported a different one of the two outcomes of interest; the study reporting pruritus (Frezza, Centini, Cammareri, et al., 1990), was included in this analysis. The other study was included in the bilirubin analysis.

As a result of these exclusions, six studies were included in the analysis. As mentioned above, one study (Nicastri, Diaferia, Tartagni, et al., 1998) had four arms: placebo, SAMe, UDCA, and SAMe plus UDCA. Two studies (Frezza, Pozzato, Chiesa, et al., 1984; Lafuenti, Plotti, Nicolanti, et al., 1988) had three arms: placebo and two SAMe arms of different doses. A fourth study (Roncaglia, Locatelli, Bellini, et al., 2000) had two arms: UDCA and SAMe. A fifth study (Frezza, Centini, Cammareri, et al., 1990) had placebo and SAMe arms, and a sixth study (Ribalta, Reyes, Gonzalez, et al., 1991) had mannitol (acting as a placebo) and SAMe arms. We sought outcome measurements closest to the time of delivery. Actual times of data collection ranged from the time of delivery (implied in one study) to 20 days post-delivery (reported in several studies).

Calculations for pruritus. Two studies that measured pruritus as an outcome did not report a measure of uncertainty. For the first study (Nicastri, Diaferia, Tartagni, et al., 1998), we assumed that the standard deviation equaled one-quarter of the theoretical range of the pruritus scale. For the second study (Roncaglia, Locatelli, Bellini, et al., 2000), we back-calculated the standard deviation from the reported p-value. The pruritus effect size was calculated as described previously. A negative effect size indicates that SAMe is associated with decreased pruritus as compared to the UDCA or placebo as appropriate. We categorized SAMe levels as low, medium, and high, as described previously.

Table 13. Effect sizes for interventions excluded from (more...)

Table 13. Effect sizes for interventions excluded from SAMe versus placebo pooled analysis (pruritus)

First Author & Year	SAMe Dose	Total N	Effect Size	95% CI
Frezza (1984)	MED	12	-0.10	(-1.23, 1.03)
Lafuenti (1988)	HIGH	29	-1.39	(-2.24, -0.55)

Meta-analysis of SAMe versus placebo. Five studies compared the effect of SAMe to that of a placebo (or mannitol, which we considered equivalent to a placebo). Two of the five studies had more than one SAMe arm. To ensure that placebo patients within one study would not be double counted, we dropped the most disparate SAMe arm from the pooled analysis so that each study contributed only one effect size to our combined analysis. For the study by Frezza and colleagues (Frezza, Pozzato, Chiesa, et al., 1984), we dropped the SAMe arm that received 200 mg of SAMe, and for the study by Lafuenti and colleagues (Lafuenti, Plotti, Nicolanti, et al., 1988), we dropped the SAMe arm that received 1800 mg. The effect sizes for these two arms are shown in Table 13.

Table 14. SAMe versus placebo pooled analysis (pruritus) (more...)

Table 14. SAMe versus placebo pooled analysis (pruritus)

First Author & Year	Comparison Arm	SAMe Dose	Total N	Effect Size	95% CI
Frezza (1990)	Placebo	HIGH	30	-0.77	(-1.51, -0.03)
Frezza (1984)	Placebo	HIGH	12	-0.94	(-2.13, 0.25)
Lafuenti (1988)	Placebo	MED	26	-1.18	(-2.11, -0.26)
	No treatment
Nicastri (1998)	Placebo	HIGH	16	-0.19	(-1.17, 0.79)
Ribalta (1991)	Mannitol	HIGH	18	-1.87	(-2.98, -0.76)
	Pooled Random Effects Estimate			-0.95(1)	(-1.45, -0.45)

(1)

Chi-squared test of heterogeneity p-value = 0.25.

Figure 12. Cholestasis of Pregnancy Outcome: Pruritus (more...)

   Figure 12. Cholestasis of Pregnancy Outcome: Pruritus SAMe Versus Placebo

The pooled analysis for SAMe versus placebo is shown in Table 14 and in Figure 12.

The Chi-squared test of heterogeneity does not detect heterogeneity among these studies. However, given the low power of this test, we still report the pooled random effects estimate of -0.95 (p<0.001). This pooled estimate shows that, in these studies, SAMe is associated with a decrease of almost a full standard deviation in the appearance of pruritus as compared with placebo (or mannitol).

Table 15. SAMe versus ursodeoxycholic acid (pruritus) (more...)

Table 15. SAMe versus ursodeoxycholic acid (pruritus)

First Author & Year	Comparison Arm	SAMe Dose	Total N	Effect Size	95% CI
Nicastri (1998)	Ursodeoxycholic acid	HIGH	16	-0.47	(-1.47, 0.52)
Roncaglia (2000)	Ursodeoxycholic acid	HIGH	22	-1.52	(-2.50, -0.54)

Meta-analysis of SAMe versus UDCA (pruritus). UDCA is the standard therapy for cholestasis of pregnancy. Two studies compared the effect of SAMe to that of this form of active therapy. One of these studies had three arms: SAMe alone, UDCA alone, and UDCA plus SAMe. The results of the analysis comparing UDCA alone to SAMe alone are shown in Table 15. The total sample size reported in the table is the sum of the sample sizes in the two arms that were used to calculate the effect size.

Since we identified only two studies that compared UDCA with SAMe, we did not pool the studies and instead discuss the results descriptively. Data from both studies support significant benefit for SAMe and UDCA compared to baseline (although both studies showed a preference for UDCA over SAMe for treatment of elevated serum bilirubin, as we will discuss later). In the Roncaglia study (Roncaglia, Locatelli, Bellini, et al., 2000), SAMe appeared more effective than UDCA in treating pruritus (p=0.02). Further, the combination of SAMe and UDCA had a greater clinical effect on both outcomes than did UDCA alone; again, the effect on pruritus was larger than the effect on serum bilirubin (p<0.0003; p<0.002).

Figure 13. Cholestasis of Pregnancy Outcome: Pruritus (more...)

   Figure 13. Cholestasis of Pregnancy Outcome: Pruritus Begg's Funnel Plot with Pseudo 95% Confidence Limits

Publication bias. We assessed publication bias only for the five studies that entered our SAMe versus placebo pooled analysis. Neither the adjusted rank correlation test (p=0.81), the regression asymmetry test (p=0.58), nor a visual inspection of the funnel plot indicated publication bias (see Figure 13). However, the small number of studies limited our ability to determine if publication bias did occur.

The Serum Bilirubin Outcome

Study inclusion/exclusion. We considered the same nine studies for inclusion in the cholestasis serum bilirubin meta-analysis as for the analysis of pruritus as an outcome. One publication (Floreani, Paternoster, Melis, et al., 1996b) was excluded, because it was an abstract that contained the same data as another full-length publication (Floreani, Paternoster, Melis, et al., 1996a); moreover, these data had been determined to be insufficient for effect size calculations. Two publications (Frezza, Centini, Cammareri, et al., 1990; Frezza, Cammareri, Di Padova, et al., 1987) reported on the same study population, but each reported only one of the two outcomes of interest. Thus, the study that reported the effects on serum bilirubin (Frezza, Cammareri, Di Padova, et al., 1987) was included in this analysis. As a result of these exclusions, seven studies were included in the analysis. One study (Nicastri, Diaferi, Tartagni, et al., 1998) had four arms: placebo, SAMe, UDCA, and SAMe plus UDCA. Two studies (Frezza, Pozzato, Chiesa, et al., 1984; Lafuenti, Plotti, Nicolanti, et al., 1988) had three arms each: placebo (no treatment) and two SAMe arms of different doses. Two studies (Floreani, Paternoster, Melis, et al., 1996a; Roncaglia, Locatelli, Bellini, et al., 2000) had two arms that consisted of UDCA and SAMe. One study (Frezza, Pozzato, Pison, et al., 1987) had two arms of placebo and SAMe. Finally, one study (Ribalta, Reyes, Gonzalez, et al., 1991) had two arms consisting of mannitol and SAMe.

Most studies reported total serum bilirubin in milligrams per deciliter (mg/dl), which was the preferred outcome. When results were not expressed in that way, other measures were considered. For two studies (Floreani, Paternoster, Melis, et al., 1996a; Frezza, Cammareri, Di Padova, et al., 1987), total bile salts and total bile acids both measured in millimicromoles per liter were used as the outcome, respectively. Although these outcomes are not equal, each varies proportionately with the total serum bilirubin. Therefore, they may be used to calculate a common effect size.

Calculations for serum bilirubin.Two studies did not report a measure of uncertainty. In one study, whose outcome measure was total bile salts (in μ-mol/l), one arm did not have an associated standard deviation, so we imputed the standard deviation from the other arm and another study (Frezza, Cammareri, Di Padova, et al., 1987) that did report a standard deviation for the same outcome measure in the same units. For the second study (Nicastri, Diaferi, Tartagni, et al., 1998), whose outcome measure was serum bilirubin (in mg/dl), we imputed the standard deviation from those four studies (Ribalta, Reyes, Gonzalez, et al., 1991; Lafuenti, Plotti, Nicolanti, et al., 1988; Frezza, Pozzato, Chiesa, et al., 1984; Roncaglia, Locatelli, Bellini, et al., 2000) that did report a standard deviation for serum bilirubin in the same units. The serum bilirubin effect size was calculated as described previously. A negative effect size indicates that SAMe is associated with decreased serum bilirubin as compared to the effect of UDCA, mannitol, or placebo, as appropriate. As described previously, we categorized SAMe levels as low, medium, and high.

Table 16. Effect sizes for interventions excluded from (more...)

Table 16. Effect sizes for interventions excluded from SAMe versus placebo pooled analysis (serum bilirubin)

First Author & Year	SAMe Dose	Total N	Effect Size	95% CI
Frezza (1984)	MED	12	-0.50	(-1.65, 0.65)
Lafuenti (1988)	HIGH	29	-1.34	(-2.18, -0.50)

Meta-analysis of SAMe versus placebo. Five studies compared the effect of SAMe to that of a placebo (or mannitol, which we considered equivalent to placebo). Two of the five studies had more than one SAMe arm. As in the pruritus analysis, we dropped the most disparate SAMe arm from the pooled analysis, so that our combined analysis would include only one effect size contributed by each study, to ensure that placebo patients in a study were not double-counted. For the study by Frezza, Pozzato, Chiesa, et al. (1984), we dropped the SAMe arm that received 200mg of SAMe, and for study by Lafuenti, Plotti, Nicolanti, et al. (1988), we dropped the SAMe arm that received 1800mg. The effect sizes for these two arms are shown in Table 16.

Table 17. SAMe versus placebo pooled analysis (serum (more...)

Table 17. SAMe versus placebo pooled analysis (serum bilirubin)

First Author & Year	Comparison Arm	SAMe Dose	Total N	Effect Size	95% CI
Frezza (1984)	Placebo	HIGH	12	-1.58	(-2.88, -0.29)
Frezza (1987)	Placebo	HIGH	30	-1.91	(-2.78, -1.05)
Lafuenti (1988)	Placebo, No treatment	MED	26	-1.22	(-2.15, -0.29)
Nicastri (1988)	Placebo	HIGH	16	-1.15	(-2.21, -0.09)
Ribalta (1991)	Mannitol	HIGH	18	-0.70	(-1.65, 0.25)
	Pooled Random Effects Estimate			-1.32(1)	(-1.76, -0.88)

(1)

Chi-squared test of heterogeneity p-value = 0.44.

Figure 14. Cholestasis of Pregnancy Outcome: Serum (more...)

   Figure 14. Cholestasis of Pregnancy Outcome: Serum Bilirubin SAMe Versus Placebo

The pooled analysis for SAMe versus placebo is shown in Table 17 and in Figure 14.

The Chi-squared test of heterogeneity does not detect heterogeneity among the studies. However, given the low power of this test, we still report the pooled random effects estimate of -1.32 (p<0.001). This pooled estimate shows that in these studies, SAMe is associated with a decrease of approximately 1.3 standard deviations in serum bilirubin as compared with placebo (or mannitol).

Table 18. SAMe versus ursodeoxycholic acid (serum bilirubin) (more...)

Table 18. SAMe versus ursodeoxycholic acid (serum bilirubin)

First Author & Year	Comparison Arm	SAMe Dose	Total N	Effect Size	95% CI
Floreani (1996)	Ursodeoxycholic acid	HIGH	20	0.92	( 0.00, 1.84)
Nicastri (1998)	Ursodeoxycholic acid	HIGH	16	-0.13	(-1.12, 0.85)
Roncaglia (2000)	Ursodeoxycholic acid	HIGH	22	1.34	( 0.39, 2.30)

Meta-analysis of SAMe versus ursodeoxycholic acid. Three studies contained UDCA arms, including one study that had both an UDCA-only arm and a SAMe plus UDCA arm. The effect sizes for the UDCA-only arm are shown in Table 18. The total sample size reported in the table is the sum of the sample sizes in the two arms that were used to calculate the effect size.

Since we identified only three studies that fit the criteria for inclusion, we did not pool the studies and instead discuss the results descriptively. One study reported no difference between SAMe and UDCA, and had very wide 95 percent confidence intervals. The other two studies reported UDCA as superior to SAMe, with an effect size favoring UDCA of about one standard deviation.

Sensitivity analysis. We conducted one sensitivity analysis on our SAMe versus placebo comparison. When we dropped the study that reported total bile acids (Frezza, Cammareri, Di Padova, et al., 1987) from the pooled analysis, the result was not significantly affected (pooled random effects estimate of -1.11 (95% CI [-1.62, -0.59]; p<0.001)).

Figure 15. Cholestasis of Pregnancy Outcome: Serum (more...)

   Figure 15. Cholestasis of Pregnancy Outcome: Serum Bilirubin Begg's Funnel Plot with Pseudo 95% Confidence Limits

Publication bias. We assessed publication bias for the five studies that entered our SAMe versus placebo pooled analysis. Neither the adjusted rank correlation test (p=0.81), the regression asymmetry test (p=0.87), nor a visual inspection of the funnel plot indicated publication bias (see Figure 15). However, the small number of studies limited our ability to determine whether publication bias did occur.

Analysis Of Studies Of Intrahepatic Cholestasis For Conditions Associated With Liver Disease (Other Than Pregnancy)

Twelve studies addressed intrahepatic cholestasis associated with a variety of liver disorders other than pregnancy. We judged that intrahepatic cholestasis is likely to be a “final common pathway” for many chronic liver conditions, and, therefore, that the studies were sufficiently clinically similar to justify statistical pooling. As we did for the previous cholestasis-of-pregnancy meta-analysis, the two outcomes we chose for pooling were pruritus and bilirubin. All outcomes were measured as closely as possible to 14 days, and follow-up times less than 10 days were considered too short to be included.

With respect to patient heterogeneity, we restricted our primary attention to the results for patients with chronic liver disease and excluded acute patients. Most studies either focused solely on chronic patients or presented results separately for the two groups. The single study that had a mix of patients whose disease status could not be discerned is noted below in the description of the results for the pruritus outcome. We report study-level data from two studies on acute patients, and discuss each outcome separately.

The Pruritus Outcome

Study inclusion/exclusion. Unlike the studies included in the cholestasis for pregnancy meta-analysis, only one of the twelve studies included in the cholestasis of chronic liver disease meta-analysis contained a mean pruritus score. Therefore, we used as our common statistic the risk ratio for resolved pruritus. For this statistic, the risk in the SAMe group is the number of patients in the SAMe group whose pruritus resolved divided by the number of patients in the SAMe group who began the study with pruritus. The risk in the comparison group, either placebo or other drug, is defined analogously. As a result, the risk ratio, which divides the risk in the comparison group by the risk in the SAMe group, is less than one if the risk of resolved pruritus is less in the comparison group, thereby indicating that SAMe is associated with a higher level of resolved pruritus. We defined the risk ratio in this manner so that a risk ratio less than one would indicate a better outcome associated with SAMe for all risk ratios across all conditions.

We considered 12 studies for inclusion in the intrahepatic cholestasis meta-analysis. Studies could be excluded if they represented data duplicated in another included study, did not have appropriate outcome measures, presented insufficient statistics, or had a crossover design.

Two studies (Manzillo, Frezza, Fiaccadori, et al., 1989); and Manzillo, Fiaccadori, Frezza, et al., 1988) were excluded, because they included duplicate data that were reported in another included study (Frezza, Surrenti, Manzillo, et al., 1990). Two studies (Adachi, Nanno, Kanbe, et al., 1986; and Bombardieri, Milani, Bernardi, et al., 1985) did not measure pruritus. Two studies (O'Donohue, Wendon and Williams, 1996; and Fiaccadori, Frezza and Di Padova, 1988) reported data that were insufficient for statistical analysis. Finally, one study had a crossover design (Cacciatore, Varriale, Cozzolino, et al., 1989): All patients were treated with a placebo before being switched to SAMe. Thus, this study was excluded. Another study (Bray, Di Padova, Tredger, et al., 1991) with a crossover design did not report data prior to the crossover and was also excluded. As a result of these exclusions, four studies were included in the analysis.

Table 19. SAMe versus placebo pooled analysis for a (more...)

Table 19. SAMe versus placebo pooled analysis for a mix of chronic and acute patients (pruritus)

Study	SAMe Dose	Total N	Risk Ratio	95% CI RR
Frezza (1990)	HIGH	81	0.43	(0.27, 0.69)
Frezza (1987)	HIGH	143	0.00	(0.25, 0.53)
Manzillo (1992)	HIGH	105	0.44	(0.30, 0.66)
Qin (2000)	HIGH	64	0.57	(0.39, 0.83)
Pooled Random Effects Estimate			0.45(1)	(0.37, 0.55)

(1)

Chi-squared test of heterogeneity p-value=0.40

Figure 16. Cholestasis Outcome: Pruritus SAMe Versus (more...)

   Figure 16. Cholestasis Outcome: Pruritus SAMe Versus Placebo

Meta-analysis of SAMe versus placebo. All four studies included in the analysis compared the effects of SAMe to that of placebo. The results of one study were presented only for a mix of chronic and acute patients (Manzillo, Piccinino, Surrenti, et al., 1992). We could extract data on chronic patients only for the other three studies. All four studies administered high doses of SAMe. The risk ratios and pooled analysis are shown in Table 19 and Figure 16.

The pooled estimate is significantly smaller than 1 (p<0.001), which indicates a higher risk of resolved pruritus associated with SAMe, as compared to the risk of resolved pruritus for placebo patients. Thus, SAMe was judged to be more effective than placebo in reducing pruritus.

Figure 17. Cholestasis Outcome: Pruritus Begg's Funnel (more...)

   Figure 17. Cholestasis Outcome: Pruritus Begg's Funnel Plot with Pseudo 95% Confidence Limits

Publication bias. Neither the adjusted rank correlation test (p=0.73), the regression asymmetry plot (p=0.76), nor a visual inspection of the funnel plot indicated a presence of publication bias (see Figure 17). However, the small number of studies limited our ability to determine if publication bias did occur.

The Serum Bilirubin Outcome

Study inclusion/exclusion for the serum bilirubin outcome. The serum bilirubin calculations for the analysis of studies of cholestasis of chronic liver disease were done as for the previous cholestasis of pregnancy analysis. Two studies (Manzillo, Frezza, Fiaccadori, et al., 1989; and Manzillo, Fiaccadori, Frezza, et al., 1988) were excluded because they included duplicate data that were reported in an included study (Frezza, Surrenti, Manzillo, et al., 1990). Two studies (Fiaccadori, Frezza, Di Padova, 1988; and O'Donohue, Wendon and Williams, 1996) reported insufficient statistics for the calculations to be performed. One study (Frezza and Di Padova, 1987) was excluded because it did not report the relevant outcome. One study (Cacciatore, Varriale, Cozzolino, et al., 1989) was excluded because it had a crossover design in which patients received the placebo prior to receiving SAMe. Finally, one study (Bray, Di Padova, Tredger, et al., 1991) was excluded because it included a crossover design and did not report data prior to the crossover. As a result of these exclusions, five studies were included in the analysis.

Table 20. SAMe versus placebo pooled analysis for cholestasis (more...)

Table 20. SAMe versus placebo pooled analysis for cholestasis in patients with chronic liver disease (serum bilirubin)

First Author & Year	SAMe Dose	Total N	Effect Size	95% CI
Adachi (1986)	HIGH	14	-0.01	(-1.07, 1.05)
Bombardieri	HIGH	18	-0.69	(-1.64, 0.26)
Frezza (1990)	HIGH	209	-0.42	(-0.70, -0.15)
Manzillo (1992)	HIGH	227	-0.18	(-0.44, 0.08)
Qin (2000)	HIGH	64	-1.78	(-2.36, -1.20)
Pooled Random Effects Estimate			-0.63(1)	(-1.16, -0.10)

(1)

Chi-squared test of heterogeneity p-value<0.001.

Figure 18. Cholestasis Outcome: Serum Bilirubin SAMe (more...)

   Figure 18. Cholestasis Outcome: Serum Bilirubin SAMe Versus Placebo

Meta-analysis of SAMe versus placebo. All five studies compared the effects of SAMe to that of placebo. We could extract data on chronic patients for all five studies, and all administered high doses of SAMe. The effect sizes and pooled analysis are shown in Table 20 and Figure 18.

The pooled estimate is significantly smaller than 0 (p=0.02), indicating a statistically significant decrease of over one standard deviation in serum bilirubin associated with SAMe treatment. However, the studies are significantly heterogeneous.

Figure 19. Cholestasis Outcome: Serum Bilirubin Begg's (more...)

   Figure 19. Cholestasis Outcome: Serum Bilirubin Begg's Funnel Plot with Pseudo 95% Confidence Limits

Publication Bias. Neither the adjusted rank correlation test (p=0.81), the regression asymmetry plot (p=0.49), nor a visual inspection of the funnel plot indicated a presence of publication bias (see Figure 19). However, the small number of studies limited our ability to determine if publication bias did occur.

Results For Acute Patients

Table 21. SAMe versus placebo effect sizes for patients (more...)

Table 21. SAMe versus placebo effect sizes for patients with acute liver disease (serum bilirubin)

First Author & Year	SAMe Dose	Total N	Effect Size	95% CI
Manzillo (1992)	HIGH	82	-0.50	(-0.95, -0.05)
Qin (2000)	HIGH	46	-1.42	(-2.07, -0.78)

The results of the two studies, each of which administered a high SAMe dose, showed large, statistically significant benefits for SAMe compared to placebo.

Table 22. SAMe versus prednisolone effect size for (more...)

Table 22. SAMe versus prednisolone effect size for patients with chronic liver eisease (serum bilirubin)

First Author & Year	Drug	Drug Dose	SAMe Dose	Total N	Effect Size	95% CI
Adachi (1986)	Prednisolone	MED	HIGH	15	0.19	(-0.85, 1.22)

This study has a small sample size and its power is extremely low. There is no statistically significant effect. The point estimate of the effect suggests at most a small difference.

Liver Studies Not Considered For Meta-Analysis

Twenty studies that included patients with a wide variety of liver diseases were found to be too heterogeneous for inclusion in any pooled analysis. The number and variety of disease states included hepatitis of mixed etiology and acuteness (eight studies); cirrhosis of the liver from a variety of causes including viral infections, autoimmune disease, and alcohol (seven studies); chronic liver disease without cholestasis or cirrhosis (four studies); and liver transplant (one study). In addition, these studies showed some variation in the outcomes measured, such that no group of studies that shared a homogeneous disease state and outcome measure was large enough to justify a pooled analysis.

Table 23. Distribution of Jadad score in other liver (more...)

Table 23. Distribution of Jadad score in other liver studies

	Jadad scores
Diagnosis	0	1	2	3	4	5
Cirrhosis	2	2	1	1	-	1
Hepatitis	4	1	3	-	-	-
Chronic liver disease	4	-	-	-	-	-
Transplantation	-	-	1	-	-	-
TOTAL	10	3	5	1	0	1

Studies Of The Effects Of SAME On Depression

For depression, a systematic review with meta-analysis (Janicak, Lipinski, Davis et al., 1988 and 1989) and a prior systematic review (Bressa, 1994) were identified. Janicak (Janicak, Lipinski, Davis et al., 1988 and 1989) reviewed both case series and comparison trials and concluded that SAMe showed greater efficacy than placebo and a slight preference for or at least comparable efficacy with SAMe relative to tricyclic antidepressants. The systematic review by Bressa (Bressa, 1994) reached the same conclusion. Both studies analyzed percentage of partial and full responders, which corresponds to our risk-ratio analyses. In general, we were able to calculate risk ratios for fewer studies than Bressa or Janicak, because both reported on unpublished data that were not available to us (some of the unpublished data appeared to come from the manufacturer). In contrast to the analyses of Bressa and Janicak, we also calculated an effect size for the HRSD measured continuously. The results of this analysis supported conclusions similar to those noted above. Bressa (1994) also identified, as we did, significant heterogeneity in the studies pooled for analysis. Neither Bressa nor Janicak performed an analysis of the frequency of side effects of SAMe compared to either placebo or other active antidepressants.

Another study was identified as a “meta-analysis” (Delle Chaie and Boissard, 1997); however, it included data from only two clinical trials, which were included in our meta-analysis. Additional review articles were identified (Friedel, Goa, and Benfield, 1989; Carney, Toone, and Reynolds, 1987; Brown, Gerbarg, and Bottiglieri, 2000; Brown and Gerbarg, 2001; Moreno and Ortiz, 1991; Tramoni and Azorin, 1988) that reviewed the literature in a nonsystematic way and were generally supportive of SAMe as an effective treatment for depression.

Studies Of The Effects Of SAMe On Osteoarthritis

Two reviews were identified that focused on the use of SAMe for osteoarthritis. A relatively limited review by Schumacher (1987) concentrated on the potential mechanism of action of SAMe, its effect on the gastrointestinal tract, and the apparently smaller number of side effects for SAMe than for standard NSAID medications, rather than the efficacy of the therapy. Our analysis is more congruent with a systematic review of controlled and uncontrolled trials performed by Di Padova (1987), which concluded that SAMe improved both subjective and objective symptoms of osteoarthritis. Further, he noted that SAMe had a slower onset of action than NSAIDs, with less efficacy for SAMe demonstrated at 2 weeks and equivalent efficacy at 4 weeks. He also asserted that SAMe is well tolerated, but no data are presented to support this assertion. In our analysis, we were able to pool data for subjective symptoms only, i.e., pain. We were able to demonstrate a modest preference for SAMe versus placebo; however the small number of studies and the marked heterogeneity within one study (Bradley, 1994) limit the assurance with which this conclusion may be asserted. In comparison with NSAIDs, SAMe appeared to be approximately equivalent in efficacy for pain, according to our analysis.

Studies Of The Effects Of SAMe On Liver Disease

In our analysis of the effects of SAMe on cholestasis for liver disease, we treated studies involving the intrahepatic cholestasis of pregnancy separately from studies of cholestasis due to other liver diseases. This treatment allowed us to reduce the clinical heterogeneity of the studies as a group sufficient to justify pooled analyses. One other meta-analysis by Tambini (Tambini, Fracassetti, Minola et al., 1997) was identified that focused solely on the intrahepatic cholestasis of pregnancy. A pooled analysis of three studies concluded that SAMe may be an effective symptomatic treatment for intrahepatic cholestasis of pregnancy because it reduced both bilirubin levels and pruritus scores. Only placebo-controlled studies were included. Our analysis included more studies and also compared SAMe to active therapy (UDCA). We concluded, as did Tambini, that SAMe was more effective than placebo. In our analysis, SAMe was likely inferior to UDCA for the treatment of pruritus and reduction of bilirubin.

Two meta-analyses and one systematic review were identified that evaluated combined data on cholestasis in both pregnant and other liver disease patients. The review by Coltorti, Bortoline, and Di Padova (1990) considered both open and placebo-controlled trials and segregated trials involving cholestasis of pregnancy from other liver diseases. Cholestasis of pregnancy trials that included active treatment were not reviewed. These authors concluded that SAMe was efficacious for relieving subjective symptoms and laboratory abnormalities in all types of intrahepatic cholestasis. Meta-analyses by Frezza and Terpin (1992) and by Frezza (1993) pooled data from pregnancy and other liver disease studies. In addition, they focused on SAMe versus placebo studies only. Both analyses concluded that SAMe was superior to placebo for cholestasis in decreasing both pruritus and bilirubin as well as improving other biochemical parameters such as liver enzymes. Comparison of our analyses with these two meta-analyses is somewhat hampered by the differences in studies included. We did not combine the intrahepatic-cholestasis-of-pregnancy patients with other liver-disease patients. However, our general conclusion for both patient populations is similar to conclusions drawn by these authors: SAMe is superior to placebo both to relieve symptoms (pruritus) and to decrease bilirubin. A systematic analysis of side effects was not done in any of the analyses discussed.

We also identified two additional nonsystematic reviews (Almasio, Bortolini, Pagliaro, et al., 1990; Chawla, Bonkovsky and Galambos, 1990). Chawla principally focused on the likely mechanism of action of SAMe, while Almasio also described preclinical data. SAMe was considered by these authors to have utility in the treatment of liver disease in general (Chawla, Bonkovsky and Galambos, 1990) and intrahepatic cholestasis in particular (Almasio, Bortolini, Pagliaro, et al., 1990).

Di Padova (Di Padova, Boissard and Brunetti, et al., 1996) performed a multivariate logistic regression analysis of pooled data from five studies of patients with chronic liver disease. Their analysis favored SAMe over placebo. They also identified the presence of a history of alcohol abuse, ascites, prior treatment, and elevated bilirubin as predictors of a positive effect.

Summary Of Comparisons With Other Analyses

Thus, the results of our meta-analyses for the four conditions examined are generally in alignment with previously published reviews and meta-analyses. In some instances, we made different choices regarding study inclusion or exclusion as well as which parameters to analyze, but these differences did not affect the broad consensus that SAMe appears to be more effective than placebo for the conditions studied and to be roughly equivalent to [standard therapy for depression and osteoarthritis]. No other analyses were identified that compared active therapy for cholestasis of pregnancy; thus, we have no comparison for our conclusion that SAMe is not preferred to standard therapy for that condition.

Further, our analysis for depression, osteoarthritis, and liver disease included studies which were published subsequent to the last systematic review. We also performed more analyses for depression (risk ratios of percent decrease in HRSD scores) than Janicak or Bressa did. We did not analyze pooled data on transaminase enzyme response to SAMe, but we did analyze both pruritus and serum bilirubin for the two liver conditions studied.

Cholestasis Of Pregnancy

The data in our review support the following conclusions regarding the use of SAMe for the treatment of cholestasis of pregnancy:

• SAMe is likely more effective than placebo in reducing pruritus and bilirubin in women with cholestasis of pregnancy.

• SAMe appears no better than and may actually be worse than ursodeoxycholic acid, the conventional treatment for reducing pruritus and bilirubin in women with cholestasis of pregnancy.

Intrahepatic Cholestasis Associated With Liver Disease

The data in our review support the following conclusions regarding the use of SAMe for the treatment of intrahepatic cholestasis associated with liver disease:

• SAMe is more efficacious than placebo in reducing pruritus and serum bilirubin in patients with intrahepatic cholestasis associated with liver disease.

• Because of an insufficient number of studies, no conclusions can be drawn about the efficacy of SAMe compared to conventional therapy in treating pruritus and bilirubin in patients with intrahepatic cholestasis associated with liver disease.