NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Myers ER, Havrilesky LJ, Kulasingam SL, et al. Genomic Tests for Ovarian Cancer Detection and Management. Rockville (MD): Agency for Healthcare Research and Quality (US); 2006 Oct. (Evidence Reports/Technology Assessments, No. 145.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Genomic Tests for Ovarian Cancer Detection and Management

Genomic Tests for Ovarian Cancer Detection and Management.

Show details

5Future Research

This chapter outlines research priorities identified through the review, both in terms of fundamental gaps in knowledge and in addressing methodological issues in existing studies.

Minimal Data Reporting

We suggest that future studies relevant to screening and diagnosis provide data on, and present results stratified by, the following minimal subject characteristics:

  • Subject age and/or menopausal status;
  • Subject race and ethnicity;
  • Presence or absence of known risk factors for ovarian cancer, particularly family history;
  • For subjects with cancer or adnexal masses, the means by which the mass was initially diagnosed;
  • For subjects with cancer or adnexal masses, the reason for the initial examination which led to diagnosis of a mass: symptoms referable to a mass or ovarian cancer, evaluation for other symptoms, asymptomatic screening for ovarian cancer, or asymptomatic screening for other conditions.

We recognize that, when using large databases for initial analysis, such as those used in many early proteomics studies, such detail may not be available; however, researchers should recognize and discuss the potential biases introduced by these factors.

Test Reproducibility

  • Data on test reproducibility - such as coefficients of variation, inter- and intra-observer agreement, or concordance of results across laboratories - should be consistently reported or referenced.
  • Whenever possible, the potential impact of this reproducibility on test characteristics should be estimated. For example, given a coefficient of variation of some percent, what proportion of test results will fall on the other side of the threshold between positive and negative due to chance alone?
  • The potential impact of reproducibility on interpretation of serial test results should also be estimated where appropriate.

Biological Variability

  • The effect of variation with time, either randomly or in relation to cyclic changes such as the menstrual cycle, should also be reported for tests which have potential use as serial markers.
  • Any variability due to age, menopausal status, or other biological processes should be tested for and noted.

Test-Negative Subjects

  • Since in many studies “control” patients never undergo the reference standard (histological examination of the ovaries), there is the potential for verification bias. Although, given the relatively low incidence of ovarian cancer, the probability of misclassification is fairly low, studies should ideally have some followup on test-negative subjects to ensure that ovarian cancer has not developed within a short time after the test was performed.

Evaluation of Tests

  • Ultimately, tests need to be evaluated based on their intended use and at the stage in the clinical pathway where they will be used. Therefore, potential screening tests must be evaluated in screening settings, with a realistic underlying prevalence of cancer. Similarly, potential diagnostic tests must be tested in settings where there is uncertainty about the diagnosis of ovarian cancer.
  • Ideally, test characteristics for a variety of tests will be compared within the same study population, in order to avoid the inherent difficulties of comparing results across studies. At a minimum, given that the performance characteristics of cancer antigen 125 (CA-125) are well established, new tests should be directly compared to CA-125.
  • Although retrospective studies based on sera or other tissues are useful for establishing estimates of test performance for sample size considerations, new screening and diagnostic tests need to be evaluated prospectively. For example:
    • For screening tests, prospective demonstration of at least one important outcome, such as (a) reduced ovarian cancer-specific mortality, or (b) improved quality of life as documented by a validated instrument. Ideally, this would be done via randomized trials; however, alternative study designs (such as prospective cohort studies with appropriate adjustment for potential confounders) are reasonable for rarer primary outcomes (such as ovarian cancer mortality). In the screening context, given the relatively low positive predictive value of any screening test, documenting the effect of the test on overall quality of life at the population level should be easily demonstrated within the context of a randomized trial.
  • Evaluation of the use of tests in predicting outcomes must ultimately be linked to some change in patient outcomes; at the least, there should be some measure of the value of the information gained from the test result is helpful in some way to the patient. Ideally, the effect of changes in management based on test results should be evaluated in properly designed studies. For example:
    • For tests which appear to reliably predict failure to respond to conventional therapies, studies should prospectively document improved patient outcomes based on this knowledge (such as improved quality of life based on more precise prognosis, or improved quality of life due to avoidance of side effects from ineffective therapy). Ideally, this would be based on randomized trials - patients could be randomized to testing with treatment based on test results, versus no testing; alternatively, testing could be done, with randomized allocation to usual care versus no care for those with test results predicting poor response.
    • For tests which predict greater response to specific agents, improved survival and quality of life need to be documented using randomized trials of those agents in those with specific test results.

Natural History of Ovarian Cancer

  • Underlying assumptions about the natural history of ovarian cancer can have a large effect on the estimated impact of screening compared to other strategies for prevention of ovarian cancer morbidity and mortality. Every effort should be made towards a better understanding of whether ovarian cancer “behaves” like cervical cancer in the sense of progressing through different stages, or whether rapid progression is the most common biological behavior.
  • The implications of these assumptions on the relative efficacy of screening compared to other strategies needs to be evaluated by more sophisticated simulation models.
PubReader format: click here to try


  • PubReader
  • Print View
  • Cite this Page

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...