A study given considerable well-deserved publicity has been published recently in the Journal of Clinical Oncology: Menon U, et al. Risk algorithm using serial biomarker measurements doubles the number of screen-detected cancers compared with a single-threshold rule in the United Kingdom Collaborative Trial of Ovarian Cancer Screening. JCO; published online on May 11, 2015; DOI:10.1200/JCO.2014.59.4945.
This study investigated the impact on cancer detection rates of using serial cancer marker results. Annual serum cancer antigen 125 (CA-125) concentration was interpreted with the risk of ovarian cancer algorithm (ROCA). Women were triaged by the ROCA: normal risk – returned to annual screening, intermediate risk – repeat CA-125, and elevated risk – repeat CA-125 and trans-vaginal ultrasound. ROCA detected 87.1% of the cancers: using fixed CA-125 cut-offs of >35, >30, and >22 U/mL would have identified 41.3%, 48.4%, and 66.5%, respectively. The authors concluded that using ROCA doubled the number of screen-detected invasive epithelial ovarian or tubal cancers (as compared with the traditional widely used fixed decision-making limit of 35 U/mL). It was stated that, in the context of cancer screening, reliance on predefined single-threshold rules may result in biomarkers of value being discarded.
The results of tests used as cancer markers are very often interpreted using single fixed decision-making limits (also termed thresholds or cut-offs). So, why is the ROCA approach, which uses serial CA-125 concentration results, better than interpretation of each annual screening test result against a fixed limit? The answer is actually very simple – biochemical individuality.
Estimates of the within-subject biological variation, the variation of results in individuals around their own homeostatic setting points, and the between-subject biological variation, the difference among the homoeostatic setting points of individuals, have been documented for many tests. There are four studies on CA-125: the median within-subject biological variation is 24.7% and the between-subject variation is 54.6%. Thus, the within-subject biological variation is much less than the between-subject variation and the ratio of these – termed the index of individuality – is 0.45. This means that CA-125 has high individuality.
The ramifications on the interpretation of test results of marked individuality such as that of CA-125 are profound. This is because the range of results from each individual is small compared to the overall range of results from all individuals; in addition, the means of the values from individuals differ. This is shown in this diagram that was created from some data in Gynecol Oncol 1999;74:12–22, a very well-executed basic study on the biological variation of CA-125 and other markers. The figure shows the mean (the dot) and range (the line) of CA-125 concentrations (U/mL) in a set of the healthy female subjects studied, arranged by age (years).
Before reading further, think about how one unusual test result from each individual, a little higher than their individual range of results, would be interpreted. Think, then about how small increases in serial results from each individual would be interpreted.
Individuality is the reason that many commonly-requested tests are less than ideal for diagnosis, case-finding and screening. Individuals can have values that are very unusual for them but, very often, these will lie below the decision-making limit (or within the reference interval – the normal range – if this is used to aid interpretation). In consequence, such results would not be flagged by laboratories on their reports as deserving of further attention. Moreover, the users of the laboratory results would be highly unlikely to pay attention to such unusual values because they would be thought to be “normal” and dismissed. Thus, taking one specimen from an individual and simply comparing the test result with a decision-making limit or reference interval will often not be an effective way of picking up the very small changes seen in early pathological processes. Moreover, often very confusingly for clinicians, a few individuals, with homeostatic setting-points close to the set limit, will have serial results which span the limit and these individuals will have values which change from usual to unusual over time, or vice versa. Further, some will have a homeostatic setting-point above the set limit, so all of their results will be above the limit without disease being present and, often, such individuals will be extensively investigated.
This influence of individuality found from studies on biological variation on the correct application and interpretation of test results has been a subject of great interest to professionals in laboratory medicine: the relevant chapter in a monograph on the subject is recommended reading to all who use the results of investigations in medicine!
It is really important to remember that most tests used in medicine do show marked individuality. In consequence, interpretation of test results using fixed decision-making limits or traditional population-based reference intervals has many difficulties and disadvantages. Following the pattern of serial results from an individual over time and using that individual as their own reference has many merits. This is an important message of relevance to interpretation of results of tests in laboratory medicine, not only of the results of cancer marker measurements. ROCA provides an interesting example of how serial test results can be used to improve cancer screening. The basic approaches of the ROCA study in JCO should be considered by other users of cancer marker tests.
– Professor Callum G Fraser, BSC, PhD, FAACB
Professor Callum G Fraser has worked in laboratory medicine in Scotland and Australia for 45 years. He has published 2 books, 12 book chapters and well over 300 papers, editorials and reviews. He has been involved with bowel screening since 1998, is Consultant to the Scottish Bowel Screening Research Unit, and is a Founding Member of the Expert Working Group on FIT for Screening, WEO.