Prognostic vs Predictive Factors in Oncology Carry Differing Intent but May Have Complementary Clinical Utility

We begin this commentary with commonly accepted definitions: In clinical medicine, a prognostic biomarker is “a clinical or biological characteristic that provides information on the likely patient health outcome irrespective of the treatment.”1 A predictive biomarker is one that “indicates the likely benef it to the patient from the treatment, compared [with] their condition at baseline.”1

Prognostic factors have been evident since the first descriptions of cancer as a clinical entity, with the recognition that patients presenting with distant metastatic disease demonstrated dismal short-term outcomes, whereas long-term survival (even cure) was a realistic possibility if the disease appeared to be localized at diagnosis.

The development of a formal cancer staging system codified an approach for estimating prognosis, with multiple generations of clinicians helping to define the specific clinical features of an individual tumor used in this process. Once the initial status is classified employing these established parameters, it is possible to inform newly diagnosed patients with cancer (analytic cases) regarding the statistical likelihood of survival (5-year) based on large numbers of previously diagnosed individuals with a similar cancer stage.

To be clear, and relevant to the discussion that follows, there is nothing in this staging metric that suggests a particular therapy may improve the prognosis, only that with available standard-of-care options, this is the reported anticipated range of survival outcomes.

For example, a decision may be made to administer systemic chemotherapy after incomplete resection of a large high-grade retroperitoneal sarcoma because of the unfortunately recognized poor outcome, but this decision based on the suggested prognosis does not predict for any clinical benef it associated with the selected management strategy.

Over the years, multiple prognostic factors related to cancers from various anatomic locations and of different cell types have been proposed with many currently widely employed in clinical practice. These include the pathological evaluation of tumor grade, abnormalities of common serum/ plasma, and a variety of immunohistochemical tests, and in recent years, molecular biomarkers as well as increasingly complex prognostic algorithms. The findings may suggest a more- or less-favorable risk of disease recurrence following initial treatment, or a slower or more rapid rate of growth and subsequent tumor spread, features that may substantially influence a patient’s ultimate survival.

Today, although newer prognostic tests and platforms may be quite interesting, lead to impressive publications in high-impact scientific journals,2 or even result in new revenue streams for the developers, one must inquire how much it really matters to an individual patient that it is possible to narrow the statistically suggested risk of disease recurrence, for example, from between 20% to 40% to 25% to 35%. This question is especially relevant if obtaining the test results will be a financial burden to the patient and if the information is not directly helpful to oncologists in their efforts to design an optimal treatment/follow-up plan.

We now turn our focus to the current, rapidly evolving, and increasingly relevant role of predictive factors in cancer management. Again, this is not a new concept. For example, breast cancer cells have been routinely examined for several decades for the presence (or absence) of estrogen and progesterone receptors to predict for the potential benefits of antiestrogen and other hormonal manipulative strategies. Multiple additional predictive factors are currently employed prior to the administration of specific antineoplastic therapeutics. Further, the rapid expansion of molecular characterization efforts has permitted the subsequent development of clinically meaningful biomarkers that help define optimal therapeutic strategies for individual patients with cancer.

In this brief discussion, it is important to note that a single factor may be effectively utilized for both defining prognosis and in the prediction of the value of a specific approach to disease management. A classic example of this is HER2 overexpression in breast cancer. Early seminal research revealed the negative prognostic effect of this biological phenomenon,3 and multiple studies over the subsequent years have demonstrated the clinical utility of therapeutic agents directed at breast cancers with HER2 overexpression.

The combined utility of a biomarker as a clinically relevant prognostic and predictive factor is not limited to molecular tests. Consider, for example, the long-established use of laboratory testing for both β-HCG and α-fetoprotein in the management of male germ cell tumors. The extremely high sensitivity and specificity of these tests permit the relatively early recognition of disease recurrence (prognostic factor). As a result, it is possible to consider observation of many patients following surgery without the immediate delivery of adjuvant chemotherapy, with a rise in the blood tumor marker indicating the need for known effective (curative) cytotoxic drug delivery in this clinical setting (predictive factor).

The final point to be highlighted in this commentary is the observation that although based on currently available data we should appropriately conclude a particular biomarker is a reasonable prognostic factor and knowledge of its presence does not predict for the use of a therapeutic strategy that can favorably influence that outcome, it is possible future therapeutic developments and research efforts may clinically meaningfully alter this scenario.

Consider, for example, the use of the serum CA-125 level in monitoring the course of patients with advanced ovarian cancer who have completed a primary chemotherapy (with or without maintenance) regimen. The previously reported randomized MRC OV05/EORTC 55955 collaborative trial (ISRCTN87786644) demonstrated that initiating treatment in individuals based solely on a rising CA-125 level (in the absence of other signs/symptoms of progressive disease) failed to improve overall survival compared with waiting until there was other evidence of disease progression, indicating the failure of the value of this biomarker as a predictive factor for improving clinical outcomes.4

However, it is possible (and one might even argue, likely) that at some point in the future an antineoplastic strategy will be developed and subsequently documented in a well-designed randomized clinical trial to be most effective when delivered at a relatively earlier point in time in a patient’s ovarian cancer journey (eg, less tumor volume, decreased presence of resistant cell populations, improved performance status and immune competence, etc) such that a rising serum CA-125 level may be found to predict for the optimal use of that approach.

Finally, it is also reasonable to speculate that other now-discarded diagnostic strategies (eg, second-look surgical assessment in ovarian cancer) might f ind a new role following solid evidence, as highlighted above, that initiation of a particular therapy at the earliest documentation of a specific clinical state predicts for a more favorable survival outcome.

References

1. Sechidis K, Papangelou K, Metcalfe PD, Svensson D, Weatherall J, Brown G. Distinguishing prognostic and predictive biomarkers: an information theoretic approach. Bioinformatics. 2018;34(19):3365-3376. doi:10.1093/bioinformatics/bty357
2. Johannet P, Abdelfattah S, Wilde C, et al. Molecular and clinicopathologic impact of GNAS variants across solid tumors. J Clin Oncol. 2024;42(32):3847-3857. doi:10.1200/JCO.24.00186 Slamon DJ, Clark GM,
3. Wong SG, et al. Human breast cancer: correlation of relapse and survival with amplification of the HER-2/neu oncogene. Science. 1987;235(4785):177-182. doi:10.1126/ science.3798106
4. Rustin GJS, van der Burg MEL, Griffin CL, et al. Early versus delayed treatment of relapsed ovarian cancer (MRC OV05/EORTC 55955): a randomised trial. Lancet. 2010; 376(9747):1155-1163. doi:10.1016/S0140-6736(10)61268-8