Target trial emulation to estimate the effects of lifestyle interventions on mortality among cancer survivors: an editorial commentary

In 1986, Robins proposed a new approach for exploring the long-term impact of exposures for which the relative risk is less than 2.0 in relation to mortality (1). In a careful paper which focused on arsenic exposure, Robins detailed the many steps and assumptions required for proving causal inference that arsenic exposure resulted in both all-cause and lung cancer mortality. Over the years, many have proposed multi-step methods to specify and emulate target trials that expand the use of Robins’ original methods (2).

In September’s issue of Epidemiology and in an article entitled, “Estimating the effects of lifestyle interventions on mortality among cancer survivors: a methodologic framework”, McGee and colleagues propose a three-step framework to estimate the impact of adhering to diet and physical activity guidelines (including alcohol intake) that are recommended for cancer survivors by the World Cancer Research Fund-American Institute of Cancer Research (WCRF-AICR) and the American Cancer Society (ACS) on mortality (3-5). They conducted their analyses among breast and prostate cancer survivors identified within the Nurses’ Health Studies (NHS) I & II and the Health Professionals Follow-Up Study (HPFS). This is a unique contribution because it is one of only a few studies that have attempted to explore causal inference relative to lifestyle factors and mortality among cancer survivors (5-7), and to our knowledge, it is the second to have tackled both diet and exercise behaviors together (5,7). Thus, this paper represents an ambitious and pioneering effort. Our intention here is to open the dialog between the authors and the scientific community at large, and to pose questions that arose as we reviewed this article, especially since the authors suggest their methods as a framework and we have concerns that some of the assumptions and applications may result in findings now (and in the future) that are interpreted as causal, but may lack foundational rigor.

As noted, Robins’ approach to causal inference began with the investigation of arsenic exposure and was conducted among copper smelters in which the industrial exposure was circumscribed, and for which the number of covariates was few (e.g., cigarette smoking history). In contrast, the current paper is far more complex as it assesses multiple exposures which are influenced by a multitude of factors. We have elected to structure this commentary around the six core questions proposed by Dahabreh and Bibbins-Domingo as guiding the use of causal interpretation (8). These questions are: (I) what is the causal question; (II) what quantity would, if known, answer the causal question; (III) what is the study design; (IV) what causal assumptions are being made; (V) how can the observed data be used to answer the causal question in principle and in practice; and (VI) is a causal interpretation of the analyses tenable?

What is the causal question?

While the introduction of this paper leads one to believe that the analysis will quantify the survival benefits of following the WCRF-AICR and ACS diet and physical activity guidelines for cancer survivors, McGee and colleagues defer to the protocol of the hypothetical trial to operationalize how the various domains of the guidelines are defined. Certainly, we appreciate the complex hurdles that such a procedure involves and the need to select domains for which the data support the analysis; however, at the same time, we lament that some key components of the guidelines are not included. For example, the leading diet and physical activity recommendation of both the WCRF-AICR and the ACS is to “Achieve and maintain a healthy weight” (3,4), yet this key recommendation is not directly explored. This is a notable shortcoming since over 70% of cancer survivors have body mass indices that indicate overweight or obesity (9), and both breast cancer and aggressive prostate cancer are considered obesity-related malignancies (10-12). Relatedly, the decisions regarding guideline adherence also do not acknowledge the separate guidelines for aerobic and muscle-strengthening exercise (which are combined in this analysis), or recommendations regarding sedentary behavior which are reflected in more current guidelines (4). Further, it is debatable whether the spirit of the guidelines that define the restriction of red meat to less than 18 oz/w (~500 g/w) for men, whereas for women the limitation is lower (<12 oz/w or ~350 g/w), is preserved within this analysis, though we recognize that in operationalizing these constructs, some compromises are necessary (13). These issues are raised not to criticize this work per se, but as “food for thought” in guiding future work to perhaps be more readily translatable through closer alignment with guidelines.

What quantity would, if known, answer the causal question?

While the previous section outlined some shortcomings on the scope and operationalization of the WCRF-AICR and ACS guidelines, we commend the authors on their fairly exhaustive investigation of various factors that resulted in modifications to eligibility criteria, strategies, outcomes and consideration of several covariates for adjustment (though some factors, such current smoking, sedentary behavior, sleep, and mental health that are time-variant and influenced by other lifestyle practices appear to have been overlooked). Our main concern, however, relates to the sixth statement within the ‘Results’, i.e., “No individual adhered to all seven components of the intervention continuously throughout the follow-up”, thus calling into question if the available data can truly support the analysis aimed at answering the causal question. Indeed, it is unlike the straightforward analysis of Robins’ in which there was the assurance of a continuous exposure.

What is the study design?

The current analysis relies on the bedrock of data collected for the NHS (I&II) and the HPFS—epidemiological studies for which there is impressive fidelity and follow-up. Nonetheless, these studies may not ideally serve to support efforts aimed at causal inference, at least with regard to determining survival in relation to WCRF-AICR and ACS guideline adherence. The scheduled assessments at 2-year intervals not only lack the ability to capture continual exposure but also may miss critical periods in relation to cancer treatments (e.g., before, during, between, and after during which time survivors either engage or not with healthful behaviors). Secondly, there may be issues in harmonizing the data to reflect the guidelines (even in the most simplistic, operationalized terms). As an example, the amount of red meat that defines a portion in the NHS (I&II) and HPFS surveys is broad and imprecise (4–6 oz) and calls into question if we can really discern guideline adherence among these survivors. Third, we had questions regarding the sample that was either included initially, or that remained at the time of follow-up; for example, in the text and as supported by the CONSORT diagrams, it appears that only prostate cancer survivors were withdrawn because of the development of subsequent metastasis, but not breast cancer survivors—a decision that lacks justification or perhaps, clarification. The NHS and HPFS samples also may not reflect the general population of cancer survivors that exist today; obviously, their educational and socio-economic status is higher (14), but there also have been substantial changes in treatment as reflected the much higher rates of surgery and radiotherapy among the prostate cancer survivors with early stage disease in this sample (a group for which active surveillance is now more the norm) (15). Furthermore, some of the eligibility criteria lack justification or are not typically used in randomized controlled trials (RCTs), such as those that appear to be “grandfathered-in” from the previous work of Dickerman et al. (6) such as stroke or myocardial infarction (MI). Indeed, while a stroke or MI within a 6-month period is considered cause for caution for unsupervised physical activity and is included on standard screening instruments (16), in the current analysis, ever experiencing these events is grounds for censoring under an assumption that these individuals may be unable to continue adherence to guidelines. Such an assumption runs counter to almost all clinical trial protocols that defer to managing physicians and promote gradual progression towards guideline adherence. It is also antithetical to the standard of care, in which patients who experience such events are expected to enroll in rehabilitation programs (where the importance of physical activity and a healthful plant-based diet is heightened, rather than downplayed). These assumptions, plus those that exclude or excuse survivors who have or later develop transient conditions such as gout, highly treatable conditions, such as type II diabetes, and who report difficulty in performing some physical function activities, such as climbing a flight of stairs are oftentimes the target population for behavioral intervention trials and for whom these interventions may be especially effective (17). Thus, excluding them may be a mistake, and runs counter to the guidelines put forth by other professional organizations, such as the American Diabetes Association (18). Moreover, given the high prevalence of comorbidity and functional limitations among cancer survivors (19), we worry that the estimates reported in this paper may be ungeneralizable to the greater population of survivors for which this work is intended.

What causal assumptions are being made?

As noted by Dahabreh and Bibbins-Domingo, causal assumptions cannot be fully evaluated with the data alone (8); however, within their study design, the authors provide a detailed description of safeguards against unmeasured confounding and positivity.

How can the observed data be used to answer the causal question in principle and in practice?

The results provided by McGee and colleagues are generated through sound statistical methods (5). Furthermore, these estimates appear stable even under various imputation approaches that account for non-response, missing data, and implausible values.

Is a causal interpretation of the analyses tenable?

In reviewing the work of McGee and colleagues (5), we reiterate our praise for the technical merits of their undertaking; however, concerns related primarily to the study design prevent us from accepting that the reductions in risk due to adhering to the guidelines (as narrowly defined) are necessarily causal. Dabraheh and Bibbins-Domingo (8) describe this step as difficult and largely subjective—a step that can be informed by triangulation of data across multiple sources and with further testing using negative controls and methods to discern the presence of assumption violations. It may be premature to make such a call at present since there are so few studies, and differences in the study design, follow-up duration, assumptions, and factors controlled for between this study and that conducted by Ergas and colleagues (7), complicate comparisons.

RCTs are still preferable

In making the case for target trial emulation studies, the authors note that “long-term randomized trials investigating the effects of lifestyle interventions on mortality among cancer survivors are often logistically complex or impractical”. While this statement is true, it is important not to abandon RCTs of lifestyle interventions and survival in oncology because they provide the strongest evidence of causality and are feasible in many clinical scenarios. One example is the recent Colon Health and Life-Long Exercise Change (CHALLENGE) trial (20). The CHALLENGE trial was the first phase 3 trial examining the effects of exercise on cancer-related survival. The trial randomized 889 patients with resected colon cancer after adjuvant chemotherapy to health education materials (n=444) or a 3-year structured exercise program (n=445). With a median follow-up of 7.9 years, the structured exercise program compared to health education materials significantly improved disease-free survival [hazards ratio (HR) =0.72; 95% confidence interval (CI): 0.55 to 0.94; P=0.017] and overall survival (HR =0.63; 95% CI: 0.43 to 0.94; P=0.022). Other RCTs with survival endpoints, such as, those aimed at determining the impact of increasing fruit and vegetable consumption and adhering to a Mediterranean, macrobiotic diet have been completed successfully (21,22), as have others that have tested lifestyle interventions and weight loss, e.g., the Lifestyle Intervention for Ovarian Cancer Enhanced Survival (LiVES) and the Breast Cancer Weight Loss (BWEL) trials, respectively for which results are imminent (23,24). Therefore, RCTs are indeed possible and should be pursued to provide the strongest level of evidence of cause and effect.

Target trial emulation studies need to answer clinically relevant questions

Even when RCTs are not possible due to logistical or ethical reasons, observational studies need to be designed to accommodate target trial emulations of clinically relevant questions (25). Most existing observational studies have not been designed with target trial emulation in mind. Assessments of physical activity, dietary intake, and other lifestyle practices often are performed at arbitrary time points well before or after the cancer diagnosis rather than at clinically important time points in relation to cancer treatments (a limitation of the current study listed under ‘study design’) (25). The age of precision medicine is upon us, and it is all about delivering the right intervention to the right patient at the right time. The right time in clinical oncology is in relation to the existing treatment combination. That is, what is the effect of a given lifestyle intervention when it is performed immediately before, during, between, or immediately after a given treatment combination? Without lifestyle assessments at these clinically relevant time points, it is not possible to apply target trial emulation to answer clinically relevant questions about lifestyle and survival. Moreover, even when broader research questions, such as the impact of adhering to lifestyle practices on quality-of-life, are the central focus, it is still imperative to assure fidelity with the recommendations and to select an appropriate sample and sampling plan that can provide sound results. Thus, observational studies must be designed prospectively to assure adequate rigor and may in fact require an outlay of time and resources that are similar to RCTs. Strategies for designing clinically relevant observational studies of physical activity and survival have been proposed (25).

In summary, we hope that our comments may be helpful as others seek to design or refine studies that bring us closer to determining causal inference from observational data as related to the impact of diet and physical activity (as well as other lifestyle exposures) on survival outcomes among cancer survivors. Given the rising numbers of cancer survivors, these data are sorely needed; however, we must be mindful to ensure that resulting analyses reflect the guidelines as written and produce accurate results that can be translated and generalized broadly.

Acknowledgments

None.

Footnote

Provenance and Peer Review: This article was commissioned by the Editorial Office, Annals of Cancer Epidemiology. The article has undergone external peer review.

Peer Review File: Available at https://ace.amegroups.com/article/view/10.21037/ace-2025-12/prf

Funding: This work was supported in part by the Division of Intramural Research at the National Institute on Minority Health and Health Disparities (NIMHD) of the National Institutes of Health (NIH) (ZIA MD000022, PI: Jayasekera) (to O.W.A.W.). The contributions of the NIH author(s) were made as part of their official duties as NIH federal employees, are in compliance with agency policy requirements, and are considered Works of the United States Government. However, the findings and conclusions presented in this paper are those of the author(s) and do not necessarily reflect the views of the NIH or the U.S. Department of Health and Human Services.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://ace.amegroups.com/article/view/10.21037/ace-2025-12/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Robins J. A new approach to causal inference in mortality studies with a sustained exposure period—application to control of the healthy worker survivor effect. Math Model 1986;7:1393-512.
2. Hernán MA, Dahabreh IJ, Dickerman BA, et al. The Target Trial Framework for Causal Inference From Observational Data: Why and When Is It Helpful? Ann Intern Med 2025;178:402-7. [Crossref] [PubMed]
3. World Cancer Research Fund/American Institute of Cancer Research. Diet, Nutrition, Physical Activity and Cancer: A Global Perspective. 2018. Available online: https://www.wcrf.org/wp-content/uploads/2024/11/Summary-of-Third-Expert-Report-2018.pdf
4. Rock CL, Thomson CA, Sullivan KR, et al. American Cancer Society nutrition and physical activity guideline for cancer survivors. CA Cancer J Clin 2022;72:230-62. [Crossref] [PubMed]
5. McGee EE, Hernán MA, Giovannucci E, et al. Estimating the effects of lifestyle interventions on mortality among cancer survivors: a methodologic framework. Epidemiology 2025;36:705-18. [Crossref] [PubMed]
6. Dickerman BA, Giovannucci E, Pernar CH, et al. Guideline-Based Physical Activity and Survival Among US Men With Nonmetastatic Prostate Cancer. Am J Epidemiol 2019;188:579-86. [Crossref] [PubMed]
7. Ergas IJ, Bradshaw PT, Cespedes Feliciano EM, et al. Hypothetical Interventions on Diet Quality and Lifestyle Factors to Improve Breast Cancer Survival: The Pathways Study. Cancer Epidemiol Biomarkers Prev 2023;32:1716-25. [Crossref] [PubMed]
8. Dahabreh IJ, Bibbins-Domingo K. Causal Inference About the Effects of Interventions From Observational Studies in Medical Journals. JAMA 2024;331:1845-53. [Crossref] [PubMed]
9. National Cancer Institute. Cancer Survivors and Weight. Cited 2024. Available online: https://progressreport.cancer.gov/after/weight
10. Calle EE, Rodriguez C, Walker-Thurmond K, et al. Overweight, obesity, and mortality from cancer in a prospectively studied cohort of U.S. adults. N Engl J Med 2003;348:1625-38. [Crossref] [PubMed]
11. Lauby-Secretan B, Scoccianti C, Loomis D, et al. Body Fatness and Cancer--Viewpoint of the IARC Working Group. N Engl J Med 2016;375:794-8. [Crossref] [PubMed]
12. Pati S, Irfan W, Jameel A, et al. Obesity and Cancer: A Current Overview of Epidemiology, Pathogenesis, Outcomes, and Management. Cancers (Basel) 2023;15:485. [Crossref] [PubMed]
13. Shams-White MM, Brockton NT, Mitrou P, et al. Operationalizing the 2018 World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) Cancer Prevention Recommendations: A Standardized Scoring System. Nutrients 2019;11:1572. [Crossref] [PubMed]
14. U.S. Bureau of Census. Quick Facts: United States Census. 2021. Available online: https://www.census.gov/quickfacts/fact/table/US/PST045221
15. Cooperberg MR, Meeks W, Fang R, et al. Time Trends and Variation in the Use of Active Surveillance for Management of Low-risk Prostate Cancer in the US. JAMA Netw Open 2023;6:e231439. [Crossref] [PubMed]
16. Shephard RJ. PAR-Q, Canadian Home Fitness Test and exercise screening alternatives. Sports Med 1988;5:185-95. [Crossref] [PubMed]
17. Morey MC, Snyder DC, Sloane R, et al. Effects of home-based diet and exercise on functional outcomes among older, overweight long-term cancer survivors: RENEW: a randomized controlled trial. JAMA 2009;301:1883-91. [Crossref] [PubMed]
18. Kalyani RR, Neumiller JJ, Maruthur NM, et al. Diagnosis and Treatment of Type 2 Diabetes in Adults: A Review. JAMA 2025;334:984-1002. [Crossref] [PubMed]
19. Wagle NS, Nogueira L, Devasia TP, et al. Cancer treatment and survivorship statistics, 2025. CA Cancer J Clin 2025;75:308-40. Erratum in: CA Cancer J Clin 2025 doi: 10.3322/caac.70036. [Crossref] [PubMed]
20. Courneya KS, Vardy JL, O'Callaghan CJ, et al. Structured Exercise after Adjuvant Chemotherapy for Colon Cancer. N Engl J Med 2025;393:13-25. [Crossref] [PubMed]
21. Pierce JP, Natarajan L, Caan BJ, et al. Influence of a diet very high in vegetables, fruit, and fiber and low in fat on prognosis following treatment for breast cancer: the Women's Healthy Eating and Living (WHEL) randomized trial. JAMA 2007;298:289-98. [Crossref] [PubMed]
22. Berrino F, Villarini A, Gargano G, et al. The Effect of Diet on Breast Cancer Recurrence: The DIANA-5 Randomized Trial. Clin Cancer Res 2024;30:965-74. [Crossref] [PubMed]
23. Thomson CA, Crane TE, Miller A, et al. Lifestyle intervention in ovarian cancer enhanced survival (LIVES) study (NRG/GOG0225): Recruitment, retention and baseline characteristics of a randomized trial of diet and physical activity in ovarian cancer survivors. Gynecol Oncol 2023;170:11-8. [Crossref] [PubMed]
24. Delahanty LM, Wadden TA, Goodwin PJ, et al. The Breast Cancer Weight Loss trial (Alliance A011401): A description and evidence for the lifestyle intervention. Obesity (Silver Spring) 2022;30:28-38. [Crossref] [PubMed]
25. Courneya KS, Friedenreich CM. Designing, analyzing, and interpreting observational studies of physical activity and cancer outcomes from a clinical oncology perspective. Front Oncol 2023;13:1098278. [Crossref] [PubMed]