Immortal time bias in survival outcomes when comparing treatment with chemotherapy versus immunochemotherapy for non-small cell lung cancer

In an interesting observational study published in June 2023, Patel et al. compared the overall survival (OS) for patients treated with chemotherapy only versus chemoimmunotherapy for non-small cell lung cancer (NSCLC) (1). This is an important topic to describe using real-world data. We are nonetheless concerned that immortal time bias can have led to misleading results that have affected the conclusions drawn by the authors, and which are being used to support more recent studies (2).

Immoral time bias can occur when subjects in a cohort study are not at risk of experiencing the outcome of interest for some duration of time and are effectively “immortal” during this period (3). This bias was first described in medical research in 1972 (4), and has been widely discussed as a potential pitfall of survival analyses conducted using observational data (5-8). It has nonetheless been estimated that as many as 40% of studies with time-dependent factors are biased due to not accounting for time-dependent effects in the analysis (9).

In the study by Patel et al., survival time was measured from diagnosis and the exposure of interest was the type of systemic treatment received. Because patients were selected for this cohort study based on having survived long enough to receive their first systemic treatment, the observed risk of death between diagnosis and treatment initiation was 0% for all study subjects. Additionally, patients in the chemoimmunotherapy group had to survive long enough to receive two systemic therapies, which would have further increased their guaranteed survival period (“immortal time”). Indeed, the authors’ description in the “variables of interest” section implies that patients in this group could have started immunotherapy up to one year after diagnosis and were therefore immortal until starting this treatment.

Fig. 1 in Patel et al. illustrates this as no deaths occurred until roughly 1.5 months of follow-up in the chemotherapy group and 3.5 months in the chemoimmunotherapy group, despite the poor prognosis associated with stage IV NSCLC. To evaluate the impact that immortal time bias could have had on the overall results of the study, we have performed a naïve analysis to partially correct for this bias by starting follow-up after treatment initiation. In our analysis, we assumed an immortal time of 1.5 and 3.5 months for all patients in the chemotherapy only and chemoimmunotherapy groups respectively. We corrected the observed survival time by subtracting the assumed immortal time. This approach demonstrates the minimum change we would expect to see in survival times, since patients would have likely experienced a longer immortal period (up to one year) based on their individual treatment plans. Lastly, we recreated the survival curves presented in Fig. 1A by Patel et al. using a previously described algorithm and the program WebPlotDigitizer (available at https://automeris.io/) (10). In the recreated survival curves from Patel et al., the difference in median survival between the treatment groups is 5.6 months (Figure 1A). After applying our correction, the difference in median survival between the two treatment groups is reduced by 40%, to 3.4 months (Figure 1B).

Figure 1 Estimated survival for patients treated with chemotherapy and chemoimmunotherapy (A) before and (B) after a naïve correction for immortal time bias. The arrows indicate the difference in median survival observed between treatment groups.

Performing a similar naïve correction to the sex-based survival curves presented in Fig. S2 by Patel et al., we find that the difference in median survival between the two treatments groups is reduced from 6.7 to 3.7 months for males, and by from 4.7 to 3.1 months for females. The sex-based corrected median survival improvement between treatment groups is thus reduced by 70%, from 6.7−4.7=2 months to just 3.7−3.1=0.6 months. In our view, this substantially weakens the authors’ claim that “chemoimmunotherapy yielded significant increases in OS over chemotherapy for males”.

We acknowledge the efforts of Patel and colleagues to generate much needed real-world evidence about the effectiveness of immunotherapy in general clinical practice. However, it appears that immortal time bias may have resulted in a substantial overestimation of survival in their study, particularly for those treated with chemoimmunotherapy; our naïve corrections indicate that the differences in median survival between groups may have been overestimated by 40–70%. We are thus concerned that immortal time bias undermines the validity of the results published by Patel et al. to the extent that jeopardizes the validity of their conclusions, and worried that these results and conclusions are further referred to in the scientific literature.

Acknowledgments

Funding: None.

Footnote

Provenance and Peer Review: This article was a standard submission to Translational Lung Cancer Research. The article did not undergo external peer review.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-24-1002/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. Patel KH, Alpert N, Tuminello S, et al. Personal and clinical characteristics associated with immunotherapy effectiveness in stage IV non-small cell lung cancer. Transl Lung Cancer Res 2023;12:1210-20. [Crossref] [PubMed]
2. Taioli E, Flores RM, Abdelhamid A, et al. Antibiotic use and survival in patients with late-stage NSCLC treated with chemoimmunotherapy. JTO Clin Res Rep 2024;5:100710. [Crossref] [PubMed]
3. Catalogue of Bias Collaboration, Lee H, Nunan D. Immortal time bias. In: Catalogue of Bias 2020 [Accessed: October 11, 2024]. Available online: https://catalogofbias.org/biases/immortal-time-bias/
4. Gail MH. Does cardiac transplantation prolong life? A reassessment. Ann Intern Med 1972;76:815-7. [Crossref] [PubMed]
5. Newman NB, Osmundson EC. Practical demonstration of time bias with administration of adjuvant therapy in lung cancer. Lung Cancer 2021;157:75-8. [Crossref] [PubMed]
6. Agarwal P, Moshier E, Ru M, et al. Immortal time bias in observational studies of time-to-event outcomes: assessing effects of postmastectomy radiation therapy using the national cancer database. Cancer Control 2018;25:1073274818789355. [Crossref] [PubMed]
7. Weberpals J, Jansen L, van Herk-Sukel MPP, et al. Immortal time bias in pharmacoepidemiological studies on cancer patient survival: empirical illustration for beta-blocker use in four cancers with different prognosis. Eur J Epidemiol 2017;32:1019-31. [Crossref] [PubMed]
8. Suissa S. Immortal time bias in pharmaco-epidemiology. Am J Epidemiol 2008;167:492-9. [Crossref] [PubMed]
9. van Walraven C, Davis D, Forster AJ, et al. Time-dependent bias was common in survival analyses published in leading clinical journals. J Clin Epidemiol 2004;57:672-82. [Crossref] [PubMed]
10. Guyot P, Ades AE, Ouwens MJ, et al. Enhanced secondary analysis of survival data: reconstructing the data from published Kaplan-Meier survival curves. BMC Med Res Methodol 2012;12:9. [Crossref] [PubMed]