Estimating treatment effects from a randomized clinical trial in the presence of a secondary treatment

Our plain-English summary of Estimating treatment effects from a randomized clinical trial in the presence of a secondary treatment, by M. Zhang, Y. Wang (Biostatistics, 2012).

The study

• Authors: M. Zhang, Y. Wang
• Journal: Biostatistics
• Published: 2012;13(4):625-36
• Cited by: 10 articles, as counted by NIH iCite
• Read it: PubMed · Publisher

Abstract

In randomized clinical trials involving survival time, a challenge that arises frequently, for example, in cancer studies (Manegold, Symanowski, Gatzemeier, Reck, von Pawel, Kortsik, Nackaerts, Lianes and Vogelzang, 2005. Second-line (post-study) chemotherapy received by patients treated in the phase III trial of pemetrexed plus cisplatin versus cisplatin alone in malignant pleural mesothelioma. Annals of Oncology 16, 923--927), is that subjects may initiate secondary treatments during the follow-up. The marginal structural Cox model and the method of inverse probability of treatment weighting (IPTW) have been proposed, originally for observational studies, to make causal inference on time-dependent treatments. In this paper, we adopt the marginal structural Cox model and propose an inferential method that improves the efficiency of the usual IPTW method by tailoring it to the setting of randomized clinical trials. The improvement in efficiency does not depend on any additional assumptions other than those required by the IPTW method, which is achieved by exploiting the knowledge that the study treatment is independent of baseline covariates due to randomization. The finite-sample performance of the proposed method is demonstrated via simulations and by application to data from a cancer clinical trial.

In plain English

This was a statistics methods paper, not a study of asbestos exposure. The authors proposed a way to estimate a treatment's effect in a randomized trial when patients later start other ('secondary') treatments. Their method builds on an existing approach called inverse probability of treatment weighting (IPTW). They tested it in computer simulations (1000 data sets of 400 subjects each). They then applied it to a phase III trial in which 448 patients with malignant pleural mesothelioma were randomized to cisplatin (222) or pemetrexed plus cisplatin (226). Median survival was 9.3 months in the cisplatin arm and 12.1 months in the combination arm, and the intent-to-treat hazard ratio was 0.77. After accounting for second-line chemotherapy, the hazard ratio for the combination versus cisplatin was 0.69 with the usual IPTW method and 0.65 with the new method. The authors wrote that this supported concern that the intent-to-treat analysis had underestimated the effect of the study treatment. They concluded that their method may give moderate to considerable gains in efficiency (statistical precision) when stabilized weights are used.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Statistical methods paper: development of a new estimation method, tested by computer simulation and applied to data from a previously conducted randomized phase III clinical trial
• Who was studied: Simulated data sets, plus 448 patients with malignant pleural mesothelioma in a randomized phase III trial: 222 assigned to cisplatin (control) and 226 to pemetrexed/cisplatin combination
• Exposure: Not asbestos. The exposures studied were the randomized chemotherapy (cisplatin versus pemetrexed plus cisplatin) and second-line (secondary) chemotherapy started during follow-up. Asbestos is mentioned only in passing as linked to mesothelioma.
• Main results:
  • Simulations were each based on 1000 data sets of 400 subjects
  • In simulations, for the randomized treatment effect, the new (augmented) estimators were more efficient than the corresponding usual ones; the gain was only very mild when simple (unstabilized) weights were used
  • In the mesothelioma trial, median survival was 9.3 months in the cisplatin arm and 12.1 months in the combination arm
  • Intent-to-treat hazard ratio for death, combination versus cisplatin: 0.77 (log HR -0.27, standard error 0.115)
  • Time to starting a secondary treatment differed significantly between the two arms (log-rank p = 0.0038)
  • Hazard ratio of death for receiving any secondary treatment versus none: 0.59 (log HR -0.52, SE 0.221) with usual IPTW and 0.58 (log HR -0.54, SE 0.227) with the new method
  • After accounting for secondary treatment, hazard ratio for combination versus cisplatin: 0.69 (log HR -0.36, SE 0.153) with usual IPTW and 0.65 (log HR -0.43, SE 0.140) with the new method
  • The authors stated this supported investigators' concern that the intent-to-treat analysis underestimated the true effect of the study treatment
  • The new method improved efficiency of estimating the randomized treatment effect by 19% relative to the usual IPTW method
• The authors' conclusion: The authors concluded that, compared with the usual IPTW method, their method may lead to moderate to considerable gain in efficiency in estimating the effect of the study treatment when stabilized weights are used, without requiring additional assumptions
• Limitations the authors state:
  • Neither the usual IPTW method nor the proposed method is robust if the effect of secondary treatments is wrongly specified (the model assumes a proportional effect on hazard)
  • The approach requires correctly modeling the time to secondary treatment based on confounders
  • The model assumes no interaction between the randomized and secondary treatments
  • The efficiency gain is mild when weights are very large (simple weights)

Topics

Medical terms in this study, explained in our Lexicon: Clinical Trial, Chemotherapy, Pemetrexed (Alimta), Cisplatin, Pleural Mesothelioma.

Related studies

• Antimetabolites in the treatment of advanced pleural mesothelioma: a network meta-analysis of randomized clinical trials — D. Tassinari, C. Cherubini, et al. (2017)
• Systemic treatment of malignant pleural mesothelioma: new agents in clinical trials raise hope of relevant improvements — D.C. Christoph, W.E. Eberhardt (2014)
• Randomized phase II trial of pemetrexed/cisplatin with or without CBP501 in patients with advanced malignant pleural mesothelioma — L.M. Krug, A.J. Wozniak, et al. (2014)
• Need for randomized clinical trials testing targeted therapies in malignant pleural mesothelioma — N. Sobhani, G. Roviello, et al. (2017)
• A Randomized Phase II Study Adding Axitinib to Pemetrexed-Cisplatin in Patients with Malignant Pleural Mesothelioma: A Single-Center Trial Combining Clinical and Translational Outcomes — W.A. Buikhuisen, M. Scharpfenecker, et al. (2016)
• Phase II study of first-line bortezomib and cisplatin in malignant pleural mesothelioma and prospective validation of progression free survival rate as a primary end-point for mesothelioma clinical trials (European Organisation for Research and Treatment of Cancer 08052) — M.E. O'Brien, R.M. Gaafar, et al. (2013)

Browse return to medical research.