[Skip to Navigation]
Sign In
Views 558
Citations 0
Comment & Response
September 5, 2019

Reanalysis of Data Comparing Prophylactic Cranial Irradiation vs Observation in Patients With Locally Advanced Non–Small Cell Lung Cancer

Author Affiliations
  • 1Department of Statistics and Actuarial Science, The University of Hong Kong, Hong Kong
  • 2Department of Biostatistics, The University of Texas, MD Anderson Cancer Center, Houston
JAMA Oncol. 2019;5(11):1638. doi:10.1001/jamaoncol.2019.3593

To the Editor In randomized clinical trials with time-to-event outcomes, hazard ratios (HRs) are often reported, and misinterpretation of the HR is common.1 In the long-term update of the NRG Oncology/RTOG 0214 phase 3 trial, Sun et al2 reported that prophylactic cranial irradiation (PCI) decreased the incidence of brain metastasis (BM) and prolonged disease-free survival but did not affect overall survival in patients with locally advanced non–small cell lung cancer. Based on an HR of 0.43 (95% CI, 0.24-0.77; P = .003) for BM, the authors posited that patients in the PCI arm were 57% less likely to develop BM than those in the observation arm. Such an interpretation is inappropriate because, unlike the relative risk, the HR is not a ratio of 2 probabilities and does not convey a message of likelihood of developing BM. Instead, we claim that the hazard of BM for patients in the PCI arm was reduced by 57% compared with patients in the observation arm. Using a stratified log-rank test, a significant difference in disease-free survival (HR, 0.76; 95% CI, 0.59-0.97; P = .03) was shown. However, the Kaplan-Meier survival curves in Figures 2A and 2B of the article by Sun et al2 are intertwined during the first year, and the spread between the 2 survival curves is quite irregular throughout the entire follow-up (sometimes the curves even cross), which suggests a possible violation of the proportional hazards (PH) assumption. The log-rank test is optimal under the PH assumption; it incurs power loss otherwise. Moreover, if the PH assumption does not hold, the HR from the Cox model is not a meaningful quantity and its interpretation is unclear. A more robust (assumption-free) approach to quantifying treatment effect uses the restricted mean survival time (RMST),3-5 which is defined as the area under the Kaplan-Meier curve. We reanalyzed the data constructed from Figure 2B,2 and the difference in 10.5-year RMSTs between the 2 arms was 0.46 (95% CI, −0.245 to 1.154; P = .203), favoring PCI. During the 10.5-year follow-up, patients receiving PCI enjoyed less than 6 months disease-free survival on average compared with those not receiving PCI, which was not statistically significant. The log-rank test without stratification yielded P = .16, which is consistent with the RMST result. We recommend that the authors reanalyze the data using the stratified RMST-based test (stratified by the AJCC stage, prior surgery, histologic characteristics, and Zubrod performance status) owing to a lack of power for the log-rank test under deviations from the PH assumption.

Add or change institution
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words