3215 - A Secondary Analysis of Radiotherapy Oncology Group (RTOG/NRG) Trials 9202, 9413, and 9910: Unknown Causes of Death and Their Impact on Prostate Cancer Specific Mortality Estimates

J. Y. Lee¹, X. Wu², M. Wang³, A. U. Kishan⁴, A. Y. Jia⁵, D. E. Spratt⁶, H. M. Sandler⁷, J. A. Efstathiou⁸, and N. G. Zaorsky⁹; ¹University Hospitals Cleveland Medical Center, Cleveland, OH, ²Penn State College of Medicine, Hershey, PA, ³Department of Population and Quantitative Health Sciences, Case Western Reserve University, Cleveland, OH, ⁴Department of Radiation Oncology, University of California, Los Angeles, Los Angeles, CA, ⁵Weill Cornell Medical College/New York Presbyterian Hospital, New York, NY, ⁶Case Western, Cleveland, OH, ⁷Cedars-Sinai Medical Center, Los Angeles, CA, ⁸Department of Radiation Oncology, Harvard School of Medicine, Boston, MA, ⁹University Hospitals Seidman Cancer Center, Case Western Reserve University, Cleveland, OH

Purpose/Objective(s): To demonstrate that deaths of unknown cause represent enough data to introduce a significant amount of uncertainty to cancer-specific mortality endpoints in randomized controlled trials.

Materials/

Methods: RTOG 9202, 9413, and 9910 were chosen for secondary analyses. Data from the 4,369 participants of these trials were reviewed. For each trial, unknown cause deaths within each treatment arm were collectively reclassified either as deaths secondary to prostate cancer or non-prostate cancer deaths. Kaplan-Meier survival analyses and hazard ratios (HR) were computed using R statistical software (R Foundation for Statistical Computing, Vienna, Austria).

Results: RTOG 9202 originally reported prostate cancer-specific survival HR of 0.6857 favoring long-term ADT (LT-ADT) over short-term ADT (ST-ADT) (Wald test p-value= 0.00156, 95% CI: 0.5427-0.8663). Secondary analysis of RTOG 9202 with unknown cause deaths recoded as prostate cancer deaths within the ST-ADT group and as other-cause deaths within the LT-ADT group yielded HR of 0.4058 (Wald test p-value <0.0001, 95% CI 0.3278 – 0.5025). Recoding unknown cause deaths in the reverse yielded HR of 1.3004 (Wald test p-value=0.00947, 95% CI 1.066 – 1.586).

RTOG 9413 originally reported no statistically significant differences in prostate cancer-specific survival in its 2x2 comparisons of prostate-only radiation with whole-pelvic radiation plus prostate boost (WPRT) and neoadjuvant hormonal therapy (NHT) with adjuvant hormonal therapy. Secondary analysis with unknown cause deaths recoded as prostate cancer death in prostate-only radiation with NHT and as other-cause mortality in WPRT with NHT yielded HR of 1.637, favoring the prostate-only radiation with NHT group (Wald test p-value = .001).

RTOG 9910 originally reported no statistically significant difference in prostate cancer specific survival between long-term neoadjuvant ADT (LT-ADT) and short-term neoadjuvant ADT (ST-ADT). Secondary analysis of prostate cancer-specific survival with deaths of unknown cause recoded as prostate cancer death within the ST-ADT group and as other-cause mortality within the LT-ADT group yielded HR of 0.3741 (Wald test p-value < 0.0001, 95% CI 0.2401 – 0.583). Recoding unknown cause deaths in the reverse yielded HR of 1.7958 (Wald test p-value = 0.00787, 95% CI: 1.166 – 2.766).

Conclusion: Recoding unknown cause deaths as prostate cancer deaths or other-cause deaths resulted in conclusions that differed from the original cancer specific mortality findings of the three published randomized controlled trials. Cancer-specific mortality as a trial endpoint should be interpreted with caution, as it can be swayed by unknown causes of death.