Adjuvant Hormone Therapy After Prostate Radiation: Is This Data Torture?

Abstract

The noble laureate Ronald Coase and American journalist Darrell Huff are both credited with the quote "if you torture the data long enough, it will confess to anything." 1(p3),2 The meta-analysis by Spratt et al 3 on adjuvant hormone therapy for prostate cancer inadvertently risks falling into this category.

Reply to S. Sundar et al and S. Höcht et al

“…In fact, we used the entire Ottawa 0101 trial and the entire prostate-only randomized cohort in RTOG 9413. Moreover, we only tested a single hypothesis from our meta-analysis, and therefore, we disagree with Sundar et al 1 that we tortured the data. Additionally, the results were highly consistent across all end points assessed, were in agreement with the trends observed in each of the included trials, and were consistent with other trial results as outlined in our Discussion section.…”

“…There may be some confusion by Sundar et al 1 and Höcht et al 2 regarding our meta-analysis. 3 Although we fully agree with the dangers of multiple testing and unplanned subgroup analyses, much of what they have stated are in fact strengths of randomization and limitations of observational research, especially in the context of our paper.…”

“…Both Sundar et al 1 and Höcht et al 2 feel strongly about continuing neoadjuvant androgen deprivation therapy (ADT). The Ottawa trial suggested that adjuvant ADT improved biochemical recurrence-free survival (0.66; 95% CI, 0.41 to 1.07), RTOG 9413 demonstrated a significant improvement in biochemical control with adjuvant ADT in patients receiving prostate only radiotherapy (HR 0.73; 95% CI, 0.59 to 0.91), and in our study, when the trials are combined our meta-analysis shows that adjuvant ADT improves biochemical control, progression-free survival, distant metastasis, and metastasis-free survival.…”

“…The primary results shown are completely unadjusted and solely leverage the randomization. Thus, the comment by Sundar et al 1 on the “fog of statistical adjustments” has no basis. Sensitivity analyses were performed adjusting for clinical trial to account for potential confounders or biases, and the direction of association was congruent with the primary analyses.…”