Program Information
Pooling Outcome Data Demonstrates Unforeseen Dose-Response Relationships
M Thor1*, A Jackson1 , M Zelefsky1 , G Steineck2 , A Karlsdottir3 , M Hoyer4 , M Liu5 , N Nasser1 , S Petersen4 , V Moiseenko6 , J Deasy1, (1) Memorial Sloan Kettering Cancer Center, New York, US, (2) Institute of Clinical Sciences, the Sahlgrenska Academy at the University of Gothenburg, Gothenburg, Sweden, (3) Haukeland University Hospital, Bergen, Norway, (4) Aarhus University Hospital, Aarhus C, Denmark, (5) Vancouver Cancer Centre, Vancouver, British Columbia, Canada, (6) UC San Diego, La Jolla, CA, US
Presentations
SU-I-GPD-T-98 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall
Purpose: To pool data from six cohorts (n=989) in order to explore if a generalizable dose-response relationship can be established for late rectal bleeding (LRB) after conventionally fractionated radiotherapy (RT) for localized prostate cancer.
Methods: Pooling feasibility was investigated by assessed best-fit cohort-specific rectal volume-effect parameters (a) from the Lyman-Kutcher-Burman (LKB) formalism, and cohorts with an a within the 95% bootstrap percentile confidence interval of other cohorts, using the heterogeneity index (I2), were pooled. In the pooled cohort, best-fit LKB parameters a/D50/m were identified and dose-response modeling was conducted in a training-and-testing-design within univariate and multivariate logistic regression on bootstrapping (n=1000). Predictability was estimated using area under the receiver-operating characteristics curve (AUC) and Hosmer-Lemeshow p-values (pHL). Since data was taken from six cohorts and included various prescription doses (70-86.4Gy@1.8-2.0Gy/fraction), all dose-volume histograms were fractionation corrected (α/β=3Gy).
Results: Four cohorts were feasible to pool since I2=0 when excluding the two cohorts with the lowest LRB rate (2% vs. 5-22%), and of which one was the only intensity-modulated RT (IMRT) cohort. Best-fit a/D50/m in the pooled cohort (n=690) were 3/41Gy/0.43; the absolute volume <55Gy and the minimum dose generated a model that best explained LRB (hold-out testing data: AUC: 0.71; pHL=0.63).
Conclusion: Pooling was feasible across four cohorts in which an unforeseen relationship to rectal sparing was emphasized. The only IMRT cohort was found unfeasible to pool, and precaution should, thus, be applied when combining 3DCRT and IMRT data.
Contact Email: