CT Based Dosimetry Following Low Dose Rate (LDR) Prostate Brachytherapy in 2787 Consecutive Cases over 11 Years

Abstract

To examine trends in CT based dosimetry within a large, multi-practitioner LDR prostate program. Our program was established on 1/11/97 and developed a planning/treatment algorithm that attempted to incorporate the most workable components of existing programs; the first implant was done 20/07/98. Initially all implants were done by four Radiation Oncologists usually working in pairs. After ˜3 years, a formal 6-12m training program was introduced; today 14 Radiation Oncologists participate and 2787 consecutive implants are included in this analysis. Model 6711 (Oncura) 125Iodine sources of 0.33 mCi (NIST99) were used for >99% of implants, 333 of the first 361 used loose seeds, the remaining 2454 employed RapidStrand. Day-30 CT was standard for the first 1604, after which day-0 CTs were substituted. CT dosimetry is available for 98% (N = 2734). In general, the Radiation Oncologist who did the implant also contoured the CT images. The first 1500 patients in the cohort have a median FU of 60m (range 0-135.4m) and this subset was used to calculate the KM biochemical no evidence of disease (bNED) rates using the Phoenix definition. Among this subset, 832 had low risk; 668 intermediate risk and 4 high risk disease. Two thirds (N = 958) received 6m of adjuvant androgen deprivation therapy (ADT). The 5, 7 and 10y KM bNED rates (±95% CI) are 96.0 ±1.2%, 93.9 ± 1.8% and 93.0 ± 2.0%. D90 and V100 were not predictive of bNED in univariate or multivariate analysis. The means (±SD) for D90 (expressed as % of prescription dose), V100, and V150 are 104.8 ± 9.7%, 91.9 ± 5.3% and 52.7 ± 12.0%. A learning curve effect is seen; the mean D90% and V100 for our first 100 implants are lower than for the rest (96.4 vs. 106.5%; 85.5 vs. 92.1%, p < 0.001) and the proportion of implants failing to reach quality assurance (QA) thresholds (D90% ≥90%, V100 ≥85%) declined with time. Nearly 1/3 (31) of first 100 implants had D90 <90%, which applied to 15.3% of the next 500 implants and only 2.8% of the remaining 2187. The V100 results are virtually identical. In contrast, the mean (±SD) V150 was 45.3 ± 14.6% in the first 100 implants, gradually rose to a wide peak of >60% for implants 400-800 inclusive, and then declined to a plateau of just under 50% (48.8 ± 10.5%) for the most recent 1000 cases. An improving therapeutic ratio is suggested by a decreasing V150 in the presence of better dose coverage and may reflect increasing reliance on stranded sources to provide both central sparing with enhanced extraprostatic coverage. About 500 cases were completed before more than 95% reliably met QA thresholds; ironically, but fortunately, these arbitrary thresholds proved to be poor predictors of biochemical failure in this series.