Study design and setting: EPICAP-QALY is an ancillary study of the EPICAP survey (8). EPICAP is a multicentric case-control, observational prospective, open, follow-up study including newly diagnosed PCa patients between 2012 and 2014 (819 patients) and 879 age-matched healthy controls from the same area. The EPICAP-QALY was performed at Nimes University Hospital between August 2015 and October 2017 and approved by the institutional review board.
Participants: All participants from the EPICAP cohort completed a screening questionnaire to determine eligibility. Patients who had received hormone therapy within the previous year or who experienced a relapse in the intervening years were excluded, except patients on salvage radiotherapy following prostatectomy for more than 6 months with a PSA level < 1ng/ml. Age-matched ± 1 year healthy controls were included in a 1:2 ratio. Individuals diagnosed with PCa following inclusion or with a PSA > 10 ng/ml were excluded. Men with PSA > 10ng/ml at the time of completing the questionnaire were not selected to exclude potential relapse for cases or cancer occurrence for controls.
Outcomes: The primary outcome was QoL 3 years after PCa treatment compared to controls, as evaluated by the QLQ-C30 questionnaire (9). Secondary outcomes were the comparison of urinary, sexual and anxiodepressive dysfunction between patients and controls using the following questionnaires: IPSS International Prostate Symptom Score (10), ICIQ-MLUTS International Consultation on Incontinence Male Lower Urinary Tract Symptoms (11), IIEF-6 International Index of Erectile Function (12), and HADS Hospital Anxiety and Depression Scale (13).
These questionnaires were used to compare QoL and symptoms according to active surveillance (AS), radical prostatectomy (RP), EBRT, brachytherapy or High-intensity Focused Ultrasound (HIFU), androgen deprivation therapy (ADT) or combined care (CC). A life situation questionnaire complemented with specific questions concerning sexuality was used to test for some potential confounders (14).
Data collection: Age, BMI, PSA level, educational level, housing, living alone, marital status, monthly income, chronic disease and regular medication were collected. Treatment at diagnosis, last treatment received, hormone therapy within previous 12 months, and employment status were also recorded. For controls, urologic consultation for urinary troubles, prostate treatment and PSA testing in the 3 previous year were recorded.
Sample size: By predicting a lower participation rate in cases than controls and a recurrence rate of cases of 10%, we originally planned a cohort of 600 patients and 300 controls paired with a ratio 2:1 on age to highlight a standardized difference in score on the QLQ-C30 of 0.25 (“small” effect according to Cocks et al (9)) with a global bilateral risk alpha of 5% and 90% power. The participation rate was lower than expected and the study included 376 patients to whom we matched 188 patients (from the 364 available).
Statistics: The comparability of age was assessed with a Student test. Descriptive statistics are reported as counts and percentages for categorical variables and means and standard deviations for continuous variables with normal distribution and median and quartiles for others. Comparisons of baseline characteristics and putative risk factors between cases and controls were performed with Mann–Whitney, Kruskal–Wallis, χ2, Student, or Fisher exact test as appropriate.
For each questionnaire, the distribution of scores was analyzed. When extreme values (0 or 100) were over-represented, scores were recoded into classes and described qualitatively with effectives and percentages.
The univariate analysis was performed with a mixed linear model for quantitative scores (QLQ-C30 summary score, VS and IS score of ICIQ-MLUTS). For recoded QLQ-C30 scores, analyses were conducted with a mixed logit model. To account for pairing, a random effect on 2: 1 trinoma was considered.
For recoded QLQ-C30 scores, distribution and links with social potential confounders was assessed. When the symptom or trouble was present in less than 20% of cases or when no apparent link was possible, multivariate analysis was not performed. For other QLQ-C30 scores and for the summary quantitative score, the effects of putative confounders were evaluated. Socio-professional integration items were selected for testing based on their reliability, their clinical pertinence of potential confounding factors and their similarity with items of the QALIPRO study (15). Putative confounders for quantitative scores were analyzed with Spearman correlation test, Kruskal-Wallis or ANOVA as appropriate, and with χ2, Fisher test, Student or Wilcoxon test for qualitative values. All variables with a P-value lower than 0.20 were considered as potential covariates and adjusted mixed linear general models or logistic models were computed with a random effect on 2:1 trinoma.
All analyses were performed using SAS software (SAS Institute, Cary, NC) version 9.3. P-values <0.05 were interpreted as statistically significant for 2-sided tests. Since multiple comparisons increase the risk of introducing a Type-I error, we applied the sequentially rejective Bonferroni correction (Holm’s correction) to control for this type of error in Table 2 and 3. This means that the p-value must be divided by the number of tests run in parallel, resulting in an adjusted level of statistical significance. The corrected p-values for Holm’s correction are reported. For multivariate analysis, Holm’s correction is also applied on p-values of interest obtained by the models.