Often in observational studies the treatment effects within subgroups are important. The question arises whether the overall propensity score (PS) should be used to adjust for confounding and to estimate subgroup treatment effects or whether PS should be recalculated within the subgroups to estimate subgroup treatment effects. This paper addresses this issue from the perspective of the PS differences and differences for the within subgroup adjusted treatment effects. A specific real world evidence oncology study is used to illustrate the findings. We show that the propensity scores obtained from the within group model are identical to the propensity scores obtained from the overall model with the addition of the interaction effects of the subgroup variable with the other confounders. This information being added to the overall propensity score model is small. Thus, to analyze the treatment effect within subgroups, either the overall propensity score or the within subgroup propensity score will yield adjusted treatment effects which are not substantively different. In addition, for both the within subgroup PS and the overall PS, the treatment effects from the within subgroups analysis are identical to the treatment effects form the overall model and including propensity score by age interaction. We conclude that it is not necessary to compute within subgroup propensity scores nor to use within subgroup analyses to estimate the adjusted within subgroup treatment effects. This is consistent with the primary analysis which considers overall confounders and not how confounders may differ within subgroups.

Objectives: The Medicines and Healthcare Products Regulatory Agency in the United Kingdom (UK) formally reclassified sildenafil citrate 50 mg tablets as a pharmacy medicine (sildenafil-P) in 2017 for adult men with erectile dysfunction (ED). A one-year prospective real-world observational study was conducted to track men’s health behavior, particularly their healthcare resource utilization (HCRU) and quality-of-life (QoL) before and after the availability of sildenafil-P. Methods: Adult men with ED aged ≥18 years provided data at baseline (prior to launch of sildenafil-P) and every three months after the launch. Demographics, health characteristics, treatments at baseline and HCRU, including number of pharmacist and physician/nurse practitioner visits over time are reported. QoL-related outcomes were assessed via Self-Esteem and Relationship Questionnaire (SEAR), 2-Item Patient Health Questionnaire, and ratings of sexual satisfaction. Generalized linear models were used to assess the association of sildenafil-P use with total physician/nurse practitioner and pharmacist visits and QoL-related outcomes at 12 months. Results: Overall, 1162 men completed the survey at all 5 time-points. The mean ± SD age was 59.02 ± 12.06 years; 55.42% reported having a moderate-to-severe ED. Hypertension (37.52%) and hypercholesterolemia (31.50%) were the most common risk factors for ED. At baseline, 62.99% were not using any ED treatment. After adjusting for baseline visits/other covariates, mean physician/nurse practitioner (3.68 vs 2.87; P = .003) and pharmacist visits for any reason (2.10 vs 1.34; P < .001) at 12 months were significantly higher among sildenafil-P users than those who never used sildenafil-P. Sildenafil-P users had significantly higher SEAR total and domain (sexual relationship and self-esteem) scores at 12 months. Conclusion: Following the reclassification to pharmacy medicine in the UK, sildenafil-P was associated with a higher number of physician/nurse practitioner and pharmacist visits for any reason. Sildenafil-P use was also associated with better QoL, although group differences were small in magnitude.