Participants from two independent research cohorts were compared. DESTINY-PWS (C601; NCT03440814) was an international, randomized, double-blind, placebo-controlled, parallel-group, Phase 3 study comparing DCCR to placebo in individuals with PWS age 4 years and older with hyperphagia. Dosing in this study was weight-based and targeted an average of 4.2 mg/kg [13]. Following the completion of the double-blind study, participants who sought continued treatment with DCCR were eligible to enroll in clinical study C602 (NCT03714373), a long-term open-label treatment study [14]. PATH for PWS (NCT03718416) is an ongoing natural history study, which enrolled individuals with PWS 5 years and older and was conducted by the Foundation for Prader-Willi Research (FPWR). The objectives of PATH for PWS include advancing the understanding of serious medical events in PWS over a 4-year period, as well as to evaluate how PWS related behaviors change over time. The data from this study are intended to inform the development of, and clinical trial designed for, potential new treatments for PWS. Prior to initiating each study (C601, C602 and PATH for PWS), ethics committee review and approval of the protocol and study related documents was completed. Protocols and study related materials for clinical studies C601 and C602 were reviewed and approved by the WCG Institutional Review Board. The protocol and study related materials for PATH for PWS were reviewed and approved by Hummingbird Institutional Review Board, and the study was subsequently overseen by the North Star Review Board. Participants or their parents/guardians provided informed consent and, as appropriate, assent, prior to being enrolled in each study.

Participants in PATH for PWS were recruited concurrently with the recruitment of participants in C601, providing a contemporaneous population of individuals with PWS from which a cohort could be selected that is comparable to participants randomized in C601, to compare changes in behavior over time.

A statistical Clinical Research Organization (CRO) independent of both Soleno and FPWR prepared and obtained approval from the sponsor for a statistical analysis plan (SAP) prior to the conduct of any analyses. The CRO utilized an unblinded biostatistician who identified participants for inclusion in the PATH for PWS cohort. Once the cohorts were identified and the SAP was approved the CRO was responsible for the analysis of the data. Data from clinical study C601 had already been unblinded, as had a portion of the data from the first 13 weeks of C602, but Soleno was blinded to most of the data from C602 prior to the start of statistical analysis for this comparison. A cohort of participants was identified by the statistical CRO from the PATH for PWS study, which met the key inclusion criteria for C601: baseline HQ-CT score ≥ 13, age 4 years and older, weight between 20 and 135 kg, not enrolled in an interventional clinical study at the first of two consecutive visits approximately 6 months apart with HQ-CT data available at both visits. Most of this cohort also had HQ-CT and data for other assessments available at a subsequent visit, which occurred approximately 1 year after the first visit.

The HQ-CT is a 9-item (range 0–36), caregiver completed questionnaire intended to assess a range of hyperphagia-related behaviors in PWS over a 2-week recall period including: food seeking behaviors, tendency toward being upset or distressed when denied food or asked to stop talking about food and the extent to which these behaviors interfere with daily life [15, 16]. The HQ-CT has been developed and validated for use in the assessment of hyperphagia related behaviors in clinical studies in PWS [13, 17]. The PWSP is a caregiver completed questionnaire that assesses PWS-related behaviors including 51 items which comprise 6 behavioral domains: aggressive behaviors (9 items, range 0–18), anxiety (11 items, range 0–22), compulsivity (10 items, range 0–20), depression (5 items, range 0–10), disordered thinking (6 items, range 0–12) and rigidity/irritability (10 items, range 0–20) [18]. It is intended for use in clinical trials.

Baseline for participants in the C601/C602 cohort was defined as the last measurement of the parameter made prior to initiating DCCR administration. For participants in the DCCR arm of C601, baseline is the C601 baseline measurement, while for participants in the C601 placebo arm, baseline is the end-of-treatment measurement from C601. Hereafter, weeks refer to weeks of DCCR administration. The primary endpoint was HQ-CT change from baseline at 26 weeks. Additional endpoints included HQ-CT change from baseline at 52 weeks, HQ-CT change from baseline at 26 weeks by subgroup of age (< 18 years, ≥ 18 years), PWS genetic subtype (deletion, non-deletion), sex (female, male), growth hormone treatment status (use at both visits, no growth hormone use), and baseline HQ-CT total score (≥ 18, < 18), and changes in behavioral complications of PWS assessed using the PWSP at 26 and 52 weeks.

All statistical tests were performed with a two-sided significance level of α = 0.05. Analyses were based on observed data without imputation for missing results. Comparability of the cohorts was assessed with respect to baseline and demographic characteristics. The primary comparison was the least square mean change from baseline in HQ-CT Total Score at Week 26 between the C601/C602 and the PATH for PWS cohorts from an analysis of covariance model adjusting for baseline HQ-CT Total Score and cohort. Subgroup analyses of HQ-CT change from Baseline were based on an analysis of covariance model adjusting for baseline HQ-CT Total Score and cohort. In addition, all available HQ-CT scores through Week 52 were analyzed using a mixed model for repeated measures with unstructured covariance, baseline HQ-CT, and post-baseline timepoint with an interaction term between cohort and post-baseline timepoint. A sensitivity analysis was conducted in which each missing HQ-CT assessment in the C601/C602 cohort was imputed with the highest possible score for the HQ-CT, including individuals missing data at both 26 and 52 weeks, and would not have been included in the analysis of the primary endpoint (an HQ-CT score of 36, i.e., the worst-case scenario). The resulting dataset was analyzed using the same mixed model for repeated measures as was used for the observed cases. PWSP domains were compared between cohorts based on the Mann-Whitney test. The domains of the PWSP at Week 26 and at Week 52 were also analyzed similarly to the primary ANCOVA for the HQ-CT total Score at Week 26, using their respective baseline scores as covariates.

The lack of randomization when comparing with observational cohorts may result in large differences or imbalances in the values of their respective baseline covariates that can create biased estimates of treatment effects. A methodology used to overcome such biases is the use of propensity scores. Several covariates that have been thought to affect the efficacy outcome as based on scientific rationale or known to influence efficacy as demonstrated in previous studies are considered for inclusion in propensity scores. These propensity scores are then incorporated in the statistical analysis in lieu of multiple covariates. Propensity scores combine all the information from the covariates into a single number that represents the probability of being assigned to the treatment given the covariates. A subset of subjects with similar propensity scores will tend to be similar with respect to values of the baseline and demographic characteristic.

Propensity score adjusted ANCOVA analyses were also conducted. Propensity scores were derived from a logistic regression model adjusting for age, gender, baseline weight (kg), baseline HQ-CT, growth hormone treatment status (on treatment vs. not), region (US vs. Outside the US [OUS]), and PWS genetic type (deletion vs. non-deletion).

Given the concurrent recruitment in the same or similar geographic regions and the use of the same age, weight and HQ-CT inclusion criteria, there was an expectation that the individuals in the two cohorts would be receiving a similar standard of care, and have similar durations of disease and disease severity, limiting the potential for introduction of a bias from these sources in the analysis.