RESUMO
It is now common to have a modest to large number of features on individuals with complex diseases. Unsupervised analyses, such as clustering with and without preprocessing by Principle Component Analysis (PCA), is widely used in practice to uncover subgroups in a sample. However, in many modern studies features are often highly correlated and noisy (e.g. SNP's, -omics, quantitative imaging markers, and electronic health record data). The practical performance of clustering approaches in these settings remains unclear. Through extensive simulations and empirical examples applying Gaussian Mixture Models and related clustering methods, we show these approaches (including variants of kmeans, VarSelLCM, HDClassifier, and Fisher-EM) can have very poor performance in many settings. We also show the poor performance is often driven by either an explicit or implicit assumption by the clustering algorithm that high variance features are relevant while lower variance features are irrelevant, called the variance as relevance assumption. We develop practical pre-processing approaches that improve analysis performance in some cases. This work offers practical guidance on the strengths and limitations of unsupervised clustering approaches in modern data analysis applications.
RESUMO
Context: Despite a high prevalence of obesity in the veteran population, antiobesity medications (AOMs) have been underused in the Veterans Health Administration. Real-world reports on outcomes when AOMs have been used in veterans is limited. Objective: To analyze weight loss outcomes from a local Veterans Health Administration pharmacotherapy-based weight management clinic (WMC). Methods: This was a retrospective cohort study of veterans enrolled in a local WMC for 15 months from August 2016 through September 2018 and followed through November 2019. Patients were offered 1 of 5 available AOMs based on their comorbidities. Factors associated with weight loss (5% or more weight loss) were assessed. Key results: A total of 159 patients were seen in a WMC, 149 (93.7%) veterans were prescribed an AOM, and 129 returned for follow-up. Overall, 61/129 (47%) patients achieved 5% or greater weight loss and 28/129 (22%) achieved 10% or greater weight loss within 15 months. Clinically significant weight loss (%) over the first 15 months was achieved with phentermine/topiramate ER (-6.3%) and liraglutide (-7.5%), but not with orlistat (-3.9%) and lorcaserin (-3.6%). Comorbid obstructive sleep apnea was negatively associated with achieving ≥5% weight loss. Conclusion: Phentermine/topiramate ER and liraglutide were found to be effective AOMs among veterans. Further work is needed to mitigate barriers to AOM initiation given the continued rise in obesity.
RESUMO
BACKGROUND AND OBJECTIVE: There is no consensus on de-escalation of monitoring during active surveillance (AS) for prostate cancer (PCa). Our objective was to determine clinical criteria that can be used in decisions to reduce the intensity of AS monitoring. METHODS: The global prospective AS cohort from the Global Action Plan prostate cancer AS consortium was retrospectively analyzed. The 24656 patients with complete outcome data were considered. The primary goal was to develop a model identifying a subgroup with a high ratio of other-cause mortality (OCM) to PCa-specific mortality (PCSM). Nonparametric competing-risks models were used to estimate cause-specific mortality. We hypothesized that the subgroup with the highest OCM/PCSM ratio would be good candidates for de-escalation of AS monitoring. KEY FINDINGS AND LIMITATIONS: Cumulative mortality at 15 yr, accounting for censoring, was 1.3% for PCSM, 11.5% for OCM, and 18.7% for death from unknown causes. We identified body mass index (BMI) >25 kg/m2 and <11% positive cores at initial biopsy as an optimal set of criteria for discriminating OCM from PCSM. The 15-yr OCM/PCSM ratio was 34.2 times higher for patients meeting these criteria than for those not meeting the criteria. According to these criteria, 37% of the cohort would be eligible for de-escalation of monitoring. Limitations include the retrospective nature of the study and the lack of external validation. CONCLUSIONS: Our study identified BMI >25 kg/m2 and <11% positive cores at initial biopsy as clinical criteria for de-escalation of AS monitoring in PCa. PATIENT SUMMARY: We investigated factors that could help in deciding on when to reduce the intensity of monitoring for patients on active surveillance for prostate cancer. We found that patients with higher BMI (body mass index) and lower prostate cancer volume may be good candidates for less intensive monitoring. This model could help doctors and patients in making decisions on active surveillance for prostate cancer.