Explainable machine learning model for predicting skeletal muscle loss during surgery and adjuvant chemotherapy in ovarian cancer.

BACKGROUND: Skeletal muscle loss during treatment is associated with poor survival outcomes in patients with ovarian cancer. Although changes in muscle mass can be assessed on computed tomography (CT) scans, this labour-intensive process can impair its utility in clinical practice. This study aimed to develop a machine learning (ML) model to predict muscle loss based on clinical data and to interpret the ML model by applying SHapley Additive exPlanations (SHAP) method. METHODS: This study included the data of 617 patients with ovarian cancer who underwent primary debulking surgery and platinum-based chemotherapy at a tertiary centre between 2010 and 2019. The cohort data were split into training and test sets based on the treatment time. External validation was performed using 140 patients from a different tertiary centre. The skeletal muscle index (SMI) was measured from pre- and post-treatment CT scans, and a decrease in SMI ≥ 5% was defined as muscle loss. We evaluated five ML models to predict muscle loss, and their performance was determined using the area under the receiver operating characteristic curve (AUC) and F1 score. The features for analysis included demographic and disease-specific characteristics and relative changes in body mass index (BMI), albumin, neutrophil-to-lymphocyte ratio (NLR), and platelet-to-lymphocyte ratio (PLR). The SHAP method was applied to determine the importance of the features and interpret the ML models. RESULTS: The median (inter-quartile range) age of the cohort was 52 (46-59) years. After treatment, 204 patients (33.1%) experienced muscle loss in the training and test datasets, while 44 (31.4%) patients experienced muscle loss in the external validation dataset. Among the five evaluated ML models, the random forest model achieved the highest AUC (0.856, 95% confidence interval: 0.854-0.859) and F1 score (0.726, 95% confidence interval: 0.722-0.730). In the external validation, the random forest model outperformed all ML models with an AUC of 0.874 and an F1 score of 0.741. The results of the SHAP method showed that the albumin change, BMI change, malignant ascites, NLR change, and PLR change were the most important factors in muscle loss. At the patient level, SHAP force plots demonstrated insightful interpretation of our random forest model to predict muscle loss. CONCLUSIONS: Explainable ML model was developed using clinical data to identify patients experiencing muscle loss after treatment and provide information of feature contribution. Using the SHAP method, clinicians may better understand the contributors to muscle loss and target interventions to counteract muscle loss.

Association of prognostic nutritional index with muscle loss and survival in patients with ovarian cancer treated with primary debulking surgery and chemotherapy.

PURPOSE: Sarcopenia is prevalent in ovarian cancer and contributes to poor survival. This study is aimed at investigating the association of prognostic nutritional index (PNI) with muscle loss and survival outcomes in patients with ovarian cancer. METHODS: This retrospective study analyzed 650 patients with ovarian cancer treated with primary debulking surgery and adjuvant platinum-based chemotherapy at a tertiary center from 2010 to 2019. PNI-low was defined as a pretreatment PNI of < 47.2. Skeletal muscle index (SMI) was measured on pre- and posttreatment computed tomography (CT) at L3. The cut-off for the SMI loss associated with all-cause mortality was calculated using maximally selected rank statistics. RESULTS: The median follow-up was 4.2 years, and 226 deaths (34.8%) were observed. With a median duration of 176 days (interquartile range: 166-187) between CT scans, patients experienced an average decrease in SMI of 1.7% (P < 0.001). The cut-off for SMI loss as a predictor of mortality was - 4.2%. PNI-low was independently associated with SMI loss (odds ratio: 1.97, P = 0.001). On multivariable analysis of all-cause mortality, PNI-low and SMI loss were independently associated with all-cause mortality (hazard ratio: 1.43, P = 0.017; hazard ratio: 2.27, P < 0.001, respectively). Patients with both SMI loss and PNI-low (vs. neither) had triple the risk of all-cause mortality (hazard ratio: 3.10, P < 0.001). CONCLUSIONS: PNI is a predictor of muscle loss during treatment for ovarian cancer. PNI and muscle loss are additively associated with poor survival. PNI can help clinicians guide multimodal interventions to preserve muscle and optimize survival outcomes.

Impact of Multiple Prescriptions With Anticholinergic Properties on Adverse Clinical Outcomes in the Elderly: A Longitudinal Cohort Study in Taiwan.

The cumulative effect of multiple pharmaceutics with anticholinergic properties (i.e., anticholinergic burden), can serve as an indicator of suboptimal prescribing in the elderly. Yet, no research is available concerning the effect of different compositions of score on adverse outcomes under the same anticholinergic burden. This population-based cohort study investigated whether different combinations of medications with anticholinergic properties have different impacts on adverse outcomes in the elderly using the Anticholinergic Risk Scale (ARS) and Anticholinergic Cognitive Burden Scale (ACB) scores. We included 116,043 people aged 65 years and older from Taiwan's Longitudinal Health Insurance Database and measured their monthly anticholinergic burden over a 10-year follow-up period (from January 1, 2002, to December 31, 2011). We analyzed the association between different anticholinergic score compositions and adverse outcomes (emergency department visits, all-cause hospitalizations, fracture-specific hospitalizations, and incident dementia) via generalized estimating equations. Cumulative effects of multiple medications with low anticholinergic activity were associated with a greater risk for emergency department visits and all-cause hospitalizations (emergency department visits for 65-74 year olds (y/o): ACB 1 + 1 + 1, adjusted odds ratio (aOR) 2.05 (1.99-2.12); ACB 1 + 2, aOR 2.04 (1.91-2.17); and ACB 3, aOR 1.62 (1.57-1.66)). In contrast, using medications with greater potency had a greater impact on central adverse outcomes (incident dementia for 65-74 y/o: ACB 1 + 1 + 1, aOR 3.30 (2.84-3.84); ACB 1 + 2, aOR 5.84 (4.59-7.41); and ACB 3, aOR 9.15 (8.38-9.99)). The quantity of anticholinergics (even with low score) an older person used matters in risk of emergency department visit and all-cause hospitalization but the potency of anticholinergics (i.e., those with high score) matters in risk of fracture-specific hospitalization and incident dementia.

Comparative Associations Between Measures of Anti-cholinergic Burden and Adverse Clinical Outcomes.

PURPOSE: No consensus has been reached regarding which anticholinergic scoring system works most effectively in clinical settings. The aim of this population-based cohort study was to examine the association between anticholinergic medication burden, as defined by different scales, and adverse clinical outcomes among older adults. METHODS: From Taiwan's Longitudinal Health Insurance Database, we retrieved data on monthly anticholinergic drug use measured by the Anticholinergic Risk Scale (ARS), the Anticholinergic Cognitive Burden Scale (ACB), and the Drug Burden Index - Anticholinergic component (DBI-Ach) for 116,043 people aged 65 years and older during a 10-year follow-up. For all 3 scales, a higher score indicates greater anticholinergic burden. We used generalized estimating equations to examine the association between anticholinergic burden (ARS and ACB: grouped from 0 to ≥4; DBI-Ach: grouped as 0, 0-0.5, and 0.5-1) and adverse outcomes, and stratified individuals by age-group (aged 65-74, 75-84, and ≥85 years). RESULTS: Compared with the ARS and DBI-Ach, the ACB showed the strongest, most consistent dose-response relationships with risks of all 4 adverse outcomes, particularly in people aged 65 to 84 years. For example, among those 65 to 74 years old, going from an ACB score of 1 to a score of 4 or greater, individuals' adjusted odds ratio increased from 1.41 to 2.25 for emergency department visits; from 1.32 to 1.92 for all-cause hospitalizations; from 1.10 to 1.71 for fracture-specific hospitalizations; and from 3.13 to 10.01 for incident dementia. CONCLUSIONS: Compared with the 2 other scales studied, the ACB shows good dose-response relationships between anticholinergic burden and a variety of adverse outcomes in older adults. For primary care and geriatrics clinicians, the ACB may be a helpful tool for identifying high-risk populations for interventions.