Researchers examined whether an externally validated clinical model could predict the individualized risk of serious muscle disorders among patients potentially eligible for statin treatment.
In a retrospective cohort study using linked electronic health record data from England's Clinical Practice Research Datalink, the researchers used the Gold database to derive the model and the independent Aurum database for external validation among male participants aged 50 years or older and female participants aged 60 years or older who were registered between January 1, 1998, and December 31, 2018. The age criteria corresponded to a population with an estimated 10-year cardiovascular disease risk of roughly 5% or greater according to QRISK2. The primary outcome was the first hospital admission or mortality with a diagnosis of a serious muscle disorder, identified using International Classification of Diseases, Tenth Revision codes. Mild muscle symptoms, including myalgia, were excluded. The researchers developed a Fine-Gray competing-risk model to estimate 1-, 5-, and 10-year risks while accounting for other causes of mortality. Candidate predictors were selected through a literature review, clinical consultation, and assessment of routinely available primary care data. Missing predictor data were handled using multiple imputation by chained equations.
The derivation cohort included nearly 1.8 million patients with a median follow-up of 6.6 years, and the external validation cohort included nearly 3.9 million patients with a median follow-up of 8.5 years. Serious muscle disorders occurred in 0.30% (n = 5,723) and 0.35% (n = 13,549) of the patients in the derivation cohort and validation cohort, respectively, corresponding to adjusted 10-year cumulative incidences of 0.45% and 0.44%. The final model included 22 predictors, including age, sex, frailty, previous muscle disorders, comorbidities, vitamin deficiencies, recorded statin prescriptions, and concomitant drugs with myotoxic potential. Previous muscle disorders were the strongest predictor of serious muscle disorders, followed by vitamin D deficiency and use of other myotoxic drugs. Recorded prescriptions for atorvastatin, rosuvastatin, simvastatin, and fluvastatin or pravastatin were each associated with higher predicted risks compared with no statin prescription.
External validation showed good overall model performance. The model achieved C-index values of 0.84, 0.81, and 0.78 for 1-, 5-, and 10-year prediction, respectively. Calibration was good overall for the 10-year prediction, although the model modestly underestimated risk for the 1- and 5-year predictions, particularly among patients in the highest-risk group. Decision curve analysis suggested greater modeled net benefit across a range of risk thresholds compared with strategies based on taking action in all patients, no patients, or patients’ experience of muscle issues. The researchers found that 99.6% of the patients in the validation cohort had a predicted 10-year risk below 10%. Among the subgroup with a recorded QRISK2 score above 10%, 98.1% had a predicted 10-year risk of muscle issues 10%, and 62.5% were not receiving statins.
“Our model could be included in a patient clinical decision aid, together with a prediction model for [cardiovascular disease] risk [...] to help patients and physicians better understand the benefit-to-harm balance of statin treatment at a personal level and support well-informed decision-making,” wrote lead study author Ting Cai, DPhil, of the Nuffield Department of Primary Care Health Sciences at the University of Oxford, and colleagues.
The researchers noted that the model was developed and validated using routine primary care data from England and requires further validation in other health care settings and populations. They acknowledged that genetic predisposition, physical activity, and other potential risk factors were not included because such information is not usually available in UK primary care records. Because cardiovascular risk was poorly recorded in the data, the researchers used a cutoff age as a proxy for predicted cardiovascular risk according to the QRISK2 model to define a population eligible for statins. The researchers concluded that the model could complement existing cardiovascular disease risk prediction tools to support shared decision-making regarding statin treatment, although additional research is needed to evaluate its clinical impact and cost-effectiveness.
The study was funded by the British Heart Foundation, Wellcome Trust, and Royal Society and supported in part by the National Institute for Health and Care Research. The study authors reported no competing interests.
Source: The Lancet Digital Health
