Modeling correlates of low bone mineral density in patients with phenylalanine hydroxylase deficiency

Phenylalanine hydroxylase (PAH) deficiency is an inherited metabolic disorder requiring life-long restriction of dietary protein and phenylalanine-free medical food. Low bone mineral density (BMD) is reported, but factors associated with BMD Z-score (standard deviations from normal) are unknown. We examined associations between clinical and dietary parameters and total BMD Z-score in PAH deficiency patients, and developed models to predict Z-score. Data collected from patients > 4 years of age (n = 88; mean age = 18.8 y; 61 % female) included demographic, clinical, laboratory, and dietary intakes. Adjusted Spearman's correlation coefficients were calculated between parameters and TBMD Z-score, measured by dual energy x-ray absorptiometry (DXA). Parameters approaching significance (p-value < 0.10) were candidate predictors for four linear regression models predicting TBMD Z-score. To validate, model-predicted Z-scores were compared to DXA Z-scores. Mean TBMD Z-score was -0.326; 18 (20.4 %) had Z-score < -1. Z-scores were positively correlated with dietary vitamin D, calcium, and medical food intake and compliance with prescription, and negatively with dietary carbohydrate, sugar, caffeine intake, glycemic load, and prescribed medical food (grams protein/day; p-value < 0.05). The best model included medical food compliance, medical food intake, caffeine intake, and bone-specific alkaline phosphatase (r-square = 0.364). This model predicted Z-score category [normal or low (<-1)] with sensitivity = 66.7 %, likelihood ratio = 14.7, and AUC = 0.83 compared to DXA Z-score. No subjects had low BMD for chronological age (Z-score a parts per thousand currency signaEuro parts per thousand a'2). Compliance with medical food prescription was the strongest predictor of TBMD Z-score. One model, if validated in a separate sample of patients with more cases of low BMD, showed potential to estimate TBMD Z-score using routine clinical patient parameters.