Bicarbonate from arterial blood gas analysis as predictor of sleep-related hypoventilation: a diagnostic accuracy study

Background Bicarbonate above the upper limit of normal is used as an indicator of sleep-related hypoventilation in patients with diseases frequently associated with hypoventilation, such as obesity or neuromuscular diseases. Research question The aim was to evaluate the diagnostic accuracy of bicarbonate from daytime arterial blood gas analysis for sleep-related hypoventilation measured via transcutaneous nocturnal capnometry. Methods In a diagnostic accuracy study, the sensitivity and specificity of bicarbonate concentration for detecting nocturnal hypoventilation were tested using a receiver operating curve (ROC) and logistic regression analysis in patients undergoing both an in-laboratory sleep study with transcutaneous capnometry and daytime arterial blood gas analysis. Results In 510 patients (mean±SD age 52±15 years, BMI 34±10 kg/m $^{2}$ ), the area under the ROC curve for bicarbonate (HCO $_{3-}$ ) and median nocturnal tcpCO $_{2}$ ≥50 mm Hg (45% yes, 55% no) was 0.70 (95% CI 0.65 to 0.75). Sensitivity of HCO $_{3-}$ of 26.5 mmol/L and 27.0 mmol/L was 61.0% and 54.1%, specificity 71.3% and 79.6%, and the likelihood ratio for sleep-related hypoventilation (median tcpCO $_{2}$ ≥50 mm Hg) 2.1 and 2.6. HCO $_{3-}$ was significantly higher in those with (n=231, 27.6±3.0 mmol/L) versus without (n=279, 25.8±1.9 mmol/L) sleep-related hypoventilation (p<0.001). Increasing awake HCO $_{3-}$ was associated with a higher likelihood of nocturnal hypoventilation (OR 1.4 (95% CI 1.3 to 1.5), p<0.001). Conclusions The sensitivity of elevated bicarbonate from awake arterial blood gas analysis using common cut-off values such as 27 mmol/L is only moderate, and bicarbonate as an integral part of the assessment of sleep hypoventilation (likelihood ratio) must be interpreted in the context of other assessments. A normal bicarbonate does not rule out sleep hypoventilation.