Verification bias occurs when the probability of confirming the diagnosis in a gold standard test depends on the results of a screening test. We sought to determine the sensitivity and specificity, before and after correction for verification bias, of estimated pulmonary artery systolic pressure (PASP) measured by echocardiogram (echo) in the diagnosis of pulmonary arterial hypertension (PAH) in patients referred to our pulmonary hypertension center.
We retrospectively reviewed all patients who had a right heart catheterization (RHC) within 120 days of echo to determine the sensitivity and specificity of estimated PASP > 35mmHg on echo in the diagnosis of PAH (mean pulmonary artery pressure > 25mmHg and pulmonary capillary wedge pressure < 18mmHg by RHC). Sensitivity and specificity corrected for verification bias was calculated using the Diamond method (Diamond GA. Work-up Bias. J Clin Epidemiol. 1993;46:207–208). This method consists of a formulation that combines the data used to calculate sensitivity and specificity with data from patients who had echo but no RHC. By ROC analysis, PASP > 45mmHg on echo was determined as the best discriminator of PAH.
242 patients received an echo (103 received RHC). 98 of 191 patients with PASP > 35mmHg on echo had no RHC. 41 of 51 patients with PASP < 35mmHg on echo had no RHC. PASP > 35mmHg on echo had sensitivity and specificity of 94% and 39% respectively. The values corrected for verification bias were 87% and 61% respectively. 64 of 147 patients with PASP > 45mmHg on echo (as determined by ROC analysis) had no RHC. 75 of 95 patients with PASP < 45mmHg on echo had no RHC. PASP > 45mmHg on echo had sensitivity and specificity of 89% and 77% respectively. The values corrected for verification bias were 75% and 90% respectively.
Verification bias exists in the use of echo for the evaluation of PAH. After correcting for verification bias we demonstrated a reduction in the previously expected sensitivity and an increase in specificity.
Physicians should be aware of verification bias in the evaluation of patients for pulmonary hypertension.
Nick Patel, No Financial Disclosure Information; No Product/Research Disclosure Information