CT measurement of central pulmonary arteries to diagnose pulmonary hypertension (PHTN): more reliable than valid?

The association between main pulmonary artery (MPA) size and pulmonary arterial hypertension (PAHTN) is well established; however, the clinical utility of routine measurement of MPA is uncertain due to considerable overlap between normal patients and those with pulmonary hypertension. The lack of diagnostic accuracy could be further degraded by variability among the radiologists. It is unknown whether the addition of right and left pulmonary artery measurements would improve accuracy or further impair it. The purposes of this study are to verify the accuracy of a proposed cutoff value for the size of MPA in the diagnosis PAHTN, to determine the interrater agreement for this measurement, and to determine whether addition of right pulmonary artery (RPA) and left pulmonary artery (LPA) measurement or simple assessment of patient comorbidities can improve the accuracy.

Materials and methods: Patients undergoing both cardiac catheterization and chest computed tomography (CT) within 3 months of each other at a large university hospital between January 2010 and December 2012 were identified. Patients with prior cardiac surgery or congenital heart disease and critically ill patients were excluded from the study population. Patients with pericardial disease or severe lung disease documented on CT examinations were also excluded. From the remaining patients, 45 patients with normal pulmonary artery pressure and 50 patients with proven pulmonary hypertension were selected. Demographic data and clinical information was collected from medical records of these patients. Three radiologists with different years of experience in cardiothoracic imaging measured the MPA, RPA, and LPA diameters on axial images using an electronic ruler on 3D work stations independently and were masked to the patient clinical symptoms, diagnosis, and each other's measurement to prevent bias. Association between MPA diameter (MPAD) and patient characteristics assessed by one-way analysis of variance for scalar measures. Each reader's measurements were used to construct a separate receiver operating curve (ROC) to assess optimal MPA threshold. The ability of an MPA measurement threshold to correctly identify PAHTN was assessed using chi-squared. Chi-squared was also used to assess the effect of categorical comorbidities on false positive diagnosis.

Results: None of the demographic data or patients' factors (age, gender, height, weight, body surface area, and body mass index) were related to the size of MPAD. The distribution of the MPAD was normal in both groups. Based on prior literature, MPAD (≥3.15cm) was selected as the cutoff value to diagnose PAHTN. Review of ROCs did not suggest a superior cutoff value for any reader. Using this threshold per case interrater agreement was good, kappa values >0.65. Based on an average measurement for all three readers, MPAD was 82% sensitive and 62% specific for PAHTN. Limiting positive diagnosis to those subjects with both MPAD ≥3.15 and either enlarged RPA diameter (RPAD) or LPAD diminished sensitivity but did not improve specificity. Defining positive study as the presence of any dilated artery (MPAD, RPAD, or LPAD) increased sensitivity to 94% but decreased specificity to 27%. Comorbidities that might cause fluctuating mean pulmonary artery pressures could not be shown to account for false positive studies. The 29 true negative patients and 16 false positive patients did not differ in the prevalence of obstructive sleep apnea/home oxygen use or documented congestive heart failure/low ejection fraction.

Conclusion: Previously proposed threshold of MPAD ≥3.15cm is likely optimal but is not specific for identifying patient with PAHTN. Interobserver differences in MPAD measurement do not account this inaccuracy. Incorporation or RPA and LPA measurement does not improve diagnostic accuracy of PAHTN, and assessment of comorbidities does not easily identify likely false positive cases. Diagnosis of PAHTN based solely on CT examinations of the chest may not be sufficiently accurate for clinical use.