A predictive tool for an effective use of (18)F-FDG PET in assessing activity of sarcoidosis

(18)F-FDG PET/CT (PET) is useful in assessing inflammatory activity in sarcoidosis. However, no appropriate indications are available. The aim of this study was to develop a prediction rule that can be used to identify symptomatic sarcoidosis patients who have a high probability of PET-positivity. We retrospectively analyzed a cohort of sarcoidosis patients with non organ specific persistent disabling symptoms (n = 95). Results of soluble interleukin-2 receptor (sIL-2R) assessment and high-resolution computed tomography (HRCT) were included in the predefined model. HRCT scans were classified using a semi-quantitative scoring system and PET findings as positive or negative, respectively. A prediction model was derived based on logistic regression analysis. We quantified the model's performance using measures of discrimination and calibration. Finally, we constructed a prediction rule that should be easily applicable in clinical practice. The prediction rule showed good calibration and good overall performance (goodness-of-fit test, p = 0.78, Brier score 20.1%) and discriminated between patients with positive and negative PET findings (area under the receiver-operating characteristic curve, 0.83). If a positive predictive value for the presence of inflammatory activity of ≥90% is considered acceptable for clinical decision-making without referral to PET, PET would be indicated in only 29.5% of the patients. Using a positive predictive value of 98%, about half of the patients (46.3%) would require referral to PET. The derived and internally validated clinical prediction rule, based on sIL-2R levels and HRCT scoring results, appeared to be useful to identify sarcoidosis patients with a high probability of inflammatory activity. Using this rule may enable a more effective use of PET scan for assessment of inflammatory activity in sarcoidosis.