Quantitative CT is used to estimate LD but is not always available. We hypothesized that non-radiological parameters gathered in clinical practice may predict LD. 136 COPD subjects from a respirology practice completed PRO (mMRC and COPD Assessment Tests), PFT (Jaeger or Vmax22, Carefusion, Germany), OS (tremoFlo C-100, Thorasys, Canada) and a non-research CT scan within 1 year. LD (at 15th percentile) was determined from CT scans using AirwayInspector (airwayinspector.acil-bwh.org). Multiple linear regressions were performed (Stata, USA) using LD vs. BM, PRO, ln transformed OS parameters and PFT from a derivation group (DG) of 68 subjects. The resulting model was applied to data from a validation group (VG) of the remaining 68 subjects. With significance set at p<0.05 after Bonferroni correction, Student t-tests were used to determine differences in BM, PRO, OS and PFT between the 2 groups and none were observed (p=0.56-1.00).
LDmodel=0.37(CAT)-3.91(mMRC)+0.27(BMI)+0.19(FEV1/FVC)+0.72(lnAX)-0.32(TLC %predicted)+ 0.03(FRC %predicted)+0.32(Kco %predicted)+50.17 resulted in a strong correlation (r=0.75, p<0.0001) using DG data. LDmodel vs. LD using VG data showed a strong correlation (r=0.76, p<0.0001). A reasonable estimate of LD, and hence emphysema, is possible by means of non-radiologic parameters. Further work improving such models is warranted.
Keywords : COPD diagnosis, Imaging, Physiology
Dandurand R, Dandurand M, San Jose Estepar R, Bourbeau J, Eidelman D. A statistical model to estimate lung density (LD) utilizing oscillometry (OS), biometrics (BM), patient reported outcomes (PRO) and pulmonary function tests (PFT). Eur Respir J. 2016;48:PA4408.