More information about text formats
I read with great interest the article “The use of pretest probability increases the value of high-resolution CT in diagnosing usual interstitial pneumonia”, by Brownell and colleagues . The study has great methodological strength, and I applaud the authors for such an elegant work. But what really caught my attention was the clear use of pre-test probability and likelihood ratio to establish the diagnosis of usual interstitial pneumonia (UIP) in patients with suspected UIP. I believe this article should change the way we care for those patients.
The study included patients with “possible UIP" and “inconsistent with UIP” patterns on high-resolution computed tomography (HRCT) of the chest. Those patients represent a diagnostic dilemma we commonly face in interstitial lung diseases clinical practice. Three different radiologists (two in the derivation and one in the validation cohort) reviewed the HRCT scans, and most importantly: they were blinded to clinical information and pathology results. All patients had the reference standard surgical lung biopsy, which were prospectively evaluated by expert pathologists.
The likelihood ratio for male patients, with ≥ 60 years-old, and possible UIP with traction bronchiectasis score ≥ 4 was as high as 47 in the derivation cohort. Since likelihood ratios are a ratio of two likelihoods (the likelihood of a test results in disease / the likelihood of the same test result in no disease ), the further away from on...
The likelihood ratio for male patients, with ≥ 60 years-old, and possible UIP with traction bronchiectasis score ≥ 4 was as high as 47 in the derivation cohort. Since likelihood ratios are a ratio of two likelihoods (the likelihood of a test results in disease / the likelihood of the same test result in no disease ), the further away from one the better. So, what a likelihood of 47 means is that male patients with ≥ 60 years-old and possible UIP with traction bronchiectasis score ≥ 4 are 47 times more likely to have UIP versus not have UIP. This would move us from a pretest probability of IPF of 60% to a post-test probability of IPF of 98% in the derivation cohort, for example.
How to translate that into clinical practice? We have to start with an estimation of the pre-test probability of UIP in our patients. For that, it would have been helpful if the authors had used the variables gender and age to stratify the pre-test probability, similarly to the pre-test probability stratification we do in the evaluation for pulmonary embolism (with the Well’s criteria ) and lung cancer (with the Mayo Clinic prediction rule ). Then, we could apply the likelihood ratios of the HRCT findings to reach a post-test probability. It would be useful to know the likelihood ratio of a possible UIP with traction bronchiectasis score ≥ 4 encountered in the present study.
In conclusion, we will probably continue to struggle to make an accurate diagnosis of UIP in our patients with suspected UIP. But at least, this study allows us to struggle with some good data in our hands.
 Brownell R, Moua T, Henry TS, et al. The use of pretest probability increases the value of high-resolution CT in diagnosing usual interstitial pneumonia. Thorax. 2017;72:424–429.
 Richardson WS, Wilson MC, Keitz SA, et al. Tips for Teachers of Evidence-based Medicine : Making Sense of Diagnostic Test Results Using Likelihood Ratios When to Use This Tip When to Use This Tip. J Gen Intern Med. 2007;23:87–92.
 Chunilal SD, Simel DL. Does This Patient Have Pulmonary Embolism ? JAMA J. Am. Med. Assoc. 2014;290:2849–2858.
 Swensen SJ, Silverstein MD, Edell ES, et al. Solitary pulmonary nodules : Clinical prediction model versus physicians. Mayo Cllinic Proc. 1999;74:319–329.