We appreciated the article from Sebastiani et al. that evaluated the SAFE biopsy algorithm, which sequentially combines the aspartate aminotransferase-to-platelet ratio index (APRI) and FibroTest.1 However, this article deserves several comments.
The authors presented their study as a validation of previously published algorithms.2 However, the algorithms were slightly different. First, the diagnosis of significant fibrosis previously relied on algorithms for normal and abnormal levels of aminotransferases, whereas only the latter was used in the validation study. Second, the APRI cutoffs changed for the cirrhosis diagnosis.
A major disadvantage of a sequential algorithm is to cumulate the misclassification rate of test at each algorithm step or to require a high liver biopsy (LB) rate, which is unsuitable in a screening setting. We evaluated the accuracy of the recent SAFE biopsy algorithm1 in 1056 patients with chronic hepatitis C.3 The SAFE biopsy algorithm provided a very good accuracy for the diagnosis of significant fibrosis (95.8%), but at the cost of a high LB rate (68.3%, Table 1). In order to decrease the LB rate, we combined APRI and Fibrotest in a new test using a synchronous algorithm by binary logistic regression. This test allowed a diagnosis of significant fibrosis with ≥90% negative or positive predictive values in 36.5% of the patients. By using LB in the 63.5% remaining patients (versus 68.3% by SAFE, P = 10−3), overall diagnostic accuracy was 96.0% (versus 95.8% by SAFE, P =0.86). Consequently, we performed the same analysis by combining the tests with the highest accuracy—FibroMeter and Fibrotest—in our population.3 This second synchronous test provided a 95.3% overall diagnostic accuracy (P = 0.58 versus SAFE, Table 1) and 55.2% required LB (P < 10−3 versus SAFE or APRI-Fibrotest synchronous combination).
|Blood Tests||Significant Fibrosis (Metavir F≥2)||Cirrhosis (Metavir F4)|
|SAFE*||APRI + FT†||FM + FT†||SAFE*||APRI + FT†||FM‡ + FT†|
The SAFE biopsy algorithm designed for cirrhosis had a moderate diagnostic accuracy and even a poor positive predictive value (PPV), resulting in a false diagnosis of cirrhosis in approximately 40% of the cases.1 In our population, SAFE biopsy algorithm for cirrhosis had also only 57.3% positive predictive value (Table 1). The synchronous combination of APRI and Fibrotest for the diagnosis of cirrhosis provided a diagnostic accuracy of 95.1% (versus 91.3% by SAFE, P < 10−3), a 9.8% LB requirement (versus 6.8% by SAFE, P = 0.02), and even a dramatic gain in PPV to 100%. A second synchronous combination of FibroMeter4 and Fibrotest provided 94.9% overall diagnostic accuracy (P < 10−3 versus SAFE), with only 4.7% of LB required (P = 0.05 versus SAFE), and a sustained gain in PPV to 98.1%.
Finally, the most accurate combinations of blood test for significant fibrosis or cirrhosis diagnosis rely on synchronous combination of accurate tests rather than on a sequential test combination including a simple test (APRI). This method might be improved, because we have observed in 390 patients with various causes of chronic liver disease that a synchronous combination of a blood fibrosis test with Fibroscan significantly reduced the LB requirement to only 20% for the diagnosis of significant fibrosis and <10% for cirrhosis.5