We read with great interest the systematic review, by Shaheen and Myers, concerning the diagnostic accuracy of the aspartate aminotransferase-to-platelet ratio index (APRI) for the prediction of hepatitis C–related fibrosis.1
The authors must be congratulated for this difficult work, but there are several methodological limitations that impaired their results and conclusions.
Major limitations of this systematic review included the absence of critiques concerning the use of upper normal limits of aminotransferases, the absence of standardization of the area under the receiver operating characteristic curves (AUROCs) according to the prevalence of stages defining advanced or nonadvanced fibrosis, the non–evidence-based statements concerning composite panels, the absence of a systematic review of direct comparisons between biomarkers and APRI, and the absence of discussion of the impossible perfect biomarker.
A major disadvantage of APRI is the use of upper limits of normal (ULNs) for aminotransferases, which are not standardized and not reproducible between laboratories.2, 3 ULN of aminotransferase activity can vary from 26 IU/L in females to 66 IU/L in males with a body mass index > 23 kg/m2,3 and therefore the use of ULN should be abandoned.4
The methodology used by the authors did not take into account the most important source of heterogeneity between studies, which is the different spectrum of fibrosis stages (spectrum bias).5 If a study is overrepresented in fibrosis extreme stages (F0 and F4), the marker sensitivity, specificity, and AUROCs will be automatically higher than those of a study including only adjacent stages (F2 and F1). Standardization analyses should be performed according to these differences in stages' prevalence defining advanced and nonadvanced fibrosis. The indirect AUROC comparisons are meaningless without these standardizations.5
The authors' statements concerning the composite panels are not evidence-based. FibroTest has been extensively and independently validated, has the same adjusted accuracy for intermediate stages,5 and is even recommended by French health authorities as a first-line estimate of fibrosis with a social security reimbursement approval.6
The authors' statements concerning comparisons between APRI with elastography and with composite panels are misleading in the absence of adjustment or in the absence of direct comparisons.5 We analyzed seven studies, including 1768 hepatitis C virus patients, directly comparing FibroTest with APRI,7–13 five being independent of the FibroTest inventor, one being nonindependent, and one being mixed. Meta-analysis (random effect model) demonstrated a greater observed AUROC for FibroTest [0.83; 95% confidence interval (CI) = 0.80-0.85] versus APRI (0.76; 95% CI = 0.73-0.80) with a 0.07 (95% CI = 0.04-0.10) mean difference (P < 0.001) without heterogeneity (Cochran's Q statistic = 7.2; P = 0.31; Fig. 1). This significant difference persisted when the nonindependent and mixed studies were excluded from analysis: 0.05 (95% CI = 0.01-0.09; P = 0.003). The authors mentioned a prospective study demonstrating a 5-year prognostic value for APRI but forgot to mention that the FibroTest had a significantly higher prognostic value (survival without complications: AUROC = 0.96; 95% CI = 0.93-0.97) than APRI (AUROC = 0.82; 95% CI = 0.66-0.91; P = 0.03).14
For liver fibrosis diagnosis, nonexpert physicians and patients are waiting for an almost perfect test, that is, a biomarker with less than 10% false-positive or false-negative results and more than 99% applicability. This is not possible, even with 40-mm liver biopsy.15 The rate of false positives/negatives of a 25-mm, unfragmented biopsy is still around 20% for the diagnosis of advanced fibrosis in comparison with the perfect gold standard, which is the whole liver.15 Therefore, among the discordances observed between biopsy and biomarker estimates of fibrosis, the cause of failure is frequently biopsy failure.14 It is an illusion to wait for an almost perfect biomarker with an adjusted AUROC greater than 90% for the diagnosis of advanced fibrosis. This point must be explained to patients and health authorities.
From this overview, we finally agree with the authors that APRI has moderate diagnostic utility for the prediction of fibrosis in hepatitis C virus–infected patients. However, we think, as other independent observers do,6 on the basis of evidence-based data that validated biomarkers such as FibroTest-ActiTest have significantly greater diagnostic value and prognostic value than APRI and also provide assessment of necroinflammatory grades.