Background The presence and progression of hepatic (liver) fibrosis into cirrhosis is a prognostic variable having impact on survival in people with alcoholic liver disease. Liver biopsy, although an invasive method, is the recommended 'reference standard' for diagnosis and staging of hepatic fibrosis in people with liver diseases. Transient elastography is a non-invasive method for assessing and staging hepatic fibrosis. Objectives To determine the diagnostic accuracy of transient elastography for diagnosis and staging hepatic fibrosis in people with alcoholic liver disease when compared with liver biopsy. To identify the optimal cut-off values for differentiating the five stages of hepatic fibrosis. Search methods The Cochrane Hepato-Biliary Group Controlled and Diagnostic Test Accuracy Studies Registers, The Cochrane Library, MEDLINE (OvidSP), EMBASE (OvidSP), and the Science Citation Index Expanded (last search August 2014). Selection criteria Diagnostic cohort and diagnostic case-control study designs that assessed hepatic fibrosis in participants with alcoholic liver disease with transient elastography and liver biopsy, irrespective of language or publication status. The study participants could be of any sex and ethnic origin, above 16 years old, hospitalised or managed as outpatients. We excluded participants with viral hepatitis, autoimmunity, metabolic diseases, and toxins. Data collection and analysis We followed the guidelines in the draft Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy. Main results Five retrospective and nine prospective cohort studies with 834 participants provided data for the review analyses. Authors of seven of those studies sent us individual participant data. The risk of bias in the included studies was high in all but three studies. We could identify no serious concerns regarding the applicability of the studies in answering the main study question of our review, namely to use transient elastography to diagnose hepatic fibrosis. We could not identify the optimal cut-off values for the fibrosis stages. The definition of the diagnosis of alcoholic liver disease was not provided in one study and was not clearly defined in two studies, but it was clear in the remaining 11 studies. The study authors used different liver stiffness cut-off values of transient elastography for the hepatic fibrosis stages. There was only one study (103 participants) with data on hepatic fibrosis stage F1 or worse, with a cut-off of 5.9 kPa, and reporting sensitivity of 0.83 (95% confidence interval (CI) 0.74 to 0.90) and specificity of 0.88 (95% CI 0.47 to 1.00). The summary sensitivity and specificity of transient elastography for F2 or worse (seven studies with 338 participants and with cut-offs around 7.5 kPa (range 7.00 to 7.8 kPa)) were 0.94 and 0.89 with LR+ 8.2 and LR- 0.07, which suggests that transient elastography could be useful to rule out the presence of significant hepatic fibrosis, thus avoiding liver biopsy. Due to the wide range of cut-off values (from 8.0 to 17.0 kPa) found in the 10 studies with 760 participants with hepatic fibrosis F3 or worse, we fitted a hierarchical summary receiver operating characteristic (HSROC) model and estimated a summary ROC (SROC) curve. The sensitivity of the 10 studies varied from 72% to 100% and the specificity from 59% to 89%. We performed an additional analysis by including the studies with a cut-off value of around and equal to 9.5 kPa (range 8.0 to 11.0 kPa). The summary sensitivity and specificity of transient elastography (eight studies with 564 participants) were 0.92 and 0.70 with LR+ 3.1 and LR- 0.11, which suggests that transient elastography could also be useful to rule out the presence of severe hepatic fibrosis (F3 or worse), avoiding liver biopsy. We carried out a sensitivity analysis by considering only the studies with a cut-off value equal to 9.5 kPa and the result did not differ. We performed an HSROC analysis and reported an SROC curve for hepatic fibrosis stage F4 (cirrhosis). The HSROC analysis suggested that when the cut-off value changes, there is a wide variation in specificity and a more limited variation in sensitivity. We performed an additional analysis with the studies with the most commonly used cut-off value of 12.5 kPa. The summary sensitivity and specificity of transient elastography (seven studies with 330 participants) were 0.95 and 0.71 with LR+ 3.3 and LR- 0.07, which again suggests that transient elastography could be useful to rule out the presence of cirrhosis, avoiding liver biopsy. Authors' conclusions We identified a small number of studies with a few participants and were unable to include several studies, which raises the risk of outcome reporting bias. With these caveats in mind, transient elastography may be used as a diagnostic method to rule out liver cirrhosis (F4) in people with alcoholic liver disease when the pre-test probability is about 51% (range 15% to 79%). Transient elastography may also help in ruling out severe fibrosis (F3 or worse). Liver biopsy investigation remains an option if the certainty to rule in or rule out the stage of hepatic fibrosis or cirrhosis remains insufficient after a clinical follow-up or any other non-invasive test considered useful by the clinician. The proposed cut-off values for the different stages of hepatic fibrosis may be used in clinical practice, but caution is needed, as those values reported in this review are only the most common cut-off values used by the study authors. The best cut-off values for hepatic fibrosis in people with alcoholic liver disease could not be established yet. In order to diagnose correctly the stage of hepatic fibrosis in people with alcoholic liver disease using transient elastography assessment, the studies should consider a single aetiology. Hepatic fibrosis should be diagnosed with both transient elastography and liver biopsy and in this sequence, and transient elastography cut-off values should be pre-specified and validated. The time interval between the two investigations should not exceed three months, which is the interval mainly valid for people without cirrhosis, and assessment of results should be properly blinded. Only studies with low risk of bias, fulfilling the Standards for Reporting of Diagnostic Accuracy may answer the review question.
