In this study, we found that serum HA levels were positively correlated with the stages of liver fibrosis. Moreover, the AUC-ROC increased with the stage of fibrosis with the highest value found for cirrhosis. HA was moderately accurate at the diagnosis of F≥2 (AUC-ROC of 0.676), while it seemed to be a very useful method for the detection of cirrhosis (AUC-ROC of 0.863). Many fibrosis experts would consider non-invasive tests for fibrosis with an AUC-ROC value of 0.85-0.90 to be as good as liver biopsies for staging fibrosis . Some authors have argued that some non-invasive markers of fibrosis might be even more accurate than biopsies and that most of the significantly discordant results between biopsies and non-invasive tests may be due to the method of obtaining biopsies that does not demonstrate the actual liver fibrosis state (sampling error when performing the biopsies) .
HA has been described as a component of several fibrosis indexes or as a single parameter for the non-invasive assessment of fibrosis/cirrhosis in HIV/HCV-coinfected patients [22–25]. In HIV/HCV-coinfected patients, there are few published studies with HA alone, and they are limited by their small sample sizes  or were designed to only evaluate significant fibrosis .
The diagnostic performance of HA was similar to the Forns, APRI, FIB-4, HGM-1 and HGM-2 indexes for our HIV/HCV-coinfected patients. We also found that the AUC-ROC of HA was similar to the AUC-ROC values of APRI, FIB-4, and Forns indexes obtained by other authors in HIV/HCV-coinfected patients [26–30] but lower than the AUC-ROC values of APRI, FIB-4, and Forns found in several studies carried out on HCV-monoinfected patients [11, 12, 31]. In summary, according to this study the performance of HA is not better than several biomarkers using parameters easily available in routine clinical practice in HIV/HCV-coinfected patients.
The clinical utility of HA in our study was low except for cirrhosis as the AUC-ROC for cirrhosis was the only one that was higher than 0.850. Also, the NPV was 99%, which could be acceptable for excluding cirrhosis. However, the PPV was only 55%, which is unacceptable for the diagnosis of cirrhosis; although this value can be explained due to the low number of cirrhotic patients in our cohort. Naturally, ruling out cirrhosis is of less importance in the management of patients than confirming such a diagnosis. According to all the available data, the practical interest of the isolated use of HA for assessing liver fibrosis in HIV/HCV-coinfected patients in clinical practice seems to be rather low.
The advantage of HA over the other simple non-invasive indexes (APRI, FIB-4, HGM-1 and HGM-2) is that these indexes could be affected by some factors associated with HIV infection such as biochemical and haematological abnormalities and antiretroviral therapy [32–34], which can lead to an increase in transaminases or cholesterol in the blood [32–34]. HAART has increased the incidence of significant metabolic disturbances. These metabolic disturbances produce clinical manifestations which have an impact on the future health of the HIV-infected patient, including hyperlipidaemia, lipodystrophy, metabolic syndrome, cardiovascular disease and type 2 diabetes [35, 36]. Moreover, hepatotoxicity is a serious complication in patients taking HAART and coinfection with HCV increases the risk of liver toxicity while taking antiretroviral therapy . HCV coinfection is associated with a 2 to 10-fold chance of developing elevated transaminase levels during HAART . The evidence of severe hepatic dysfunction (coagulopathy or elevation of ammonia levels) is suggestive of severe toxicity and HAART should be discontinued. However, the simple indexes (APRI, FIB-4, HGM-1 and HGM-2) are calculated in a relatively easy way using parameters easily available in routine clinical practice. Even though HA is a single molecule, its quantification is not commonly measured in hospitals, it cannot be obtained from normal clinical data, and it is more expensive.
While HA has been shown to be accurate when used in combination with other parameters in HIV/HCV-coinfected patients (SHASTA index , HGM-3 ), its effectiveness at assessing liver fibrosis as an isolated marker is poorer. For instance, according to data published by our group very recently , the AUROC of HGM-3, a combination which includes HA, was 0.939 for F≥3, whereas the corresponding figure for HA alone reported here was 0.772. However, others authors have reported AUC-ROC values for Hepascore, Fibrometer and SHASTA (three indexes which include HA), and Fibrotest  similar to AUC-ROC values of HA in our patients.
Moreover, we used a commercial HA-ELISA test (Echelon Biosciences) different from the enzyme-linked protein binding assay (Hyaluronic Acid Test Kit, Corgenix, Westminster, CO, USA) or the sandwich enzyme binding assay kit (Pharmacia, Uppsala, Sweden) used by others authors [23, 25, 37–41]. To the best of our knowledge, this test has not previously been reported as a fibrosis test. So, this paper is a validation study of the HA-ELISA test (Echelon Biosciences) although the HA levels in this study are quite different (about 10 times larger) from those previously reported in HCV or HIV/HCV patients [23, 25, 38–41].
Aside from these laboratory biomarkers, liver fibrosis is evaluated using transient elastography (FibroScan) . Our group reported an excellent diagnostic performance of liver stiffness for fibrosis and cirrhosis in HIV/HCV-coinfected patients , which was higher than the diagnostic performance of HA shown here.
The diagnostic performance analysis in our cohort had several limitations: a) the low number of patients; b) this study was made on patients with well preserved immune function and the extrapolation to individuals with more marked immune suppression would require further study; c) we did not directly compare HA with SHASTA, Fibrotest, Hepascore or Fibrometer because we did not have all the clinical routine variables needed to calculate these indexes the day the liver biopsy was undertaken, and as a result of the use of the HA-ELISA test (Echelon Biosciences) the values of these combining scores would be quite different from those previously reported; d) we could not give exact information regarding biopsy length or portal tracts, but we found that only 1.68% of biopsies were defective for pathological diagnosis, and these cases were excluded from this study; e) only one pathologist read the biopsies and the biopsies were not validated by someone else; f) the uneven distribution of the stages of fibrosis in our cohort with a high proportion of absent to mild fibrosis and a low proportion of cirrhosis (11%). However, we carried out an analysis using the DANA method, which is used when the distribution of fibrosis stages are highly asymmetric , and we did not find a significant increase in AUC-ROC values (data not shown).