HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

Electronic Letters to:

Proteomics and Protein Markers: Enrico Rossi, Leon Adams, Alexander Prins, Max Bulsara, Bastiaan de Boer, George Garas, Gerry MacQuillan, David Speers, and Gary Jeffrey Validation of the FibroTest Biochemical Markers Score in Assessing Liver Fibrosis in Hepatitis C Patients Clin Chem 2003; 49: 450-454 [Abstract] [Full text] [PDF]

eLetters: Submit a response to this article

Electronic letters published:

Fibrotest even better than liver biopsy ?

Thierry Poynard, Thierry Poynard, Françoise Imbert-Bismut, Vlad Ratziu, Rob Myers, Vincent Di Martino, Dominique Thabut, Joseph Moussalli and Yves Benhamou.   (10 March 2003)

Response to Poynard et al

Enrico Rossi, Leon Adams, Alexander Prins, Max Bulsara, Bastiaan de Boer, George Garas, Gerry MacQuillan, David Speers, Gary Jeffrey   (21 March 2003)

Fibrotest even better than liver biopsy ? 10 March 2003
Thierry Poynard, Prof Hepato-Gastroenterologie Groupe Hospitalier Pitie-Salpetriere, Thierry Poynard, Françoise Imbert-Bismut, Vlad Ratziu, Rob Myers, Vincent Di Martino, Dominique Thabut, Joseph Moussalli and Yves Benhamou. Send letter to journal: Re: Fibrotest even better than liver biopsy ? tpoynard{at}teaser.fr Thierry Poynard, et al.	Sir We read with interest the article by Rossi et al entitled “Validation of the Fibrotest biochemical markers score in assessing liver fibrosis in hepatitis C patients”. We acknowledge the results observed on a small sample size study, but disagree with the interpretation and the conclusion. We were surprised with the opposite conclusion presented in this article in comparison with that of the same work when presented in abstract form: “The FibroTest is reasonably predictive of the absence or presence of METAVIR fibrosis grades F2 to F4, respectively.” (1). As far as the statistical analysis, we were unable to recalculate some indices. In the discussion it is stated that “among 33 patients with scores <0.1, F2F3F4 was observed in 6”. As there were 48 F2F3F4 patients, the sensitivity should be 88 % (42/88) instead of 92 % as described in Table 1 and on page 451; the specificity should be 35% (27/77) and not 29% and the positive predictive value 46% (42/92). It is also stated that “of the 24 patients with scores >0.6 ...F0F1 was observed in 5 patients”. Therefore 19 patients were F2F3F4 and the sensitivity should be 19/48, which is 39.5% instead of 42% as stated in Table 1 and page on 451. This validation study indeed confirms the excellent specificity (94% for a Fibrotest cutoff >0.6) for the diagnosis of significant fibrosis (F2F3F4). This is similar to what we observed in our first study (95% specificity in the second year population). The specificity of Fibrotest can be further improved if identified causes of false positives have been excluded. Our experience in France with 4,000 Fibrotests has shown two main causes of false positives, hemolysis (mainly due to cardiac valvular prosthesis) and Gilbert syndrome (2-5). We wonder if these diagnoses had been ruled out in their 5 false positive cases. Since the specificity is similar in this and our study, the 13% difference observed in positive predictive values, (78% in the validation versus 91% in our study) could be related to differences between prevalence of significant fibrosis stages and/or sensitivity of the Fibrotest. There was indeed a difference in the prevalence of fibrosis stages, 38 % (F2F3F4) in this study versus 45% (F2F3F4) in our study. However the main difference concerns the sensitivity: this was 92% in the Rossi study vs 100% in ours for the 0.10 cutoff, and was 42% vs. 72% respectively, for the 0.60 cutoff. One explanation is the difference in the prevalence of cirrhosis, which was present in only 7% in the Rossi study versus 16 % in our study. Among F2F3F4 patients, the sensitivity of Fibrotest is higher for F4 than for F2. Another reason for discrepancy between the two studies, as discussed by Rossi et al, is a difference in instrumentation. We recently published a study documenting significant variability in Fibrotest results obtained using different kits and automates (5). It is therefore important that the laboratories apply the quality charter when performing the Fibrotest (5), which is available at www.biopredictive.com. A true objective validation of Fibrotest should use this recommended procedure. Rossi et al do not discuss the important limitations of liver biopsy: sampling error, intra and inter pathologist errors and risks. Sampling error is paramount and leads to inaccurate estimates of sensitivity and specificity in small sample size studies. In hepatitis C there is a 33% discordance rate in the same patient for fibrosis staging when assessed in the right and left lobes of the liver (6). This 33% discordance rate should be weighted against the 27% discordance rate observed between Fibrotest and one biopsy in the Rossi study. Therefore discordances between biochemical markers and liver biopsy can be either due to limitations of the biochemical markers or those of liver biopsy. As there is no other “gold standard”, we recently compared the diagnostic values of Fibrotest and Actitest for significant features (A2A3 -F2F3F4 vs A0A1-F0F1) in relation to the size of liver biopsy (greater or less than 15 mm) (3). The area under the ROC curve (AUROC) among 200 patients with a biopsy size <15 mm was 0.705 ± 0.04 (se) which was lower than 0.879 ± 0.03 in 152 patients with biopsy size >15 mm (p<0.001). The AUROC was also significantly lower in 39 patients with less than 6 portal tracts on the histological sample (0.615 ± 0.09) in comparison with 313 patients with 6 or more portal tracts (0.803 ± 0.03 p<0.001). In conclusion we think that a true validation study of Fibrotest must apply the recommended procedures, include several hundred cases and use sensitivity analysis according to biopsy size and number of portal tracts. Even with the diagnostic value observed in this report, (94 % specificity for a 0.6 cutoff and 92% sensitivity for a 0.1 cutoff), the Fibrotest seems to have a much better benefit-risk ratio than liver biopsy, because of the biopsy sampling error (33% discordance between left and right lobe stages) and the biopsy risk (3 deaths out of 10,000 procedures and, 3 severe adverse events out of 1,000). References 1. Adams L, Rossi E, DeBoer B, Speers D, Macquillan G, Garas G, Jeffrey G. Use of Fibrotest to predict liver fibrosis in hepatitis C: a replacement for liver biopsy? Gastroenterology 2002;122: 1615 A.(abstract) 2. Myers RP, Messous D, Thabut D, Imbert-Bismut F, Ratziu V, Mercadier A, Poynard T. Life is possible without biopsy in patients with chronic hepatitis C: validation of biochemical markers of liver fibrosis and activity in 1570 patients and blood donors. Hepatology 2002; 36:351A. (Abstract) 3. Poynard T, McHutchison J, Manns M, Myers RP, Albreht J. Biochemical markers as surrogate markers of liver fibrosis and activity in patients infected by hepatitis C virus: an example in a randomized trial of pegylated-interferon alfa-2b and ribavirin combination. Hepatology 2002;36:351A. (Abstract) 4. Myers RP, Messous D, Thabut D, Imbert-Bismut F, Ratziu V, Mercadier A, Poynard T. The prediction of fibrosis with serum biochemical markers in patients with chronic hepatitis C: Prospective validation in 534 patients. Hepatology 2002;36:351A. (Abstract) 5. Halfon P, Imbert-Bismut F, Messous D, Antoniotti G, Benchetrit D, Cart- Lamy P, Delaporte G, Doutheau D, Klump T, Sala M, Thibaud D, Trepo E, Robert P. Myers RP, Poynard T. A prospective assessment of the inter- laboratory variability of biochemical markers of fibrosis (FibroTest) and activity (ActiTest) in patients with chronic liver disease. Comparative Hepatology 2002;30 December; 2:3. 6. Regev A, Berho M, Jeffers LJ, Milikowski C, Molina EG, Pyrsopoulos NT, Feng ZZ, Reddy KR, Schiff ER. Sampling error and intraobserver variation in liver biopsy in patients with chronic HCV infection. Am J Gastroenterol. 2002;97:2614-8.
Response to Poynard et al 21 March 2003
Enrico Rossi, Biochemist PathCentre, Leon Adams, Alexander Prins, Max Bulsara, Bastiaan de Boer, George Garas, Gerry MacQuillan, David Speers, Gary Jeffrey Send letter to journal: Re: Response to Poynard et al ric.rossi{at}health.wa.gov.au Enrico Rossi, et al.	Sir We presented our work in abstract form [1] with the following conclusion: Nearly half of the patients chronically infected with hepatitis C in our population had a FibroTest score of less than 0.1 or greater than 0.6. In these patients the FibroTest is reasonably predictive of the absence or presence of METAVIR fibrosis grades F2 to F4, respectively. Thus we found FibroTest reasonably predictive only when the score was either less than 0.1 or greater than 0.6, and even in these we subsequently felt the accuracy was insufficient to guide clinical decisions [2]. Poynard et al had difficulty recalculating statistical indices due to the following anomaly. The sensitivity, specificity and predictive values data presented in Table 1 were calculated based on FibroTest scores used as cutoffs, we believe in the same manner as the original report of Imbert -Bismut et al [3]. Thus the sensitivity at a cutoff of 0.1 refers to the 26 patients who had scores below but not equal to 0.1, comprising four of the 48 cases with fibrosis grades F2 to F4, and giving a sensitivity of 44/48 or 0.92. Similarly of the 77 cases with F0 and F1 fibrosis, 22 had FibroTest scores less than 0.1, giving a specificity of 22/77 or 0.29. In the Discussion we state; of the 33 patients with scores less than 0.1, a false negative result, F2 to F4 fibrosis, was observed in six [18%] patients, compared to none in the original report. This should have read of the 33 patients with scores equal or less than 0.1, etc . To be consistent with our definition of cutoff values, we present findings in patients at a cutoff score of 0.1. Of 26 patients who had scores below 0.1, F2 to F4 fibrosis was observed in four patients giving a proportion of false negatives of 15% [4/26], similar to the 18% calculated above. Poynard et al report the main causes of false positives in their experience were hemolysis mainly due to cardiac valvular prosthesis and Gilbert syndrome and asked if we could exclude these in our five patients with FibroTest scores greater than 0.6 and F0 F1 fibrosis. Four of the five cases had serum bilirubin results less than 20 ƒÝmol/L, the fifth had a raised bilirubin at the time of biopsy which later returned to normal [7 ƒÝmol/L], excluding Gilberts in all five cases. Although we cannot exclude hemolysis per se, we can exclude cardiac valvular prosthesis as a possible cause as none of the study patients had such a prosthesis. The causes of the discrepancy between the two studies may include a lower proportion of cases with F4 fibrosis, 7% in our study [2] compared to 16% in the original report [3]. Another possible cause is acknowledged in our paper, we used a different brand of automated nephelometer and a true objective comparison should use the instrumentation specified in the FibroTest quality charter. The limitations of liver biopsy including substantial complications in 0.3% of patients and sampling error were briefly mentioned in our opening paragraph [2], and we welcome the discussion and additional recent references on liver biopsy sampling error given by Poynard et al. Continuing research into the possible applications of serum markers for assessing liver fibrosis is important. References 1. Adams L, Rossi E, DeBoer B, Speers D, Macquillan G, Garas G, Jeffrey G. Use of fibrotest to predict liver fibrosis in hepatitis C : a replacement for liver biopsy ? Gastroenterology 2002; 122: 1615A [abstract]. 2. Rossi E, Adams L, Prins A, Bulsara M, DeBoer B, Garas G, Macquillan G, Speers D, Jeffrey G. Validation of the FibroTest biochemical markers score in assessing liver fibrosis in hepatitis C patients. Clin Chem, 2003; 49: 450-454. 3. Imbert-Bismut F, Ratziu V, Pieroni L, Charlotte F, Benhamou Y, Poynard T; MULTIVIRC Group. Biochemical markers of liver fibrosis in patients with hepatitis C virus infection: a prospective study. Lancet 2001; 357: 1069-75.