Estimating the Prognosis of Canadians Infected With the Hepatitis C Virus Through the Blood Supply, 1986-1990
Date Posted: 2011/04/02
Hepatitis C virus (HCV) is one of the most common causes of liver disease in Canada. Before serologic testing for the presence of hepatitis C became available in 1990, blood transfusion and blood product use were a major source of HCV infection. Between 1986 and 1990, surrogate marker testing was employed to screen blood donors in the United States to reduce the risk of HCV infection in the general population. In Canada, surrogate marker testing was not employed in most jurisdictions. As a result, many individuals in Canada became infected by HCV through blood transfusion and blood products during this time window.
On March 27, 1998 federal, provincial, and territorial governments announced an offer of financial assistance to individuals who were infected with HCV through the blood system between January 1, 1986 and July 1, 1990. In 1999, court orders in British Columbia, Ontario and Québec were obtained approving a settlement agreement which made approximately $1.2 billion available to compensate claimants, who included individuals with transfusion-acquired HCV infection (including hemophilics), those with HIV who became co-infected with HCV, and secondarily infected individuals.
The Canadian compensation program is unique in that it links compensation levels to stage of liver disease. However, the long-term prognosis of HCV infection is uncertain and variable, and experts disagree. In order to assist in ensuring the long-term sufficiency of the fund, a working group was formed in November of 1998 to provide best possible estimates of the prognosis of the HCV-infected post-transfusion compensation claimant (PTCC) cohort. This "medical model", a Markov state-transition model, served as the basis of the actuarial model which estimated future payments from the compensation fund.
The compensation agreement between governments and plaintiffs calls for an estimate of the sufficiency of the compensation fund every three years. In order to assist in the process of assuring the sufficiency of the fund, the original prognostic model has undergone four subsequent revisions. This document describes the fourth revision of the original model. Serial
revision is required because new information regarding both the characteristics of compensation claimants (e.g. HCV stage distribution and size of claimant cohort) and HCV outcome data (e.g. natural history prognostic data, treatment patterns and treatment intensity) continues to become available. Older projections become less accurate as time passes.
The first revision took place in 2002 by a working group which included some members of the original group (Murray Krahn, Jenny Heathcote & Linda Scully) and two new members (Peter Wang & Qilong Yi). There were two major differences in the 2002 prognostic model, in comparison to the original model. The first was that the prognosis of the PTCC cohort was explicitly linked to liver fibrosis stage. This made it considerably easier to use the "medical model" to estimate future payments, as compensation levels were closely linked with fibrosis stage. The second major difference was that we had detailed clinical and demographic data from 2,466 compensation claimants.
The second revision included one new member (Morris Sherman), and differed from the first revision in several aspects. First, the number of compensation claimants increased from 2,446 to 4,530 or by 85%. Thus, the results reflected were more representative of the target cohort. Second, the stage transition probabilities were revised by incorporating data from newly published prognostic studies and transition rates derived directly from the PTCC cohort. In contrast to the previous models, a new method (Markov maximum likelihood estimation, MMLE developed by our group), which does not assume constant fibrosis progression rate, was used to obtain stage-specific transition probabilities. Third, antiviral therapy improved substantially, with combination pegylated interferon (PEG-IFN) and ribavirin therapy proving to be more effective than the standard interferon-based therapies, and became the standard of care in the past few years. A meta-analysis was thus performed to estimate sustained virologic response (SVR) rates in patients treated with PEG-IFN and ribavirin. Fourth, a revised survey of hepatologists to evaluate practice patterns with respect to antiviral therapy was incorporated into the 2005 model.
The third revision included two members from the previous revisions (Murray Krahn & Qilong Yi) and one new member (Hla-Hla Thein). It retains all the objectives of the second revision: i) update literature review regarding transition probabilities; ii) use the most current data from the post-transfusion compensation claimants; and iii) project future outcomes. Besides that the number of compensation claimants increased from 4,530 to 5,004 in this revision, the methodology used is almost the same as in the 2005 revision. However, in order to obtain more precise transition estimates, the stage-specific transition probabilities were revised by performing a new comprehensive meta-analysis involving transition probabilities derived from the 111 literature-based studies (both English and non-English) and from transition probabilities derived directly from the PTCC cohort. Further, these transition probabilities were adjusted, taking into account study design and clinical factors.
The third revision differs from the previous revisions in a number of ways: i) conducted more comprehensive systematic reviews (i.e., transition probabilities, effect of HIV on fibrosis progression, excess mortality associated with HIV infection, and HCV treatment efficacy); ii) adjustment for the effect of study design and clinical factors on disease progression; iii) and revision of the link between compensation level and fibrosis stage distribution (i.e., level 3, non-bridging fibrosis has been distributed to F1/F2 and level 4, bridging fibrosis to F3 instead of F1 and F2/F3, respectfully). This does not appear to change the overall results substantially in the short-term.
This fourth revision was conducted by the same members as in the 2007 revision. It differed from the previous revisions in several aspects: i) in all models, claimants were categorized into different disease stages based on clinical symptoms and laboratory testing results, which is independent of the compensation level. In the previous models, claimants who were classified as Level 2, but had negative RNA or did not have RNA result, other symptoms to support fibrosis, or cirrhosis diagnosis were reclassified as Compensation Level 1 (HCV antibody positive/RNA negative). In our current model, there were 254 claimants who were classified as Level 2; all were HCV antibody positive, but 44 had negative RNA and 210 had missing RNA results. There were 20 deaths and 234 were still alive. These were a group of claimants whose physicians had not confirmed Level 2 but on discussion with the Crawford Class Action Services Administrator, it was confirmed that PCR positive tests had been provided to the Administrator for all claimants approved at Level 2. Therefore, these cases remained classified as Compensation Level 2 (HCV RNA positive); ii) the model was revised with an additional health state transition from HCC to liver transplantation; iii) annual HCV treatment rates from the cohort rather than expert estimates were used; and iv) previous HCV treatment was considered. Approximately 30% of living patients in 2010 whose current stage between fibrosis stage 1 and fibrosis stage 4 (compensated cirrhosis) were assumed cured after treatment and would have a lower progression rates. Transition probabilities (particularly, clinical stages), age and sex distribution, and initial stage distribution were updated.
For the overall living patients, our model predicts that the prevalence of cirrhosis in August 2010 is 10.0% (Table 8.1.1). The cumulative lifetime incidence of cirrhosis is 38.5%, and 24.0% will ultimately die of liver disease. Our model also predicts that 34.5% of non-hemophilic patients
alive in 2010 will ultimately develop cirrhosis, and 20.4% will ultimately die of liver disease. Because hemophilic patients are younger, and are frequently co-infected with HIV, they will have higher cumulative rates of cirrhosis and liver-related death (51.6% and 35.6%, respectively).
Predictions of the current model relative to those of the earlier three models are reported in Tables 7.2 and 7.3. There is no substantial difference in the prognostic projections between the current model and the 2007 model. The differences between more recent revisions (2004, 2007 and 2010) and 2002 revision are attributable to several factors. First, there are now more claimants in early HCV stages (F0 and F1) than in the 2002 model (76.4% in 2004 and 82.6% in 2007 and 79.7% in 2010 vs. 63.9% in 2002) (Table 7.2). Second, the stage transition probabilities used in the recent projections were adjusted downward after incorporating several newly published studies and data from the compensation cohort. Third, more effective treatment (combination PEG-IFN and ribavirin) is now available.
This document reports specific projections for 10-year age strata for individuals with transfusion acquired hepatitis C infection who are hemophilics as well as those who are non-hemophilics. We also report sensitivity analyses that estimate the degree of uncertainty associated with these projections.
As in the previous reports, the limitation of the prognostic model relates to the availability of biopsy data of PTCC cohort. The true extent of liver damage at the time of claim is unknown in approximately 80% of the PTCC cohort. However, the current model likely represents the state- of-the-art in estimation of HCV prognosis among this cohort. It is possible that the number and stage distribution of compensation claimants might be almost complete as the number of claimants has increased by only 4.4% (from 5,004 in 2007 to 5,225 in 2010). The number of outstanding claims is believed to be relatively small.
This work represents a multidisciplinary effort between experts in hepatitis C clinical care, epidemiology, biostatistics, and decision modeling. It represents a unique application of decision modeling methods to a public policy question of great import to Canadians. It provides separate estimates for hemophilics and non-hemophilics, and allows estimates of the sufficiency of the compensation fund to rest on the best current evidence. The model incorporates meaningful clinical data to estimate stage distribution and the direct estimation of annual treatment rates from the cohort. Analysis of the large PTCC cohort dataset makes it possible to more accurately estimate the stage distribution of compensation claimants and to estimate patient-derived transition probabilities. The best possible current and future predictions are produced using both literature- and patient-derived stage-specific transition probabilities, taking into account study design and clinical factors. Finally, this work provides uniquely detailed prognostic estimates that will be of value to HCV patients and their physicians who want to know what the future holds for them.
Read the Paper here