Section 2: Modelling the Incidence and Prevalence of Hepatitis C Infection and its Sequelae in Canada, 2007 – Methods

2. Methods

The present updated modelling study was carried out using the methodology originally developed in 2003 to estimate HCV incidence and prevalence in Canada in 2002. However, while the basic structure of the approach remains the same, several important refinements, modifications and updates were incorporated into the present project.

2.1 HCV infection and outcomes model - overview

The HCV modelling study was carried out in three stages using spreadsheet software (Excel 97, Microsoft Corporation 1999). The first stage obtained estimates of the population at risk stratified by age and gender. The second stage modelled HCV incidence rate among those born in Canada and HCV prevalence at the time of arrival and subsequent incidence for persons immigrating to Canada. The third stage projected outcomes of chronic HCV infection among those infected as estimated in Stage 2.

Further details of the approach used in each of the three stages as follows:

Stage 1

Stage 1 was carried out in two phases. In the first phase, we estimated the populations potentially at risk for HCV infection. The population was treated in two major categories, namely persons born in Canada and persons born elsewhere who immigrated to Canada. In the second phase, we estimated the numbers of persons in four mutually exclusive exposure categories defined as a function of risk of HCV infection, within each of these two populations and together. The four exposure categories were: injection drug users (IDU), recipients of blood transfusion, hemophilia patients and others. For those born in Canada, the birth cohort was modelled from 1960 to 2027 and life table values for mortality specific for gender and age were applied to obtain an estimate of the persons alive at each age by gender for each year from 1960 to 2027. Projections from Statistics Canada were used for the population after 2006³.

Data from Statistics Canada and Citizenship and Immigration Canada were used to estimate the number of immigrants arriving in Canada and the distribution by age and sex of those arriving for each of 84 major countries or groupings of countries produced routinely by Statistics Canada. The resulting populations were compared with Census Canada data stratified on age, gender and whether born in Canada or elsewhere and, for those who immigrated, by country of birth, since 1960. Where digressions from census statistics were greater than 5%, the input parameters, including births, immigrants and life table mortality were adjusted to fit the observed census data.

In the second phase, populations in each of the four exposure categories were obtained using an approach specific to each category as described below. In general, the rate of uptake of HCV-risk associated behaviours and mortality rates were incorporated to estimate age and sex-specific prevalence of the behaviour and vital status. These were then compared to values from observed studies and from previous modelling exercises and adjusted accordingly.

Limited data are available to estimate the number of IDUs in Canada. One study used the capture-recapture methodology to estimate the number of active IDUs in Montreal, Toronto and Vancouver⁴. An independent study estimated the number of IDUs in these three cities based on HIV testing data⁵. In 2000, Eric Single estimated the number of IDUs in Canada to be from 75,000 to 125,000⁶, a number consistent with the results of both these studies.

Triangulation techniques used HIV diagnostic data and estimated HIV prevalence rates for the three major cities (Montreal, Toronto and Vancouver), the rest of the three provinces (Quebec, Ontario and British Columbia) and the other seven provinces to help converge the estimate.

For IDUs, we took into account the proportion of the population initiating injection as a population rate and the proportion of drug users who stop injecting drugs in the course of each year to generate a second population of ex-IDUs. It is important to take this into account in estimating the extent of hepatitis C infection since the burden of HCV infection is considerable among persons who are not actively injecting but may have injected at sometime in the past. A preliminary model of this type had been developed by the contractor for HIV infection among IDUs in New York City, Montreal and Toronto which closely fit the observed HIV prevalence data. In this approach, the number of IDUs was obtained by varying the rates of initiation and cessation of injection drug use so as to “generate” estimates which approximated the existing, though imprecise, estimates of the number of IDUs in Canada. This analysis incorporated values for both active IDUs and ex-IDUs.

Patterns of blood transfusion were based on the modelling work carried out by the Expert Working Group in 1998¹.

For hemophilia patients, a fixed proportion was initially applied to male births and then modified to fit the known numbers of hemophilia patients in Canada, after taking into account mortality rates.

The “Other” category included persons infected by HCV primarily through sexual transmission and through parenteral exposures as health care workers and was estimated by subtraction from the total population after accounting for persons in the three higher risk categories, namely IDUs, blood transfusion recipients and hemophilia patients.

To estimate the population size in each year accurately, the components and rates of mortality were modelled specifically for each of the three major exposure categories. For IDUs, mortality was considered in three categories, namely, 1) life table mortality; 2) mortality directly related to injection drug use, including overdose, serious systemic infections, infection and trauma and 3) mortality due to HIV disease. HIV-related mortality among IDUs began, for the most part, in around 1985. The excess mortality related to overdose, etc. was greater among those who are HIV-infected since these persons tend to be more active injectors.

For blood transfusion recipients, the high level of mortality following transfusion associated with the illness for which the patient was transfused was incorporated. Mortality for blood transfusion recipients was based on the 1998 study¹ which incorporated high level of mortality within the first three years following transfusion with gradually declining excess mortality until 10 years after which life table mortality was used. However, to simplify the model construction, we used a two rather than a three stage approach with an appropriately high mortality in the year following the transfusion and a weighted mortality thereafter to approximate the mortality rates from the 1998 study. The numbers specific for each year by age and sex were generated and then compared to available estimates of populations so that the size of the population at risk was plausible.

For hemophilia patients, data on mortality were derived from the published studies^7-9 and included the effect of the advent of specific treatment for hemophilia in the 1970s, in decreasing mortality. In this population, mortality due to HIV infections acquired in the period 1978-1985 was incorporated in the model.

For persons in all exposure categories, mortality based on lifetable values was applied. This was the only mortality applied to persons in the “Other” exposure category.

It is important to note that HIV incidence was included in the model only to take into account competing mortality from HIV infection. Therefore, HIV infection was limited to exposure categories with relatively high rates of HIV, namely IDUs and hemophilia patients. Also, to simplify the model and due to the higher incidence and prevalence of HCV in these groups, incident HIV infections were considered to occur only in persons already infected by HCV. Thus, the final estimate of HIV among those HCV-infected is not an accurate estimate of HCV-HIV co-infection in Canada since it does not include co-infected men who have sex with men. Estimating HCV-HIV co-infection was beyond the scope of the current project.

Stage 2

HCV incidence rates derived from published studies and previous modelling studies were used to estimate the number of HCV-infected persons for each of the four groups defined above. With respect to immigrants, the prevalence in their country of origin stratified by gender and appropriate for the age at arrival was used to generate the number of prevalent HCV infections at time of arrival. Immigrant populations were also subjected to incident HCV infection related to injection drug use, blood transfusion and “Other” modes of transmission category but not risk associated with hemophilia.

Many hemophilia patients may not be admissible for immigration into Canada. More importantly, the number of non-Canadian-born hemophilia patients would likely be small due to prior higher mortality in countries where limited specialty care is available and to restricted admissibility for immigration. In any case, no data were available on hemophilia patients by region of birth. For these reasons and for the sake of simplicity, hemophilia patients were modelled within the Canadian-born population.

Data on HCV incidence rates from the Enhanced Hepatitis Strain Surveillance System (EHSSS)10, 11 adjusted for under-reporting and asymptomatic infection were used for initial values for overall HCV incidence. In a second approach, data from the limited available epidemiologic studies were also used to guide initial HCV incidence values. In this approach, incidence was applied only to susceptible persons; this is particularly important for IDUs. The numbers of prevalent HCV infections were subsequently compared to available data from special studies and previous modelling exercises to fit with observed HCV prevalence data12-43.

Stage 3

HCV-infected persons by stage of infection was progressed using annual transition parameters of a Markov model based on both observed data and modelling studies previously published. With respect to the natural history of HCV infection, the true transition probabilities and the important covariates are still somewhat incompletely characterized but there is the emergence of a consensus around the likely values of many of these parameters. Several studies have examined or reviewed the progression from HCV infection through serious sequelae ^44-49. A critical review of a large number of natural history studies published recently by Freeman⁵⁰ was of particular interest in this regard as was the modelling study by Salomon⁵¹, which used observed data to constrain the true values of these progression parameters. Finally, the report of Krahn and colleagues^{52, 53} was also extremely helpful since it included a systematic review of transition parameters and the important modifying factors such as age, sex and alcohol intake.

For the purposes of this study, the following stages of HCV-related morbidity were included: cirrhosis, decompensated cirrhosis (liver failure), hepatocellular carcinoma, liver transplant and liver-related death. The final estimates were compared to reported and modelled numbers of liver-related deaths and incidence of hepatocellular carcinoma published by investigators at Health Canada^54-56. See Sections 2.6 and 2.7 below for further details of the methodology used to model HCV sequelae.

The above modelling exercise did not include the impact of treatment mainly because the numbers of persons under treatment are poorly defined and would be low in relation to the reservoir of HCV infections in Canada.

2.2 Modelling HCV prevalence

To estimate HCV prevalence in Canada in 2007, the reference year, we considered separately each population stratum defined by exposure category and place of birth (Canada versus elsewhere). We reviewed studies carried out in Canada^12-43 including those done in prisons available as of September 2008^21-31. We used HCV prevalence and incidence in IDUs as observed in through the Canadian Street Youth Surveillance³², studies in Montreal³³, Vancouver^34,35, and Toronto³⁶ as well as from studies in other countries. We also reviewed the results from I-track³⁷ and SurvUDI, a sentinel surveillance project of IDUs recruited primarily at needle exchanges based in multiple centres in Quebec and in Ottawa^38,39.

The data for modelling HCV infection acquired through blood transfusion were based on work carried out in a previous study¹ and integrated into the model of HCV prevalence and incidence.

A model developed for hemophilia patients carried out by the contractor for the purpose of estimating HCV and HCV-HIV co-infection was adapted for the purposes of the present study. Data from two studies of hemophilia patients in the UK were also useful in this regard^{7, 8}.

The number of HCV transmissions in the “Other” category was developed on a proportional basis to fix the percentage of infections related to non-parenteral drug use, sexual transmission, and transmission in the health care setting based on data from the EHSSS as well as studies carried out in the United States.

2.3 HCV infection among persons not born in Canada

Extensive analyses were carried out to determine the contribution of HCV infections in Canada related to infections acquired before arrival in Canada among persons born elsewhere but who have immigrated to this country. An initial model was developed for each country in the world using data from the World Health Organization on country-specific HCV prevalence and applied to the population of each country. Adjustments were made as necessary such that the total world prevalence of HCV matched the 170 million HCV-infected persons estimated by the World Health Organization. Data from the 2001 census from Statistics Canada on persons living in Canada by country of birth were then applied to the HCV prevalence in the country of origin to determine the number of HCV-infected persons living in Canada and as well to make a preliminary characterization of the extent of HCV infection among immigrants from the most important countries of origin.

2.4 Use of data from the Enhanced Hepatitis Strain Surveillance System (EHSSS)

Data were kindly provided by the Public Health Agency of Canada on acute and chronic HCV infections collected at eight sentinel sites in Canada beginning in 1998. Custom outputs were kindly provided by the Public Health Agency of Canada on the distribution of cases by mutually exclusive risk factor by year, country of birth, acute versus chronic infection for each sentinel site. We performed a cross-tabulation of risk factors by country of birth. The purpose of this analysis was, for each sentinel site, to compare the EHSSS data to census data to determine the degree to which HCV prevalence may be different than among persons born outside Canada. We also wished to validate the initial results from the country-specific HCV prevalence analysis as noted in Section 2.3 above to identify the most important countries involved. Finally, we wished to test our initial hypothesis that the majority of HCV infection among persons born outside Canada was related to exposures other than injection drug use, the most important exposure for persons born in Canada.

2.5 HCV incidence

HCV incidence is unknown for most groups in Canada with the possible exception of IDUs in selected urban centers in Canada We used two approaches to obtain plausible estimates of the annual number of new HCV infections in Canada. The first approach consisted of examining the incidence of clinically apparent and reported cases of confirmed acute HCV infection through data collected by EHSSS. The “true” HCV incidence was derived by dividing this rate by the proportion of acute HCV infections that are icteric (and therefore presumably severe enough as to motivate the seeking of medical care and serologic diagnosis) reporting. The value for the proportion that would be symptomatic was 20% and the proportion of symptomatic infections that would be reported was 75% since most of such infections would be laboratory diagnosed and thus be reported.

We also used a second, independent approach to estimate HCV incidence. We estimated the proportion and number of active IDUs in Canada who would be susceptible to HCV infection (i.e. not already HCV-infected) and multiplied this number by the HCV incidence observed in epidemiologic studies among IDUs^{55, 56}. This number of new infections among IDUs was then divided by the estimated proportion of new HCV infections that are thought to occur among IDUs, based largely on the observations at the eight sites of the EHSSS from 1998-2002.

The methodology for estimating HCV incidence was based on collaborative work carried out with Dr. Shimian Zou in 2001, who worked at Health Canada at the time.

2.6 Modelling HCV outcomes

We carried out an extensive review of the medical literature to determine the annual rates of transition from initial HCV infection to the more advanced stages of HCV disease and its sequelae. We took into account the proportion of persons newly infected with HCV who remained viremic (i.e. had detectable HCV RNA) and, subsequently, the annual rate of HCV RNA loss and the rate of HCV antibody loss. Thus, the model incorporated three stages of HCV serologic status: RNA+ Ab+, RNA- Ab+, and RNA- Ab-. Following HCV infection, the natural history of hepatitis C was simulated using a Markov model through the following stages: infection (pre-cirrhotic), cirrhosis, decompensation, transplantation, hepatocellular carcinoma and death. The values of the annual transition probabilities were obtained from published studies and reports^44-53. The schematic model and transition probabilities used are shown in Figure 1.

Figure 1: Markov Model of Transition Through Stages of HCV Infection

Legend:

Inf = HCV infection
Ci = Cirrhosis
De = Decompensated liver failure
H = Hepatocellular carcinoma
Tr = Liver transplant
D = HCV-related death

2.7 The integrated analytic HCV model

The entire model was treated as a Markov model in an integrated continuum from entry through birth or immigration and then transition to exposure-related behaviours or experiences, mortality, HCV infection and progression to HCV disease.

All tables and rates were developed in Excel spreadsheets. Required parameters were consolidated in a consistent and usable format in two Excel “input” spreadsheets.

The model engine was written in the programming language APL+Win Version 4.0.03 supplied by APL2000 Inc. Data from the two parameters spreadsheets were copied into the APL workspace and saved. Any subsequent changes were recopied and saved. Adjustment and control parameters were developed and stored directly in the APL workspace.

The raw product of the model was a series of arrays of the sub-populations as defined above and the decrements of every type to which each population is subjected. For example, the shape of the array MaleResults is 63 (years 1960-2027) by 111 (ages 0-110) by 386 (number of population and decrement columns). Note that, except for death, each decrement column is the new-entrant column for a subsequent sub-population.

Subsidiary programs condensed and consolidated the raw output in various ways such as for 5-year periods, summing over age and combining some columns. Finally, selected condensed and consolidated data were exported directly by program to Excel spreadsheets. In the spreadsheets, simple further calculations such as summing and ratioing plus formatting were performed.

For immigrants, a “super-model” was developed which ran the basic model for each country/region and summed the results. It differed in basic model processes only as follows:

• Instead of births at age* zero each year, new immigrants entered the model at age of arrival
• A proportion of new immigrants were considered to be infected with HCV on arrival
• Hemophilia patients were all modelled within the Canadian-born population for the reasons stated in Section 1 above and because this simplified the analytic approach by modelling this group as one rather than two populations.

To account for competing mortality, HIV infection (including HIV incidence and HIV-related mortality) was incorporated into the model. This was carried out only for IDUs and hemophilia patients since the contribution of mortality among blood transfusion recipients (likely fewer than 100 persons also infected with HIV among an estimated 35,000 HCV-infected blood transfusion recipients) and other persons (likely fewer than 1,000 HIV-HCV coinfected persons among the estimated 74,000 HCV-infected persons +in the “Other” category). Among ex-IDUs, HIV incidence was considered to be zero since essentially all their HIV risk is related to active drug injection.

As indicated in Section 2.1 in the description of Stage 1, we also incorporated mortality due to other causes than HCV and HIV infection into the final model. For this purpose, we used life table values specific for each age and sex.

To model populations at risk and HCV incidence and prevalence, we adjusted model parameters so that model estimates were within 2% of the epidemiologically modelled estimates for these values.

2.8 Prevalence of HCV infection by province/territory

We used the number of reported HCV cases and populations of each Canadian province and territory that was reported to the Canadian Notifiable Surveillance System as of September 2008, the results of the 1998 provincial/territorial HCV model and the number of modelled HCV infection by exposure category in 2007 to impute the number and rate of HCV-infected persons in each province/territory by exposure category for 2007.

2.9 Prevalence of HCV infection in incarcerated populations

The number and rate of HCV infections among incarcerated persons was determined by reviewing the results of epidemiologic studies of HCV infection in this population. Rates of HCV infection were examined according to history of injection drug use for federal and provincial prisons independently. We concentrated on studies of penal institutions in Canada examining key studies and reviews available as of September 2008^21-31. We also reviewed special studies from Montreal³³ and Vancouver^{34, 35} and a situation report on blood-borne infections among IDUs in Ontario³⁶.

It is clear from the above and other studies that a prisoner's history of injection drug use is by far the most important factor for acquiring HCV infection. Drug injection while in prison may account for some HCV infections. This, however, is likely relatively rare and, in any case, such infections would be primarily among prisoners who injected drugs before being incarcerated and would therefore be reflected in the results of seroepidemiologic studies.

Based on our review, we estimated that approximately 30% of prisoners had injected drugs at some time in their life. We also concluded that, in Canada, HCV prevalence among prisoners in provincial institutions with an IDU history was 56% and in federal institutions 70%. Among prisoners without such a history, HCV prevalence was 0.10% and 1.0%, respectively.

2.10 Prevalence of HCV infection in Aboriginal populations

Few studies of HCV infection in Aboriginal populations have been carried out^40-43 and those that have usually focus on high risk non-representative populations. However, Minuk et al.⁴¹ recently published a review of viral hepatitis among Aboriginal populations which draws on both published and unpublished data and was extremely useful in beginning to quantify the risk of HCV infection in this population. One study⁴³ of a “street-involved” population in Winnipeg, of whom 63% were Aboriginal, found an HCV prevalence of 22.3% in Metis, 19.4% in First Nations, and 14.4% in non-Aboriginal subjects. Overall, HCV prevalence was 47.7% in IDUs but only 3.7% among non-IDUs.

We also used a recent analysis from the Ontario AIDS surveillance program on reported AIDS cases among Aboriginal IDUs⁵⁸ to help shed light on the relative risk of HCV in this population. This was done for each exposure category but special attention was given to AIDS cases among the men who have sex with other men and inject drugs (MSM-IDU) and IDU categories. Among MSM-IDUs, six of 219 AIDS cases to 2004 with known ethnicity (or 2.7%) were Aboriginal persons and among IDUs, 18 of 236 AIDS cases to 2004 with known ethnicity (7.6%) were Aboriginal persons. Thus, Aboriginals are over-represented among IDU-related AIDS cases since they constituted only about 1.5% of the Ontario population in 2004 (188,000 of 12,407,000). AIDS cases in the other exposure categories taken together among Aboriginal persons accounted for 0.7% of cases. Thus, the number of HIV infections among Aboriginal IDUs was five-fold greater than for the rest of the population. This is likely a reflection of higher rates of injection drug use in Aboriginal populations, though it is difficult to eliminate higher HIV prevalence in Aboriginal IDUs as a contributing factor. Both would lead to a higher HCV prevalence in this population.

Based on his review of both published and unpublished data, Minuk et al. concluded the prevalence of HCV among Aboriginal populations in Manitoba was approximately three-fold greater than in non-Aboriginal Manitobans. In contrast, Calzavara et al.²⁷ found HCV prevalence among inmates in an Ontario provincial prison to be similar among Aboriginal and non-Aboriginal participants.

Thus, it is difficult to come to any firm conclusion about HCV prevalence among Aboriginal persons based on the limited and conflicting data available. Nevertheless, we believe it is likely that HCV prevalence is higher among Aboriginal population and that a three-fold difference is plausible.

2.11 Proportion of HCV infections diagnosed and reported

One of the objectives of the present study was to estimate the number and proportion of HCV-infected persons living in Canada as of 2007 who had been diagnosed. To estimate this, data on HCV cases (including both acute and chronic HCV infections) reported from 1991 to 2007 and that reported to the Canadian Notifiable Disease Reporting System as of September 2008, were provided by the Public Health Agency of Canada. For jurisdictions in which HCV data from 1991 were not available via the Canadian Notifiable Disease Reporting System, the number of HCV diagnoses before case reporting data were available was estimated by using relative weights among the provinces derived from years for which data on reported cases were available. Cases reported to the Canadian Notifiable Disease Reporting System include chronic, acute and undifferentiated cases.

2.12 Number of persons with diagnosed HCV infection still living

Some persons diagnosed with HCV infection and reported from 1991 to 2007 died during this period, both related to their disease as well from other causes. This may represent a significant proportion of diagnosed cases during the 18-year period. Therefore, we used the annual mortality rates derived for the actuarial model over this period to estimate the number of deaths in each year subsequent to HCV diagnosis. Thus, we were able to estimate the number of diagnosed HCV cases alive as of December 2007 and, in turn, more accurately calculate the proportion of modelled HCV-infected persons alive as of 2007 who have been diagnosed.