Final Report: Estimating the Number of Persons co-infected with Hepatitis C Virus and Human Immunodeficiency Virus in Canada

2. Methods

As indicated above, the exposure categories included in the analysis were: MSM, MSM-IDU, IDU, persons from HIV-endemic countries, other persons infected by heterosexual contact, patients with hemophilia, and transfusion recipients. In the analysis, we realized early that estimates of hepatitis C prevalence for each group could not be based simply on data which did not stratify on HIV status. There is a strong correlation between HIV and HCV infection due to the fact that their modes of transmission are similar. This correlation is particularly high in the case of IDUs, transfusion recipients and hemophilia patients; the situation regarding MSM is less clear. Therefore, it is not surprising that almost all studies have shown a much higher HCV prevalence among HIV-positive persons than among HIV-negative persons. Thus, for available data on HCV prevalence not stratified on HIV infection, prevalence estimates had to be adjusted upwardly to account for this dependency.

2.1 Number of HIV-infected persons in Canada as of December 1999

Estimates of HIV prevalence to December 1999 were obtained as spreadsheet format from Dr. Chris Archibald, Chief, HIV/ AIDS Epidemiology, Centre for Infectious Disease Prevention and Control, Health Canada. These analyses had originally been carried out in 1997 and 1998 and were presented elsewhere <1,2>. The estimate for December 1999 was updated by Health Canada in 2000 and released in November 2000 <3>.

The 1999 estimate was carried out using the direct and indirect methods that were used for the 1996 estimates. In addition, the estimated HIV incidence and mortality in the period (1997-1999) were independently estimated and added and subtracted, respectively, from the 1996 estimate to ensure the plausibility of the updated estimate. The author of the present report collaborated in these analyses.

The national estimates of HIV prevalence for 1999 were provided to the author of the present report with aggregate values for several exposure categories: heterosexual and HIV-endemic categories were aggregated, as were those infected by clotting factors and blood transfusion (indicated under the category A Other @). With respect to region, estimates were aggregated for the prairie provinces and the territories. For the present analysis, we used less aggregated categories for both exposure category and region. To estimate these values for the number of HIV-infected persons, we interpolated the estimates provided by Health Canada using weights determined from an earlier, provisional estimate of national HIV infections carried out in the context of the present study.

To estimate the plausible limits of the number of HIV-infected persons by exposure category and region, we used Monte-Carlo simulation using Crystal Ball, Version 4.0c (Decisioneering Inc. Aurora, Colorado, USA). Monte-Carlo estimates were carried out using estimates of uncertainty such that the output ranges matched as closely as possible the Health Canada estimates.

2.2 Estimating HCV prevalence

2.2.1 Literature review

We reviewed published studies and available reports containing data on the prevalence of hepatitis C antibody among HIV-infected and uninfected persons for each of the above-noted exposure categories. A Medline search was carried out using the following key words: hepatitis C, HIV, homosexual, injection drug use, hemophilia, transfusion, aboriginal and incarceration. For this purpose, we used the National Library of Medicine Medline literature search engine (PubMed) accessible on the World Wide Web. When key articles were identified, we used two techniques to locate additional articles of particular interest. First, published articles were reviewed and the relevant references cited were also obtained. Second, we used the hyperlink Other related articles in PubMed to extend the search to ensure a systematic and comprehensive capture of relevant published studies. Several unpublished Canadian studies available as either technical reports or abstracts of presentations were kindly provided by Marcel DuBois of the Hepatitis C Division, Health Canada. Finally, data on HCV prevalence among HIV-infected MSM derived from a study carried out in Vancouver was obtained through a personal communication with Kevin Craib of the British Columbia Centre for Excellence in HIV/AIDS.

For each value of HCV prevalence identified through the literature review, we calculated the exact 95% binomial confidence limits through the use of Epi-Info Version 6.04 (Centers for Disease Control).

2.2.2 Consultation on HCV prevalence in Canada

As noted above, only very limited data are available on HCV prevalence among HIV-infected persons in Canada. The only data in Canada on co-infection was obtained from a study conducted by Kevin Craib in Vancouver of MSM and estimates derived by the present investigator for hemophilia patients in the context of a private consultation. To compensate for the lack of data, we chose to confer with key informants to establish reasonable point estimates and plausible ranges for the prevalence of HCV among HIV-infected persons by exposure category. The four experts selected for this exercise were Kevin Craib, British Columbia Centre for Excellence in HIV/AIDS; Dr. Morris Sherman, Toronto Hospital; Dr. Shimian Zou, Division of Bloodborne Pathogens, Health Canada; and Marcel DuBois, Hepatitis C Division, Health Canada.

These experts were provided with a copy of the report from Phase I, including the tables showing the estimates and confidence limits for a large number of studies that were reviewed in this phase of the study. They were also provided with copies of selected articles that had larger and more likely representative samples and were stratified on HIV.

The consultation was conducted via teleconference. The published studies on each exposure category were briefly presented and reviewed. The likely differences between the Canadian experience (for which there was limited data) and that of countries for which data was available was also considered. On the basis of these analyses, a point estimate was derived and agreed upon by consensus. The informants also agreed upon the plausible ranges around the point estimate based on the degree of convergence of this point estimate. In general, a 10% range around the point estimate was used where the point estimate was considered to be relatively precise, 20% where intermediate, and 30% in circumstances where it was relatively uncertain. A similar procedure was carried out for all seven exposure categories for Canada as a whole.

Following this estimation procedure, we determined for each exposure category using the same consensus process whether weighting was required to take into account possible differences in HCV prevalence between the seven regions of Canada, since A background @ HCV prevalence rates vary from one region to another. Based mainly on data from blood donors, HCV prevalence appears to be substantially higher in British Columbia, intermediate in Ontario, and relatively lower in Manitoba, Saskatchewan and the Atlantic provinces.

For the second stage of this analysis, the estimates for HCV prevalence, taking into account regional weightings as described above, were applied to the number of HIV-infected persons for each exposure category, then summed across exposure categories and regions (the Atlantic provinces were grouped for this analysis). To determine plausible ranges, we carried out Monte-Carlo simulation for each of the cells defined by exposure category and geographic region. The Monte-Carlo simulation was carried out using Crystal Ball, Version 4.0 (Decisioneering, Aurora, Colorado, USA); 10,000 iterations were performed.

The analysis to estimate the number of co-infected persons among Aboriginal and prisoner populations was somewhat different. For this purpose, we relied on the HCV prevalences determined for each HIV-defined exposure category as described above. Preliminary work on estimates for the Aboriginal and prison populations, begun in Phase I, was refined in the context of additional information obtained during Phase II. More details about the methods used are described in Sections 2.3 and 2.4 below.

2.2.3 Estimates of HCV prevalence among hemophilia patients and transfusion recipients

To determine the prevalence of HCV among HIV-infected hemophilia patients, we based our estimates on analyses carried out for the class action settlement 1986 to 1990 in Canada <4>.

Studies that could be used to obtain meaningful estimates of HCV prevalence among transfusion recipients in Canada were, however, not available. Therefore, we used the HCV per unit risk for the period 1980 to 1990 and applied it to the mean number of units received by surviving HIV-infected transfusion recipients (35) derived through an earlier modeling study <5> using the formula (1-[(1-p)n]) to calculate HCV prevalence among HIV-infected persons.

2.3 Estimates of HCV-HIV co-infection among the Aboriginal population

We obtained data on the Aboriginal population by province and by status as of 1996 from the Statistics Canada website. We then estimated the proportion of the population in each exposure category to determine the population at risk. Several available studies on the Aboriginal population suggest that injection may be more frequent among Aboriginal populations than among Canadians as a whole, and that HIV prevalence among Aboriginal injection drug users may be higher than for non-Aboriginal injection drug users. In this regard, several published studies from Vancouver were especially helpful <6-9 >; these observations were taken into account in the final estimates. This estimate was then multiplied by the HIV prevalence specific to each risk group.

The derived estimate of the number of HIV-infected Aboriginals was then reviewed in the light of data on recently reported AIDS cases published by the Division of HIV Epidemiology <10> and a personal communication from the same Division <11>. With respect to the former, we obtained also data on the distribution of diagnosed HIV infections by exposure category and sex and, from the latter, on the distribution of HIV infections by region.

To ensure our final number was a plausible estimate of the total number of HIV-infected Aboriginal persons, we applied the proportion of AIDS cases who were Aboriginal as a proportion of the total AIDS cases with known ethnicity to the national HIV estimate. Adjustments were made in the prevalence of risk factors and HIV prevalence to ensure that the estimate was consistent with both the distribution by exposure category and by region of reported AIDS cases. The final results for HIV distribution were also adjusted based on the proportion of reported AIDS cases diagnosed from 1996-99 comprising Aboriginals in each cell defined by exposure category and region.

Finally, the number of dually infected Aboriginal persons was determined by multiplying the HCV prevalence among HIV-infected persons defined by exposure category as determined above by the estimated number of HIV-infected Aboriginal persons in each exposure category.

The above method was used to obtain preliminary estimates of HIV infection and HCV-HIV infection among the Aboriginal population in an earlier phase of this study. More recently, the Bureau of HIV/AIDS, STD and TB estimated the number of HIV-infected Aboriginal persons in Canada as part of the national analyses referred to in Section 2.1 above. These were kindly provided to the author of the present report by exposure category and region of residence. The Health Canada estimates of the number of HIV-infected Aboriginal people was used in the final calculation, applying the HCV prevalence for the population as a whole. Where Health Canada had aggregated exposure categories and regions, the specific cells were estimated through interpolation as for the overall numbers described in Section 2.1.

2.4 Estimates of HCV-HIV co-infection among prisoners

This analysis was carried out in four stages:

2.4.1 - The numbers of persons incarcerated in 1996 in federal and provincial penal institutions were obtained from the Statistics Canada website. Where individual provincial estimates were not available, they were determined by interpolation using the federal prison data.

2.4.2 - To calculate the number of dually infected persons, we needed to know, most importantly, the proportion and the number of prisoners in each exposure category related to HIV infection.

For this purpose, a focused literature search was carried out to examine seroepidemiologic studies of HIV and HCV infection, as well as those on the prevalence of risk behaviours among prisoners in Canada and elsewhere <12-23>. The proportion of prisoners who were also injection drug users was estimated on this basis. For MSM, we used population-based estimates from the surveys carried out in Canada and elsewhere. The number of prisoners in each exposure category by region was derived by multiplying the estimated proportion of prisoners in each exposure category by the number of prisoners in each region.

2.4.3 - In the next stage of the analysis, we wished to determine HIV prevalence among each HIV exposure category in the prison population. In addition, we incorporated the estimates for the denominator populations, i.e the number of persons incarcerated in Canadian prisons, to calculate HIV prevalence as a proportion. This calculation was carried out in a preliminary fashion using initial, plausible estimates, and the numbers of HIV-infected persons thus derived were then compared to the results of the identified published studies and reports as noted in Section 2.4.2 used to determine the proportion of persons having injection drug use history. We then adjusted estimates of HIV prevalence so that the overall prevalence rate was within the span of observed HIV prevalence from the approximately 10 studies so identified. For the Yukon and Northwest Territories, we used estimates midway between those for Alberta and British Columbia.

2.4.4 - In the final stage of the analysis, the proportion of HIV-infected persons who are also HCV infected was derived by the same approach as that described above for each category for Canada as a whole. The estimated number of HIV-infected prisoners in each exposure category in each region was multiplied by the exposure category-specific HCV prevalence to derive the number of persons with HCV-HIV co-infection in each exposure category and geographic region. Using Monte-Carlo simulation as described above, we derived plausible limits around the point estimates for the number of dually infected prisoners.

Back