br The CHMS is a household based survey of
The CHMS is a household-based survey of non-institutionalized populations (Statistics Canada, 2010). Thus, groups with higher HCV prevalence, namely intravenous drug users, were underrepresented by excluding those who were homeless or in prison. Moreover, although a diagnosis of either HBV or HCV in Canada are reported to national public health agencies (Public Health Agency of Canada, 2018), many of these infections remain undiagnosed and therefore uncaptured in this data source. We thus obtained the modeled chronic HCV prevalence by birth cohort from Trubnikov, Yan and Archibald who accounted for high-risk groups and undiagnosed infections in their analyses (Trubnikov et al., 2014). To obtain chronic HCV prevalence by sex, we partitioned the estimates using the sex distribution of HCV prevalence from another study that modeled HCV prevalence in Canada in 2007 (Remis, 2010). Since the latency period between initial HCV infection and hepatocellular carcinoma is 25–30 years (Lingala and Ghany, 2015), we used the midpoint of a 15-year latency in our estimates.
We did not estimate a PAR for HBV and HCV coinfection and he-patocellular carcinoma because data on coinfection prevalence were not available. To estimate the combined impact of HBV and HCV on hepatocellular carcinoma, we combined their PARs with the following equation: 1 – (1-HBV PAR) * (1-HCV PAR) (Miettinen, 1974).
2.7. Helicobacter pylori
Few studies have assessed H. pylori prevalence in Canadian popu-lations. Although most of these studies were conducted with specialized populations (Cheung et al., 2014; Sethi et al., 2013), one study included 1306 residents aged 50–80 in Canada's most populous province, On-tario (Naja et al., 2007). As population-based data covering a broad age range were required, we opted to utilize other data. H. pylori serostatus was assessed in one NHANES RGDS peptide collected from 1999 to 2000 (Centers for Disease Control and Prevention, 2001) which resulted in a 15–16 year latency period. The weighted NHANES data were re-weighted by sex, five-year age groups, and race/ethnicity (Black, Latin American, White, and Other) to better reflect the composition of the Canadian population in 2001 (the closest year for which Canadian census ethnicity data were available). Missing H. pylori results, ac-counting for 5.0–6.6% of the reweighted data, were assumed to be missing completely at random and excluded. Additionally, half of the 1–2% ‘equivocal’ results, which were the results of IgG levels between the cut-offs for positive and negative results, were re-assigned as posi-tive or negative. NHANES used enzyme-linked immunosorbent assay (ELISA) to detect H. pylori. ELISA has a sensitivity of 95.6% and spe-cificity of 92.6% (Monteiro et al., 2001). We corrected our reweighted prevalence data according to these reported diagnostic accuracy mea-sures (Franco, 1992).
Since immunoblot is more sensitive than ELISA for the detection of H. pylori in gastric cancer cases (Gonzalez et al., 2012; Mitchell et al., 2008), we also corrected the association measures from matched case-control studies that used ELISA by deriving a formula used to adjust the OR, (Franco, 1992) and calculating sensitivity and specificity para-meters. The latter were derived by pooling the sensitivity and specifi-city from three studies (Gonzalez et al., 2012; Mitchell et al., 2008; Peleteiro et al., 2010), that directly compared ELISA and immunoblot in the same patients.
2.8. Estimating infection prevalence in cancer cases
considered attributable to HPV if genotype 16 was found via the de-tection of E6 and/or E7 oncoproteins which indicates viral activity and replication.
2.9. Cancer incidence
To determine the number of cases attributable to a given infection, the calculated PAR is multiplied by the number of incident cases. Incident cancer data were obtained from the Canadian Cancer Registry for 2015, which was the most recent year available. When data were requested for rare or subsite cancers, they were aggregated to maintain privacy; for example, cancer incidence counts were combined into two age groups (ages < 50, and ≥ 50), instead of five-year age groups. To preserve the granularity in the incidence data, we estimated the pro-portion of liver cancer estimated to be hepatocellular carcinoma. A study using SEER (Surveillance, Epidemiology, and End Results) data reported that there were 55,344 primary liver cancers diagnosed from 1978 to 2007, of which 44,080 were hepatocellular carcinoma (Altekruse et al., 2011). We applied the ensuing proportion of 0.797 (44,080/55,344) to liver cancer incidence to get the estimated number of hepatocellular carcinoma cases.