br For the province of
For the province of Quebec, the most recent year for which cancer incidence data were available was 2010. Quebec's 2015 cancer in-cidence was estimated in one of two ways. For cancers with fewer cases (< 500 in Canada in 2015), the last five years of available incidence data for Quebec, 2006–2010, were averaged and applied to Quebec's 2015 population. For other cancers, Quebec's 2015 incidence was im-puted by fitting a Poisson regression on Canada's 2008–2015 incidence.
2.10. Statistical analysis
To obtain the prevalence of a given infection in its associated cancer, individual studies were pooled with a random effects model; a fixed effect model was adopted if the index of consistency (I2) was < 25% and if the test for heterogeneity was not statistically significant (p > 0.05). To pool the proportions and measures of association, we used the commands metaprop (Nyaga et al., 2014) and metan (Harris et al., 2008), respectively. To calculate 95% confidence intervals (CIs) for the pooled proportions, the exact method was used with the com-mand “cimethod (exact)”. When studies were excluded by the software because of inadmissible 95% CIs (e.g. proportions of 1.0 can yield CIs over 1.0), the Freeman-Tukey double arcsine transformation was en-abled to calculate admissible 95% CIs bounded by 0.0–1.0 (stata com-mand is: “ftt”). All meta-analyses were conducted in Stata v14 (Stata-Corp., College Station, TX, USA). R was used to calculate PARs via formula 1 (R Foundation for Statistical Computing [Internet], 2017). For infections where the PAR was approximated by the prevalence of the infection in cancer cases, no additional calculations were necessary after pooling the prevalence. The CIs for PARs calculated via formula 1 were calculated as previously described (Brenner et al., 2018; Brenner et al., 2019).
3. Results
A summary of the overall methods and findings for HBV, HCV, H. pylori is presented in Table 2, and for infections where the prevalence in cases approximated the PAR in Table 3. Specific results and tabulations on the characteristics of included studies as well as forest plots, are provided under the respective infection and cancer sites (Supplemen-tary Tables 6–13 and Figs. 1–8).
Table 2
Infections for which the attributable risk was estimated using the prevalence in the Brequinar and measures of association.
Infection cancer (ICD-03 code)
Method of infection measurement
Source of prevalence data
Range of prevalence
Data used to estimate measure of association
Odds ratio (95%
estimates by sex
Helicobacter pylori
Serology with ELISA or
NHANES (1999–2000) data reweighted by Canada's sex, age,
Men:
Stomach, non-cardia
(C16.1–16.9)
immunoblot detection
and race/ethnicity distribution.
12.8% (aged < 50) to
studies with fixed effects: 3 corrected studies that used
Stomach, MALT lymphoma
Serology with ELISA detection
Estimates were corrected for sensitivity and specificity.
27.9% (aged ≥50)
ELISA and 3 studies that used immunoblot
6.3 (2.0–19.9)
Women:
One study of 33 cases matched to 134 controls (Parsonnet
Hepatitis B virus
Serology with HBsAg detection
CHMS HBsAg data (2007–2011) partitioned with NHANES
Men:
Hepatocellular carcinoma
(aged 70–79) to
Women:
Hepatitis C virus
Estimates from modeling studies
Chronic HCV prevalence modeled for the Canadian
Men:
Hepatocellular carcinoma
(aged 16–20) to
Non-Hodgkin lymphoma
2010)
sex distribution from another modeling study
1.9%
(aged 46–50)
Adjusted OR calculated from SEER Medicare data with
Women:
33,940 cases matched to controls on sex, age, and year of
Abbreviations: CI = confidence interval, MALT = mucosa-associated lymphoid tissue, NHANES = National Health and Nutrition Examination Survey, CHMS = Canadian Health Measures Survey, HBsAg = Hepatitis B surface antigen, SEER = Surveillance, Epidemiology, and End Results (United States).
Table 3
Methods used for the infections where population attributable risks were estimated using the prevalence of infection in cancer cases.
Infection cancer (ICD-03 code)
Method of infection
Source of
Cases used to
Sex/age group
PAR (prevalence of infection in
measurement
prevalence
estimate PAR, n
cancer cases)
estimatesd
Epstein-Barr virus
EBER ISH
1 study
EBER ISH and/or LMP1 IHC
4 studies
Hodgkin lymphoma (C81)
Nasopharynx (C11)
EBER ISH
2 studies
Human papillomavirus, high-risk types,a
anogenital tract cancers
PCR detection with
5 studies
Cervix (C53)
genotyping of at least HPV 16
Necessary cause
6 studies
2 studies
2 studies
Human papillomavirus, type 16, head and neck
3 studies
cancers
PCR with E6 and/or E7 for