### Meta-analyses

Fourteen of the 18 studies in the systematic review were included in the meta-analysis. Two studies^{[24,34]} were excluded, as they had insufficient statistical information to compute an effect size, as were two kin–cohort studies^{[35,36]} because the outcome was the frequency of a family history of colorectal cancer. All studies included in the meta-analysis had moderate quality (NOS = 4–6) based on the Newcastle-Ottawa Scale quality assessment (Table 1). The overall meta-analysis (Figure 3) of 18 assessments of colorectal cancer risk in *BRCA1* or *BRCA2* mutation carriers found a statistically significant increase in the odds of colorectal cancer for *BRCA* carriers using a random-effects model (OR = 1.24, 95% CI = 1.02 to 1.51, *P* = .03).

Figure 3.

Forest plot of the association between overall *BRCA* mutation carriers and colorectal cancer risk expressed as unadjusted odds ratio. Squares indicate study-specific risk estimates; horizontal lines indicate 95% confidence intervals (CI); diamond indicates summary estimate with the corresponding 95% CI. DerSimonian and Laird random-effects model was used and all statistical tests were two-sided. Test for heterogeneity: χ^{2} = 19.24, *df* = 18 (*P* = .38); *I* ^{2} = 8.5%. CRC = colorectal cancer.

Four subgroup meta-analyses that focused on only *BRCA1* were performed. The first subgroup meta-analysis included the 10 studies that focused on *BRCA1* (Figure 4A) and found that *BRCA1* mutation carriers were at higher risk of colorectal cancer (OR = 1.49, 95% CI = 1.19 to 1.85, *P* < .001). The second subgroup meta-analysis by *BRCA1*only included the seven cohort studies that reported age- and sex-adjusted estimates (Figure 5A), and yielded a statistically significant risk of colorectal cancer (OR = 1.56, 95% CI = 1.23 to 1.98, *P* < .001). The third subgroup meta-analysis included four studies that screened for colon cancer and focused on *BRCA1* (Supplementary Figure 1A, available online). There was a statistically significant difference in risk of colorectal cancer for *BRCA1* carriers (OR = 1.85, 95% CI = 1.39 to 2.47, *P* < .001). The fourth subgroup meta-analysis included four studies that ascertained the *BRCA* mutation and focused on *BRCA1* (Supplementary Figure 2A, available online). There was no statistically significant difference in the risk of colorectal cancer for *BRCA1* carriers (OR = 1.20, 95% CI = 0.85 to 1.71, *P* = .31).

Figure 4.

Forest plots of the association between *BRCA* mutation carriers and colorectal cancer risk expressed as unadjusted odds ratio: **A**) *BRCA1*; **B**) *BRCA2*. Squares indicate study-specific estimates; horizontal lines indicate 95% confidence intervals (CI); diamond indicates summary estimate with the corresponding 95% CI. DerSimonian and Laird random-effects model was used and all statistical tests were two-sided. Test for heterogeneity: **A**) χ^{2} = 8.02, *df* = 9 (*P* = .53); *I* ^{2} = 0.0%. **B**) χ^{2} = 1.33, *df* = 6 (*P* = .97); *I* ^{2} = 0.0%. CRC = colorectal cancer.

Figure 5.

Forest plots of the association between *BRCA* mutation carriers and colorectal cancer from studies reporting estimates adjusted for age and sex: **A**) *BRCA1*; **B**) *BRCA2*. Squares indicate study-specific risk estimates; horizontal lines indicate 95% confidence intervals (CI); diamond indicates summary estimate with the corresponding 95% CI. DerSimonian and Laird random-effects model was used and all statistical tests were two-sided. Test for heterogeneity: **A**) χ^{2} = 5.76, *df* = 7 (*P* = .58); *I* ^{2} = 0.0%. **B**) χ^{2} = 1.30, *df* = 5 (*P* = .94); *I* ^{2} = 0.0%. CRC = colorectal cancer.

Supplementary Figure 1.

Forest plots of the association between *BRCA* mutation carriers and colon cancer from studies screened for colon cancer A) *BRCA1.* B) *BRCA2.* Squares indicate study specific risk estimates; horizontal lines indicate 95% confidence intervals (CI); diamond indicates summary estimate with the corresponding 95% CI. DerSimonian and Laird random-effects model was used and all statistical tests were two-sided. Test for heterogeneity: A) Chi-squared= Chi-squared=0.64, df=3 (P=0.89); I-squared=0.0%. B) Chi-squared=0.01, df=1 (P=0.93); I-squared=0.0%.

Four subgroup meta-analyses that focused on *BRCA2* only were performed. The first subgroup meta-analysis of seven studies focused on *BRCA2* (Figure 4B) found no statistically significant risk of colorectal cancer for mutation carriers (OR = 1.10, 95% CI = 0.77 to 1.58, *P* = .61). The second subgroup meta-analysis by *BRCA2* only included the cohort studies that reported age- and sex-adjusted estimates*.* The analysis of six studies (Figure 5B) revealed no statistically significant risk of colorectal cancer between carriers and noncarriers (OR = 1.09, 95% CI = 0.75 to 1.58, *P* = .66). The third subgroup meta-analysis of two studies that screened for colon cancer and focused on *BRCA2* (Supplementary Figure 1B, available online) found no statistically significant risk of colorectal cancer for *BRCA2* carriers (OR = 1.22, 95% CI = 0.70 to 2.12, *P* = .49). Lastly, a subgroup meta-analysis of three studies that ascertained *BRCA* mutation and focused on *BRCA2* (Supplementary Figure 2B, available online) found no statistically significant risk of colorectal cancer for *BRCA2* carriers (OR = 1.02, 95% CI = 0.58 to 1.82, *P* = .94).

Supplementary Figure 2.

Forest plots of the association between *BRCA* mutation carriers and colorectal cancer from studies that ascertained mutation A) *BRCA1.* B) *BRCA2.* Squares indicate study specific risk estimates; horizontal lines indicate 95% confidence intervals (CI); diamond indicates summary estimate with the corresponding 95% CI. DerSimonian and Laird random-effects model was used and all statistical tests were two-sided. Test for heterogeneity: A) Chi-squared=3.00, df=5 (P=0.70); I-squared=0.0%. B) Chi-squared=0.94, df=3 (P=0.82); I-squared=0.0%.

The contour-enhanced funnel plot of standard error by effect estimate (Figure 6) showed, in general, symmetry. Coupled with the Egger test (*P* = .17) indicating symmetry in the funnel plot, there was no evidence of publication bias in the studies included in the meta-analysis.

Figure 6.

Contour-enhanced funnel plot of standard error by effect estimate for overall meta-analysis of the association between *BRCA* mutation and colorectal cancer.

J Natl Cancer Inst. 2018;110(11):1178-1189. © 2018 Oxford University Press

## Comments