Modifiable risk factors for ectopic pregnancy: a Mendelian randomization study

ObjectiveEctopic pregnancy is a condition where the fertilized ovum implants outside the main cavity of the uterus, and it is an important cause of pregnancy-related mortality.1Stulberg D.B. Cain L.R. Dahlquist I. Lauderdale D.S. Ectopic pregnancy rates and racial disparities in the Medicaid population, 2004-2008.Fertil Steril. 2014; 102: 1671-1676Abstract Full Text Full Text PDF PubMed Google Scholar Several modifiable risk factors are associated with ectopic pregnancy, in particular tobacco smoking, although the role of residual confounding remains unclear.2Gaskins A.J. Missmer S.A. Rich-Edwards J.W. Williams P.L. Souter I. Chavarro J.E. Demographic, lifestyle, and reproductive risk factors for ectopic pregnancy.Fertil Steril. 2018; 110: 1328-1337Abstract Full Text Full Text PDF PubMed Scopus (33) Google Scholar Because genetic variations are randomly assigned at conception, the presence or absence of risk-increasing alleles for a trait of interest is unaffected by disease status and lifestyle factors. Thus, one may use genetic variants as instruments in instrumental variable analyses—often called Mendelian randomization (MR) analyses—to greatly reduce the risk of reverse causation and confounding (Supplemental Figure 1).3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google ScholarThis study aimed to evaluate the causal association between 5 modifiable risk factors—smoking initiation, alcohol consumption, low-density lipoprotein (LDL) cholesterol, systolic blood pressure, and body mass index (BMI)—and risk of ectopic pregnancy using MR.Study DesignWe conducted a 2-sample MR study. The instruments for the 5 exposures were collected from genome-wide association studies (GWASs) of subjects of European ancestry (Supplemental Table 1). For each exposure, we used as instruments single-nucleotide polymorphisms strongly associated with the exposure (P<.001) and independent (R2<0.001 in 10 MB windows, European sample in the 1000 Genomes Project) from each other.As there was no published GWAS of ectopic pregnancy, we retrieved publicly available summary statistics from genome-wide association analyses from UK Biobank, FinnGen, and Michigan Genomics Initiative (Supplemental Table 2). Next, we performed a fixed-effect meta-analysis in METAL (version 2011-03-25; University of Michigan, Ann Arbor, MI) with standard errors as weights and used these results as the outcome in the MR analyses (3556 cases and 327,733 controls).MR analyses were performed using the TwoSampleMR package (version 0.5.6) in R (version 3.6.2). We used the inverse-variance–weighted analysis as the main analysis. For a genetic instrument to be valid, its effect on the outcome needs to solely go through the exposure, and thus, estimates can be biased by horizontal pleiotropy (a genetic variant has direct effects on other pathways other than exposure).3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google Scholar Moreover, estimates can be biased by weak genetic instruments.4Hemani G. Bowden J. Davey Smith G. Evaluating the potential role of pleiotropy in Mendelian randomization studies.Hum Mol Genet. 2018; 27: R195-R208Crossref PubMed Scopus (336) Google Scholar Bias because of pleiotropy was explored through 3 sensitivity analyses: weighted mode, weighted median, and MR Egger. F statistics were calculated for all instruments using the formula F ≈ (beta/standard error),2Gaskins A.J. Missmer S.A. Rich-Edwards J.W. Williams P.L. Souter I. Chavarro J.E. Demographic, lifestyle, and reproductive risk factors for ectopic pregnancy.Fertil Steril. 2018; 110: 1328-1337Abstract Full Text Full Text PDF PubMed Scopus (33) Google Scholar and an instrument with an F statistic of <10 was considered weak.3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google Scholar Finally, as sample overlap between the exposure and outcome analyses may bias toward the confounded estimate,3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google Scholar we performed a sensitivity analysis using only the coefficient for the association with ectopic pregnancy from FinnGen. To account for testing 5 exposures, we set the threshold for statistical significance to P=.01.Our study used publicly available data from sources with relevant ethical approvals.ResultsSmoking initiation (ever-smokers vs never-smokers) was significantly associated with the risk of ectopic pregnancy in the main analysis, with an odds ratio of 2.02 (95% confidence interval [CI], 1.22–3.36) per standard deviation increase in the prevalence of smoking initiation (Figure). Systolic blood pressure, LDL cholesterol, and BMI showed no clear effect. Alcohol intake showed some evidence of a positive association, but with wide CIs in part because of a smaller explained variance (0.6%) of the combined genetic instruments compared with the other exposures (2.0%–4.8%) (Supplemental Table 1). Our findings were robust in sensitivity analyses that accounted for pleiotropy (Figure) and sample overlap (Supplemental Figure 2). There was no weak instrument included in the analyses (Supplemental Table 1).ConclusionThe concordance between our MR study and previous observational studies strongly suggested a causal relationship between tobacco smoking and risk of ectopic pregnancy. The underlying mechanism for this is unclear but may be because of an impairment of oocyte or embryo transportation because of smoking.2Gaskins A.J. Missmer S.A. Rich-Edwards J.W. Williams P.L. Souter I. Chavarro J.E. Demographic, lifestyle, and reproductive risk factors for ectopic pregnancy.Fertil Steril. 2018; 110: 1328-1337Abstract Full Text Full Text PDF PubMed Scopus (33) Google Scholar ObjectiveEctopic pregnancy is a condition where the fertilized ovum implants outside the main cavity of the uterus, and it is an important cause of pregnancy-related mortality.1Stulberg D.B. Cain L.R. Dahlquist I. Lauderdale D.S. Ectopic pregnancy rates and racial disparities in the Medicaid population, 2004-2008.Fertil Steril. 2014; 102: 1671-1676Abstract Full Text Full Text PDF PubMed Google Scholar Several modifiable risk factors are associated with ectopic pregnancy, in particular tobacco smoking, although the role of residual confounding remains unclear.2Gaskins A.J. Missmer S.A. Rich-Edwards J.W. Williams P.L. Souter I. Chavarro J.E. Demographic, lifestyle, and reproductive risk factors for ectopic pregnancy.Fertil Steril. 2018; 110: 1328-1337Abstract Full Text Full Text PDF PubMed Scopus (33) Google Scholar Because genetic variations are randomly assigned at conception, the presence or absence of risk-increasing alleles for a trait of interest is unaffected by disease status and lifestyle factors. Thus, one may use genetic variants as instruments in instrumental variable analyses—often called Mendelian randomization (MR) analyses—to greatly reduce the risk of reverse causation and confounding (Supplemental Figure 1).3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google ScholarThis study aimed to evaluate the causal association between 5 modifiable risk factors—smoking initiation, alcohol consumption, low-density lipoprotein (LDL) cholesterol, systolic blood pressure, and body mass index (BMI)—and risk of ectopic pregnancy using MR. Ectopic pregnancy is a condition where the fertilized ovum implants outside the main cavity of the uterus, and it is an important cause of pregnancy-related mortality.1Stulberg D.B. Cain L.R. Dahlquist I. Lauderdale D.S. Ectopic pregnancy rates and racial disparities in the Medicaid population, 2004-2008.Fertil Steril. 2014; 102: 1671-1676Abstract Full Text Full Text PDF PubMed Google Scholar Several modifiable risk factors are associated with ectopic pregnancy, in particular tobacco smoking, although the role of residual confounding remains unclear.2Gaskins A.J. Missmer S.A. Rich-Edwards J.W. Williams P.L. Souter I. Chavarro J.E. Demographic, lifestyle, and reproductive risk factors for ectopic pregnancy.Fertil Steril. 2018; 110: 1328-1337Abstract Full Text Full Text PDF PubMed Scopus (33) Google Scholar Because genetic variations are randomly assigned at conception, the presence or absence of risk-increasing alleles for a trait of interest is unaffected by disease status and lifestyle factors. Thus, one may use genetic variants as instruments in instrumental variable analyses—often called Mendelian randomization (MR) analyses—to greatly reduce the risk of reverse causation and confounding (Supplemental Figure 1).3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google Scholar This study aimed to evaluate the causal association between 5 modifiable risk factors—smoking initiation, alcohol consumption, low-density lipoprotein (LDL) cholesterol, systolic blood pressure, and body mass index (BMI)—and risk of ectopic pregnancy using MR. Study DesignWe conducted a 2-sample MR study. The instruments for the 5 exposures were collected from genome-wide association studies (GWASs) of subjects of European ancestry (Supplemental Table 1). For each exposure, we used as instruments single-nucleotide polymorphisms strongly associated with the exposure (P<.001) and independent (R2<0.001 in 10 MB windows, European sample in the 1000 Genomes Project) from each other.As there was no published GWAS of ectopic pregnancy, we retrieved publicly available summary statistics from genome-wide association analyses from UK Biobank, FinnGen, and Michigan Genomics Initiative (Supplemental Table 2). Next, we performed a fixed-effect meta-analysis in METAL (version 2011-03-25; University of Michigan, Ann Arbor, MI) with standard errors as weights and used these results as the outcome in the MR analyses (3556 cases and 327,733 controls).MR analyses were performed using the TwoSampleMR package (version 0.5.6) in R (version 3.6.2). We used the inverse-variance–weighted analysis as the main analysis. For a genetic instrument to be valid, its effect on the outcome needs to solely go through the exposure, and thus, estimates can be biased by horizontal pleiotropy (a genetic variant has direct effects on other pathways other than exposure).3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google Scholar Moreover, estimates can be biased by weak genetic instruments.4Hemani G. Bowden J. Davey Smith G. Evaluating the potential role of pleiotropy in Mendelian randomization studies.Hum Mol Genet. 2018; 27: R195-R208Crossref PubMed Scopus (336) Google Scholar Bias because of pleiotropy was explored through 3 sensitivity analyses: weighted mode, weighted median, and MR Egger. F statistics were calculated for all instruments using the formula F ≈ (beta/standard error),2Gaskins A.J. Missmer S.A. Rich-Edwards J.W. Williams P.L. Souter I. Chavarro J.E. Demographic, lifestyle, and reproductive risk factors for ectopic pregnancy.Fertil Steril. 2018; 110: 1328-1337Abstract Full Text Full Text PDF PubMed Scopus (33) Google Scholar and an instrument with an F statistic of <10 was considered weak.3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google Scholar Finally, as sample overlap between the exposure and outcome analyses may bias toward the confounded estimate,3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google Scholar we performed a sensitivity analysis using only the coefficient for the association with ectopic pregnancy from FinnGen. To account for testing 5 exposures, we set the threshold for statistical significance to P=.01.Our study used publicly available data from sources with relevant ethical approvals. We conducted a 2-sample MR study. The instruments for the 5 exposures were collected from genome-wide association studies (GWASs) of subjects of European ancestry (Supplemental Table 1). For each exposure, we used as instruments single-nucleotide polymorphisms strongly associated with the exposure (P<.001) and independent (R2<0.001 in 10 MB windows, European sample in the 1000 Genomes Project) from each other. As there was no published GWAS of ectopic pregnancy, we retrieved publicly available summary statistics from genome-wide association analyses from UK Biobank, FinnGen, and Michigan Genomics Initiative (Supplemental Table 2). Next, we performed a fixed-effect meta-analysis in METAL (version 2011-03-25; University of Michigan, Ann Arbor, MI) with standard errors as weights and used these results as the outcome in the MR analyses (3556 cases and 327,733 controls). MR analyses were performed using the TwoSampleMR package (version 0.5.6) in R (version 3.6.2). We used the inverse-variance–weighted analysis as the main analysis. For a genetic instrument to be valid, its effect on the outcome needs to solely go through the exposure, and thus, estimates can be biased by horizontal pleiotropy (a genetic variant has direct effects on other pathways other than exposure).3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google Scholar Moreover, estimates can be biased by weak genetic instruments.4Hemani G. Bowden J. Davey Smith G. Evaluating the potential role of pleiotropy in Mendelian randomization studies.Hum Mol Genet. 2018; 27: R195-R208Crossref PubMed Scopus (336) Google Scholar Bias because of pleiotropy was explored through 3 sensitivity analyses: weighted mode, weighted median, and MR Egger. F statistics were calculated for all instruments using the formula F ≈ (beta/standard error),2Gaskins A.J. Missmer S.A. Rich-Edwards J.W. Williams P.L. Souter I. Chavarro J.E. Demographic, lifestyle, and reproductive risk factors for ectopic pregnancy.Fertil Steril. 2018; 110: 1328-1337Abstract Full Text Full Text PDF PubMed Scopus (33) Google Scholar and an instrument with an F statistic of <10 was considered weak.3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google Scholar Finally, as sample overlap between the exposure and outcome analyses may bias toward the confounded estimate,3Davies N.M. Holmes M.V. Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians.BMJ. 2018; 362: k601Crossref PubMed Scopus (676) Google Scholar we performed a sensitivity analysis using only the coefficient for the association with ectopic pregnancy from FinnGen. To account for testing 5 exposures, we set the threshold for statistical significance to P=.01. Our study used publicly available data from sources with relevant ethical approvals. ResultsSmoking initiation (ever-smokers vs never-smokers) was significantly associated with the risk of ectopic pregnancy in the main analysis, with an odds ratio of 2.02 (95% confidence interval [CI], 1.22–3.36) per standard deviation increase in the prevalence of smoking initiation (Figure). Systolic blood pressure, LDL cholesterol, and BMI showed no clear effect. Alcohol intake showed some evidence of a positive association, but with wide CIs in part because of a smaller explained variance (0.6%) of the combined genetic instruments compared with the other exposures (2.0%–4.8%) (Supplemental Table 1). Our findings were robust in sensitivity analyses that accounted for pleiotropy (Figure) and sample overlap (Supplemental Figure 2). There was no weak instrument included in the analyses (Supplemental Table 1). Smoking initiation (ever-smokers vs never-smokers) was significantly associated with the risk of ectopic pregnancy in the main analysis, with an odds ratio of 2.02 (95% confidence interval [CI], 1.22–3.36) per standard deviation increase in the prevalence of smoking initiation (Figure). Systolic blood pressure, LDL cholesterol, and BMI showed no clear effect. Alcohol intake showed some evidence of a positive association, but with wide CIs in part because of a smaller explained variance (0.6%) of the combined genetic instruments compared with the other exposures (2.0%–4.8%) (Supplemental Table 1). Our findings were robust in sensitivity analyses that accounted for pleiotropy (Figure) and sample overlap (Supplemental Figure 2). There was no weak instrument included in the analyses (Supplemental Table 1). ConclusionThe concordance between our MR study and previous observational studies strongly suggested a causal relationship between tobacco smoking and risk of ectopic pregnancy. The underlying mechanism for this is unclear but may be because of an impairment of oocyte or embryo transportation because of smoking.2Gaskins A.J. Missmer S.A. Rich-Edwards J.W. Williams P.L. Souter I. Chavarro J.E. Demographic, lifestyle, and reproductive risk factors for ectopic pregnancy.Fertil Steril. 2018; 110: 1328-1337Abstract Full Text Full Text PDF PubMed Scopus (33) Google Scholar The concordance between our MR study and previous observational studies strongly suggested a causal relationship between tobacco smoking and risk of ectopic pregnancy. The underlying mechanism for this is unclear but may be because of an impairment of oocyte or embryo transportation because of smoking.2Gaskins A.J. Missmer S.A. Rich-Edwards J.W. Williams P.L. Souter I. Chavarro J.E. Demographic, lifestyle, and reproductive risk factors for ectopic pregnancy.Fertil Steril. 2018; 110: 1328-1337Abstract Full Text Full Text PDF PubMed Scopus (33) Google Scholar