(C) PLOS One [1]. This unaltered content originally appeared in journals.plosone.org.
Licensed under Creative Commons Attribution (CC BY) license.
url:https://journals.plos.org/plosone/s/licenses-and-copyright

------------



Genetically proxied therapeutic inhibition of antihypertensive drug targets and risk of common cancers: A mendelian randomization analysis

['James Yarmolinsky', 'Mrc Integrative Epidemiology Unit', 'University Of Bristol', 'Bristol', 'United Kingdom', 'Population Health Sciences', 'Bristol Medical School', 'Virginia Díez-Obrero', 'Biomarkers', 'Susceptibility Unit']

Date: 2022-02 

We performed a mendelian randomization analysis to examine the association between genetically proxied inhibition of 3 antihypertensive drug targets and risk of 4 common cancers (breast, colorectal, lung, and prostate). Single-nucleotide polymorphisms (SNPs) in ACE, ADRB1, and SLC12A3 associated (P < 5.0 × 10 −8 ) with systolic blood pressure (SBP) in genome-wide association studies (GWAS) were used to proxy inhibition of angiotensin-converting enzyme (ACE), β-1 adrenergic receptor (ADRB1), and sodium-chloride symporter (NCC), respectively. Summary genetic association estimates for these SNPs were obtained from GWAS consortia for the following cancers: breast (122,977 cases, 105,974 controls), colorectal (58,221 cases, 67,694 controls), lung (29,266 cases, 56,450 controls), and prostate (79,148 cases, 61,106 controls). Replication analyses were performed in the FinnGen consortium (1,573 colorectal cancer cases, 120,006 controls). Cancer GWAS and FinnGen consortia data were restricted to individuals of European ancestry. Inverse-variance weighted random-effects models were used to examine associations between genetically proxied inhibition of these drug targets and risk of cancer. Multivariable mendelian randomization and colocalization analyses were employed to examine robustness of findings to violations of mendelian randomization assumptions. Genetically proxied ACE inhibition equivalent to a 1-mm Hg reduction in SBP was associated with increased odds of colorectal cancer (odds ratio (OR) 1.13, 95% CI 1.06 to 1.22; P = 3.6 × 10 −4 ). This finding was replicated in the FinnGen consortium (OR 1.40, 95% CI 1.02 to 1.92; P = 0.035). There was little evidence of association of genetically proxied ACE inhibition with risk of breast cancer (OR 0.98, 95% CI 0.94 to 1.02, P = 0.35), lung cancer (OR 1.01, 95% CI 0.92 to 1.10; P = 0.93), or prostate cancer (OR 1.06, 95% CI 0.99 to 1.13; P = 0.08). Genetically proxied inhibition of ADRB1 and NCC were not associated with risk of these cancers. The primary limitations of this analysis include the modest statistical power for analyses of drug targets in relation to some less common histological subtypes of cancers examined and the restriction of the majority of analyses to participants of European ancestry.

Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: WZ and GDS are on the Editorial Board of PLOS Medicine. GP reported receiving funding from Sanofi (research funding); RKP reported receiving funding from the NIH U01 CA167551-09 (grant paid to institution) and Alimentiv, Inc, Allergan, AbbVie, PathAI, Eli Lilly (consulting fees); CIA reported receiving funding from the National Institutes of Health (U19CA203654) and the Cancer Prevention Research Institute of Texas (RR170048); and VM reported receiving funding from the Agency for Management of University and Research Grants (AGAUR) of the Catalan Government (Grant 2017SGR723), the Instituto de Salud Carlos III, co-funded by FEDER funds (Grant PI17-00092), and the National Institutes of Health (Grant R01CA201407). TGR s employed part-time by Novo Nordisk outside of the work presented in this manuscript. HH is on the Scientific Advisory Board for Invitae, Promega, and Genome Medical and has stock in Genome Medical and stock options in GI OnDemand. All remaining authors declare no potential competing interests.

Funding: JY is supported by a Cancer Research UK Population Research Postdoctoral Fellowship (C68933/A28534) ( https://www.cancerresearchuk.org/ ). JY, EEV, VT, GDS, and RMM are supported by Cancer Research UK (C18281/A29019) programme grant (the Integrative Cancer Epidemiology Programme) ( https://www.cancerresearchuk.org/ ). JY, TGR, VMW, EEV, VT, GDS, and RMM are part of the Medical Research Council Integrative Epidemiology Unit at the University of Bristol which is supported by the Medical Research Council (MC_UU_00011/1, MC_UU_00011/3, MC_UU_00011/6, and MC_UU_00011/4) and the University of Bristol ( https://mrc.ukri.org/ ; https://www.bristol.ac.uk/ ). RMM is also supported by the NIHR Bristol Biomedical Research Centre which is funded by the NIHR and is a partnership between University Hospitals Bristol NHS Foundation Trust and the University of Bristol ( https://www.nihr.ac.uk/ ). EEV is supported by Diabetes UK (17/0005587) and the World Cancer Research Fund (WCRF UK), as part of the World Cancer Research Fund International grant programme (IIG_2019_2009) ( https://www.diabetes.org.uk/ ; https://www.wcrf.org/ ). MOS received a post-doctoral fellowship from the Spanish Association Against Cancer (AECC) Scientific Foundation ( https://www.aecc.es/es ). VDO, MOS, and VM are part of the group 55 of CIBERESP and AGAUR 2017SGR723 ( https://www.ciberesp.es/en ). CIA is a Cancer Prevention Research Institute of Texas Research Scholar and is supported by RR170048 ( https://www.cprit.state.tx.us/ ). INTEGRAL-ILCCO is supported by a National Institutes of Health grant (U19 CA203654) ( https://www.nih.gov/ ). The International Lung Cancer Consortium is supported by a National Cancer Institute grant (U19CA203654) ( https://www.cancer.gov/ ). GC is supported by R01 NIH/NCI CA143237 and CA204279. Cancer consortia funding is presented in the Supplementary Material. The funders of this study played no role in the design, data collection and analysis, decision to publish, or preparation of this manuscript.

Data Availability: Approval was received to use restricted summary genetic association data from GECCO, INTEGRAL ILCCO, and PRACTICAL consortia after submitting a proposal to access this data. Summary genetic association data from these consortia can be accessed by contacting GECCO ( kafdem@fredhutch.org , INTEGRAL ILCCO ( rayjean.hung@lunenfeld.ca ) ( https://ilcco.iarc.fr/ ), and PRACTICAL ( practical@icr.ac.uk ). Summary genetic association data from the BarcUVa-Seq analyses included in this manuscript can be downloaded from https://doi.org/10.34810/data132 (“eqtls_annot_hg38.txt.xz”). Summary genetic association data from the Finngen consortium can be accessed by visiting https://www.finngen.fi/en/access_results . To obtain data from the Asia Colorectal Cancer Consortium and the Korean National Cancer Center CRC Study, please contact the Vanderbilt Epidemiology Center ( epidemiology@vumc.org ). All other relevant data are within the manuscript and its Supporting Information files.

Naturally occurring variation in genes encoding antihypertensive drug targets can be used as proxies for these targets to examine the effect of their therapeutic inhibition on disease outcomes (“mendelian randomization”) [ 31 , 32 ]. Such an approach should be less prone to conventional issues of confounding as germline genetic variants are randomly assorted at meiosis. In addition, mendelian randomization analysis permits the effect of long-term modulation of drug targets on cancer risk to be examined. Drug-target mendelian randomization can therefore be used to mimic the effect of pharmacologically modulating a drug target in clinical trials and has been used previously to anticipate clinical benefits and adverse effects of therapeutic interventions [ 33 – 36 ].

Some observational epidemiological studies have suggested potential adverse effects of long-term use of these drugs on risk of common cancers (i.e., breast, colorectal, lung, and prostate) [ 7 – 10 ], though findings have been largely inconsistent (i.e., null and protective associations have also been reported for the relationship between ACE inhibitor use and cancer risk) [ 11 – 15 ]. Interpretation of the epidemiological literature is challenging for several reasons. First, pharmaco-epidemiological studies are susceptible to residual confounding due to unmeasured or imprecisely measured confounders, including those related to indication [ 16 ]. Second, several studies examining ACE inhibitor use and cancer risk have included prevalent drug users, which can introduce bias because prevalent users are “survivors” of the early period of pharmacotherapy and because covariates at study entry can be influenced by prior medication use [ 12 , 17 – 20 ]. Third, some prior studies may have suffered from time-related biases, including immortal time bias, which can arise because of misalignment of the start of follow-up, eligibility, and treatment assignment of participants [ 17 , 18 , 20 , 21 ]. These biases can produce illusory results in favor of the treatment group, while other biases often pervasive in the pharmaco-epidemiological literature (e.g., detection bias due to more intensive clinical monitoring and testing of individuals receiving treatment) can alternatively generate upward-biased effect estimates among those receiving treatment.

Angiotensin-converting enzyme (ACE) inhibitors are commonly prescribed antihypertensive medications [ 1 ]. These medications lower blood pressure by inhibiting the conversion of angiotensin I to angiotensin II, a vasoconstrictor and the primary effector molecule of the renin–angiotensin system (RAS). Though clinical trials have supported the relative safety of these medications in the short term (median follow-up of 3.5 years), concerns have been raised that long-term use of these medications could increase risk of cancer [ 2 , 3 ]. These safety concerns relate to the multifaceted role of ACE, which cleaves various other substrates beyond angiotensin I, including several peptides that have proliferative effects. For example, ACE inhibition leads to the accumulation of bradykinin, an inflammatory mediator involved in tumor growth and metastasis [ 4 ]. In addition, substance P is elevated in ACE inhibitor users, which can promote tumor proliferation, migration, and angiogenesis [ 5 , 6 ].

Bioinformatic follow-up of findings from transcriptome-wide analysis was performed to further interrogate downstream perturbations of the ACE wGRS on gene expression profiles using gene set enrichment analysis and coexpression network analysis. In brief, these methods can either evaluate whether expression levels of genes associated with the ACE wGRS are enriched in relation to an a priori defined set of genes based on curated functional annotation (gene set enrichment analysis) or permit the identification of clusters of genes (termed “modules” and assigned arbitrary color codes), which show a coordinated expression pattern associated with the wGRS (coexpression network analysis). Further information on gene set enrichment and coexpression network analysis is presented in S4 Methods .

To explore potential mechanisms governing associations and to further evaluate potential violations of mendelian randomization assumptions, we examined associations between a genetic risk score for serum ACE concentrations and gene expression profiles in normal (i.e., nonneoplastic) colon tissue samples. Gene expression analysis was performed using data from the University of Barcelona and the University of Virginia Genotyping and RNA Sequencing Project (BarcUVa-Seq) [ 66 ]. This analysis was restricted to 445 individuals (mean age 60 years, 64% female, 95% of European ancestry) who participated in a Spanish colorectal cancer risk screening program that obtained a normal colonoscopy result (i.e., macroscopically normal colon tissue, with no malignant lesions). Further information on RNA-Seq data processing and quality control is presented in S3 Methods .

For analyses showing evidence of colocalization across drug target and cancer endpoint signals, we then examined whether there was evidence of an association of genetically proxied inhibition of that target with previously reported risk factors for the relevant cancer endpoint (i.e., BMI, low-density lipoprotein cholesterol, total cholesterol, iron, insulin-like growth factor 1, alcohol intake, standing height, and physical activity for colorectal cancer risk) [ 57 – 65 ]. If there was evidence for an association between a genetically proxied drug target and previously reported risk factor (P < 0.05), this could reflect vertical pleiotropy (i.e. “mediated pleiotropy” where an instrument has an effect on 2 or more traits that influence an outcome via the same biological pathway) or horizontal pleiotropy. In the presence of an association with a previously reported risk factor, multivariable mendelian randomization can then be used to examine the association of drug target inhibition in relation to cancer risk, accounting for this risk factor [ 45 ].

We evaluated the “exclusion restriction” assumption by performing various sensitivity analyses. First, we performed colocalization to examine whether drug targets and cancer endpoints showing nominal evidence of an association in MR analyses (P < 0.05) share the same causal variant at a given locus. Such an analysis can permit exploration of whether drug targets and cancer outcomes are influenced by distinct causal variants that are in linkage disequilibrium with each other, indicative of horizontal pleiotropy (an instrument influencing an outcome through pathways independent to that of the exposure), a violation of the exclusion restriction criterion [ 56 ]. Colocalization analysis was performed by generating ±300 kb windows from the top SNP used to proxy each respective drug target. As a convention, a posterior probability of ≥0.80 was used to indicate support for a configuration tested. An extended description of colocalization analysis including assumptions of this method is presented in S1 Methods .

We tested the “relevance” assumption by generating estimates of the proportion of variance of each drug target explained by the instrument (r 2 ) and F-statistics. F-statistics can be used to examine whether results are likely to be influenced by weak instrument bias, i.e., reduced statistical power when an instrument explains a limited proportion of the variance in a drug target. As a convention, an F-statistic of at least 10 is indicative of minimal weak instrument bias [ 55 ].

To validate the serum ACE concentrations instrument, we examined the association between genetically proxied ACE inhibition and (i) SBP; (ii) risk of stroke in the MEGASTROKE consortium (40,585 cases; 406,111 controls of European ancestry); (iii) risk of coronary artery disease in the CARDIoGRAMplusC4D consortium (60,801 cases; 123,504 controls, 77% of whom were of European ancestry); and (iv) risk of type 2 diabetes in the DIAGRAM consortium (N = 74,124 cases; 824,006 controls of European ancestry) and compared the direction of effect estimates obtained with those reported for ACE inhibitor use in meta-analyses of randomized controlled trials [ 47 – 49 ]. Likewise, we validated ADRB1 and NCC instruments by examining the association between inhibition of these targets and risk of stroke and coronary artery disease, as reported in meta-analyses of clinical trials [ 49 ].

We explored construction of genetic instruments to proxy ACE inhibition using 2 approaches: (i) by obtaining genome-wide significant variants in weak linkage disequilibrium (r 2 < 0.10) in or within ±100 kb from ACE that were associated with SBP in previously described pooled GWAS analyses (resulting in 2 SNPs); and (ii) by obtaining genome-wide significant variants in weak linkage disequilibrium (r 2 < 0.10) in or within ±100 kb from ACE that were associated with serum ACE concentrations in a GWAS of 4,174 participants in the Outcome Reduction with Initial Glargine INtervention (ORIGIN) study (resulting in 14 SNPs) [ 46 ]. Approximately 46.6% of participants in the ORIGIN study were of European ancestry, and 53.4% were of Latin American ancestry. Effect allele frequencies for these 14 SNPs were broadly similar across both ancestries ( S2 Table ). We then compared the proportion of variance in either SBP or serum ACE concentrations explained (r 2 ) across each respective instrument to prioritize the primary instrument to proxy ACE inhibition. The “serum ACE concentrations instrument” (r 2 = 0.34 to 0.39, F = 2,156.5 to 2,594.9) was prioritized as our primary instrument to examine genetically proxied ACE inhibition because of stronger instrument strength as compared to the “SBP instrument” (r 2 = 0.02, F = 128.5). In sensitivity analyses, we also examined the association between genetically proxied ACE inhibition and cancer endpoints using the “SBP instrument.”

To proxy ADRB1 inhibition, 8 single-nucleotide polymorphisms (SNPs) associated with SBP at genome-wide significance and within ±100 kb windows from ADRB1 were obtained. To proxy NCC inhibition, 1 SNP associated with SBP at genome-wide significance and within a ±100-kb window from SLC12A3 (alias for NCC) was obtained. For both of these drug targets, SNPs used as proxies were permitted to be in weak linkage disequilibrium (r 2 < 0.10) with each other to increase the proportion of variance in each respective drug target explained by the instrument, maximizing instrument strength.

To generate instruments to proxy ACE, ADRB1, and NCC inhibition, we pooled summary genetic association data from 2 previously published GWAS of systolic blood pressure (SBP) using inverse-variance weighted fixed-effects models in METAL [ 42 ]. The first GWAS was a meta-analysis of ≤757,601 individuals of European descent in the UK Biobank and International Consortium of Blood Pressure-Genome Wide Association Studies (ICBP) [ 43 ]. The second GWAS was performed in 99,785 individuals in the Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort, of whom the majority (81.0%) were of European ancestry [ 44 ]. Both GWAS were adjusted for age, sex, body mass index (BMI), and antihypertensive medication use. Estimates that were genome-wide significant (P < 5.0 × 10 −8 ) in pooled analyses (N ≤ 857,386) and that showed concordant direction of effect across both GWAS were then used to generate instruments.

For primary analyses, summary genetic association data were obtained from 4 cancer genome-wide association study (GWAS) consortia. Summary genetic association estimates for overall and estrogen receptor (ER)–stratified breast cancer risk in up to 122,977 cases and 105,974 controls were obtained from the Breast Cancer Association Consortium (BCAC) [ 37 ]. Summary genetic association estimates for overall and site-specific colorectal cancer risk in up to 58,221 cases and 67,694 controls were obtained from an analysis of the Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO), ColoRectal Transdisciplinary Study (CORECT), and Colon Cancer Family Registry (CCFR) [ 38 ]. Summary genetic association estimates for overall and histological subtype-stratified lung cancer risk in up to 29,266 cases and 56,450 controls were obtained from an analysis of the Integrative Analysis of Lung Cancer Risk and Etiology (INTEGRAL) team of the International Lung Cancer Consortium (ILCCO) [ 39 ]. Summary genetic association estimates for overall and advanced prostate cancer risk in up to 79,148 cases and 61,106 controls were obtained from the Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome (PRACTICAL) consortium [ 40 ]. These analyses were restricted to participants of European ancestry.

In transcriptome-wide analyses, the serum ACE wGRS was most strongly associated with ACE expression levels in the colon (trimmed mean of M-values [TMMs] per SD increase in wGRS: −0.42, 95% CI −0.49 to −0.36; P = 2.29 × 10 −31 ). Genetically proxied ACE expression in the colon was associated with increased odds of colorectal cancer (OR per SD increase in expression: 1.02, 95% CI 1.00 to 1.04; P = 0.01). However, colocalization analysis suggested that colon ACE expression and colorectal cancer risk were unlikely to share a causal variant within the ACE locus (29.1% posterior probability of a shared causal variant) ( S12 Table , Figs 7 and 8 ). The serum ACE wGRS was also associated with expression levels of CYB561 (TMMs per SD increase in wGRS: −0.17, 95% CI −0.21 to −0.12; P = 8.28 × 10 −11 ) and FTSJ3 (TMMs per SD increase in wGRS: −0.19, 95% CI −0.24 to −0.13; P = 2.95 × 10 −10 ) in the colon after correction for multiple testing. ACE, CYB561, and FTSJ3 are neighboring genes on chromosome 17, suggesting that associations between the ACE wGRS and CYB561 and FTSJ3 could be driven through their coexpression. Genetically proxied CYB561 expression in the colon was associated with increased odds of colorectal cancer (OR per SD increase in expression: 1.06, 95% CI 1.02 to 1.10; P = 0.005). However, multivariable mendelian randomization analysis examining the association of genetically proxied ACE inhibition with colorectal cancer risk adjusting for CYB561 expression in the colon was consistent with univariable analyses (OR 1.13, 95% CI: 0.96 to 1.32; P = 0.14). Genetically proxied FTSJ3 expression in the colon was not associated with odds of colorectal cancer (OR per SD increase in expression: 1.00, 95% CI 0.98 to 1.03; P = 0.77).

There was little evidence that genetically proxied NCC inhibition was associated with overall risk of breast, colorectal, lung, or prostate cancer ( Table 4 ). In ER–stratified breast cancer analyses, there was weak evidence that NCC inhibition was associated with an increased risk of ER− breast cancer (OR equivalent to 1 mm Hg lower SBP: 1.20, 95% CI 1.02 to 1.40; P = 0.03). Colocalization analysis provided little support for NCC and ER− breast cancer association sharing a causal variant within the SLC12A3 locus (11.5% posterior probability of a shared causal variant) ( S11 Table , Figs 5 and 6 ).

There was little evidence that genetically proxied ADRB1 inhibition was associated with overall risk of breast, colorectal, lung, or prostate cancer ( Table 3 ). In lung cancer subtype-stratified analyses, there was weak evidence to suggest an association of genetically proxied ADRB1 inhibition with lower risk of small cell lung carcinoma (OR equivalent to 1 mm Hg lower SBP: 0.87, 95% CI 0.79 to 0.96; P = 0.008). Colocalization analysis suggested that ADRB1 and small cell lung carcinoma were unlikely to share a causal variant within the ADRB1 locus (1.5% posterior probability of a shared causal variant) ( S7 Table , Figs 3 and 4 ). Findings for overall and subtype-specific cancer risk did not differ markedly when using an instrument for ADRB1 inhibition constructed from a GWAS unadjusted for BMI ( S8 Table ). In sensitivity analyses restricting the ADRB1 instrument to SNPs that are eQTLs for ADRB1, findings were consistent with those obtained when using the primary instrument for this target ( S10 Table ).

Genetically proxied ACE inhibition was associated with an increased odds of colorectal cancer (OR equivalent to 1 mm Hg lower SBP: 1.13, 95% CI 1.06 to 1.22; P = 3.6 × 10 −4 ). Likewise, in analyses using SBP SNPs in ACE, genetically proxied SBP lowering via ACE inhibition was associated with an increased odds of colorectal cancer (OR equivalent to 1 mm Hg lower SBP: 1.11, 95% CI 1.04 to 1.18; P = 1.3 × 10 −3 ). When scaled to represent SBP lowering achieved in clinical trials of ACE inhibitors for primary hypertension (equivalent to 8 mm Hg lower SBP), this represents an OR of 2.74 (95% CI 1.58 to 4.76) [ 69 ]. In site-specific analyses, this association was stronger for colon cancer risk (OR 1.18, 95% CI 1.07 to 1.31; P = 9.7 × 10 −4 ) than rectal cancer risk (OR 1.07, 95% CI 0.97 to 1.18; P = 0.16). Similar associations were found across risk of proximal colon cancer (OR 1.23, 95% CI 1.10 to 1.37; P = 1.9 × 10 −4 ) and distal colon cancer (OR 1.15, 95% CI 1.03 to 1.27; P = 0.01).

Findings from genetic instrument validation analyses for drug targets were broadly concordant (i.e., in direction of effect) with findings from meta-analyses of randomized trials for these medications. Genetically proxied ACE inhibition was associated with lower SBP (mm Hg per SD lower serum ACE concentration: −0.40, 95% CI −0.21 to −0.59, P = 4.2 × 10 −5 ) and a lower risk of type 2 diabetes (odds ratio (OR) equivalent to 1 mm Hg lower SBP: 0.90, 95% CI 0.85 to 0.95, P = 1.3 × 10 −4 ). There was weak evidence for an association of genetically proxied ACE inhibition with lower risk of stroke (OR 0.94, 95% CI 0.88 to 1.01; P = 0.06) and coronary artery disease (OR 0.95, 95% CI 0.89 to 1.02; P = 0.16).

Across the 3 drug targets that we examined, conservative estimates of F-statistics for their respective genetic instruments ranged from 269.1 to 2,156.5, suggesting that our analyses were unlikely to suffer from weak instrument bias. Characteristics of genetic variants in ACE, ADRB1, and SLC12A3 used to proxy each pharmacological target are presented in Table 1 . Estimates of r 2 and F-statistics for each target are presented in S3 Table .

Discussion

In this mendelian randomization analysis of up to 289,612 cancer cases and 291,224 controls, genetically proxied long-term ACE inhibition was associated with an increased risk of colorectal cancer. This association was restricted to cancer of the colon, with similar associations across the proximal and distal colon. There was little evidence to support associations of genetically proxied ACE inhibition with risk of other cancers. Genetically proxied ADRB1 and NCC inhibition were not associated with risk of breast, colorectal, lung or prostate cancer.

Our findings for genetically proxied ACE inhibition and colorectal cancer risk are not consistent with some previous conventional observational analyses. A meta-analysis of 7 observational studies reported a protective association of ACE inhibitor use with colorectal cancer risk (OR 0.81 95% CI 0.70 to 0.92), though with substantial heterogeneity across studies (I2 = 71.1%) [14]. Interpretation of these findings is complicated by variable use of prevalent drug users, heterogenous comparator groups (both active controls and nondrug users), and the potential for immortal time bias across most included studies. Further, this meta-analysis did not include data from an earlier large Danish population-based case–control analysis with 15,560 colorectal cancer cases and 62,525 controls, which reported an increased risk of colorectal cancer (OR 1.30, 95% CI 1.22 to 1.39) among long-term users of ACE inhibitors (≥1,000 daily doses within 5 years of study entry), as compared to never-users [7].

The potential mechanisms underpinning an association between genetically proxied ACE inhibition and colorectal cancer risk are unclear. ACE is a multifaceted enzyme, capable of cleaving several different peptide substrates with potential roles in carcinogenesis [70]. Along with ACE inhibition leading to an accumulation of bradykinin and substance P, both potential inducers of tumor proliferation, ACE inhibition can also lead to an increase in Ac-SDKP, an endogenous antifibrotic peptide that is capable of inducing angiogenesis [71]. The observed restriction of an association of genetically proxied ACE inhibition with risk of colon, but not rectal, cancer is consistent with evidence that mRNA and protein levels of ACE are enriched in the colon but not in rectal tissue [72]. There was limited evidence of association of a serum ACE genetic risk score with distinct gene expression profiles in transcriptome-wide analyses. However, gene set enrichment analysis of these findings suggested enriched expression of genes involved in immunological pathways relating to memory CD8 T cells and coexpression network analysis identified ACE expression in a cluster of coexpressed genes enriched for colorectal cancer risk susceptibility genes (e.g., LAMA5, PNKD, TOX2, PLEKHG6) [73]. These findings suggest potential future avenues of exploration to uncover mechanistic pathways linking ACE with colorectal cancer risk.

Meta-analyses of randomized trials have not reported increased rates of cancer among ACE inhibitor users, though these analyses have not reported findings separately for colorectal cancer [2,3]. Potential discrepancies in findings for colorectal cancer between this mendelian randomization analysis and previous clinical trials could reflect the relatively short duration of these trials (median follow-up of 3.5 years) given long induction periods of colorectal cancer. For example, the “adenocarcinoma sequence” proposes that transformation of normal colorectal epithelium to an adenoma and ultimately to invasive and metastatic cancer may occur over the course of several decades [74,75]. Consistent with this long induction period, in randomized controlled trials examining the chemopreventive effect of aspirin on colorectal cancer risk, protective effects of aspirin are not seen until 7 years after initiation of treatment, with clear risk reductions becoming apparent only after 10 years of follow-up [76]. It is therefore possible that adverse effects of ACE inhibition on colorectal cancer risk may likewise not emerge until many years after treatment initiation. Alternatively, it may be possible that an effect of ACE inhibition on cancer is restricted solely to the earliest stages of the adenoma-carcinoma sequence and therefore may not influence cancer risk among largely middle-aged participants of clinical trials if dysplasia is already present. Finally, it is possible that lower levels of circulating ACE concentrations may influence colorectal cancer risk only during a particular critical or sensitive period of the life course (e.g., in childhood or adolescence), given some evidence to suggest a potential role of early-life factors in colorectal carcinogenesis [77].

Our largely null findings for genetically proxied ACE inhibition and risk of breast, lung, and prostate cancer risk are not consistent with some previous observational reports that compared ACE inhibitor users to nonusers or to users of β blockers or thiazide diuretics [7,9,17]. However, our findings for genetically proxied ACE inhibition are in agreement with those from short-term randomized controlled trials for these site-specific cancers and suggest that long-term use of these drugs may not influence cancer risk, though we cannot rule out small effects from their long-term use [3]. Likewise, our findings for ADRB1 and NCC are in agreement with short-term trial data reporting no association of β blockers and thiazide diuretics use with overall cancer risk [2].

Strengths of this analysis include the use of cis-acting variants in genes encoding antihypertensive drug targets to proxy inhibition of these targets, which should minimize confounding, the employment of various sensitivity analyses to rigorously assess for violations of mendelian randomization assumptions, and the use of a summary-data mendelian randomization approach, which permitted us to leverage large-scale genetic data from several cancer GWAS consortia, enhancing statistical power and precision of causal estimates. As with prior mendelian randomization analyses of antihypertensive drug targets that used similar approaches to instrument construction to our analysis, the general concordance of estimates of the effect of these instruments on cardiometabolic endpoints with those reported in prior clinical trials for these medications supports the plausibility of these instruments [78,79]. Finally, the use of germline genetic variants as proxies for antihypertensive drug targets facilitated evaluation of the effect of the long-term inhibition of these targets, which may be more representative of the typically decades-long use of antihypertensive therapy as compared to periods of medication use typically examined in conventional observational studies and randomized trials. Despite the evidence suggesting a link between genetically proxied ACE inhibition and colorectal cancer, however, these findings cannot demonstrate a causal relationship between ACE inhibitor use and colorectal cancer risk; only a randomized controlled trial could establish this relationship.

There are several limitations to these analyses. First, mendelian randomization analyses are restricted to examining on-target (i.e., target-mediated) effects of therapeutic interventions. Second, statistical power was likely limited in some analyses of less common cancer subtypes. Limited statistical power in analyses of genetically proxied ACE inhibition and colorectal cancer risk in East Asians (instrumented by rs4343) may also have accounted for the lack of association between these traits within this population. Identification of stronger genetic instruments for ACE inhibition in East Asian populations can help to uncover whether the lack of transportability of ACE and colorectal cancer findings from Europeans to East Asians reflects lower statistical power in the latter or differences in local LD structure across ancestries. In addition, statistical power can be limited in colocalization analysis, which can reduce the probability of shared causal variants across traits examined being detected (i.e., leading to “false negative” findings). We therefore cannot rule out the possibility that some colocalization findings suggesting low posterior probabilities of shared causal variants across traits (e.g., colocalization analyses for genetically proxied NCC inhibition and ER− breast cancer risk) reflected the limited power of this approach. Third, while we were able to perform sensitivity analyses for ADRB1 inhibition by restricting instruments to SNPs that were also eQTLs for ADRB1, we were unable to find evidence that the SNP used to instrument NCC (rs35797045) was also an eQTL for the gene encoding this target. Fourth, while these analyses did not account for previously reported associations of genetically proxied elevated SBP with antihypertensive medication use within the colorectal cancer datasets analyzed, such correction would be expected to strengthen, rather than attenuate, findings presented in this study [80]. Likewise, in “instrument validation” analyses for ADRB1, our inability to recapitulate known effects of ADRB1 inhibition (via beta blockers) on stroke risk could reflect the aforementioned inability to account for blood pressure medication users in the stroke datasets analyzed. It is also possible that these findings reflect the presence of horizontal pleiotropy in the instrument biasing the effect estimate toward the null and/or a nontarget-mediated effect of these medications on stroke risk. Fifth, though mixed-ancestry GWAS were used to construct instruments for serum ACE concentrations, effect allele frequencies for variants used in this instrument were similar across European and Latin American ancestry participants, suggesting that mendelian randomization findings were unlikely to be influenced by confounding through residual population stratification. Sixth, effect estimates presented make the additional assumptions of linearity and the absence of gene–environment and gene–gene interactions. Seventh, our genetically proxied ADRB1 findings are of greater relevance to second generation β blockers (e.g., atenolol and metoprolol), which selectively inhibit ADRB1 as compared to first generation β blockers (e.g., propranolol and nadolol), which equally inhibit ADRB1 and ADRB2 [81]. Future mendelian randomization analyses examining the potential effects of long-term first generation β blocker use incorporating both ADRB1 and ADRB2 variants is warranted. Eighth, we cannot rule out findings presented being influenced by canalization (i.e., compensatory processes being generated during development that counter the phenotypic impact of genetic variants being used as instruments). Finally, while various sensitivity analyses were performed to examine exchangeability and exclusion restriction violations, these assumptions are unverifiable.

Colorectal cancer is the third most common cause of cancer globally [82]. Given the prevalence of ACE inhibitor use in high- and middle-income countries and growing use in low-income countries, and the often long-term nature of antihypertensive therapy, these findings, if replicated in subsequent clinical trials, may have important implications for choice of antihypertensive therapy [4]. Importantly, given that hypertension is more prevalent among those who are overweight or obese (risk factors for colorectal cancer), these findings suggest that long-term use of this medication could increase colorectal cancer risk among populations who are already at elevated risk of this disease. There are different types and classes of ACE inhibitors, which vary in their pharmacodynamic and pharmacokinetic properties [83]. These differing pharmacological properties (e.g., differential absorption rate, affinity for tissue-bound ACE, and plasma half-life) can influence therapeutic benefit (or experience of adverse effects) of ACE inhibitors [84,85]. Future evaluation of the potential effects of long-term ACE inhibitor use on cancer risk should therefore include assessment of whether findings are specific to individual agents or classes of ACE inhibitors. As data on circulating levels of ADRB1 and NCC become available in future GWAS, there would be merit in evaluating whether findings presented in this analysis can be replicated when using alternate instruments developed from protein quantitative loci for these targets. Further work is warranted to unravel molecular mechanisms underpinning the association of ACE with colorectal cancer risk. In addition, extension of the analyses presented in this study to a survival framework could inform on whether concurrent use of ACE inhibitors may have an adverse effect on prognosis among colorectal cancer patients. Finally, findings from this analysis should be “triangulated” by employing other epidemiological designs with orthogonal (i.e., nonoverlapping) sources of bias to each other to further evaluate the association of ACE inhibition and colorectal cancer risk [86].

[END]

[1] Url: https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1003897

(C) Plos One. "Accelerating the publication of peer-reviewed science."
Licensed under Creative Commons Attribution (CC BY 4.0)
URL: https://creativecommons.org/licenses/by/4.0/


via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/