Association Between the ACE Insertion/Deletion Polymorphism and Risk of Lower-Limb Amputation in Patients With Long-Standing Type 1 Diabetes

Diabetes is a leading cause of nontraumatic lower-limb amputation (LLA), which is a common and devastating complication (1–3). LLA is a major public health problem worldwide. It is eight times more frequent in people with diabetes than in those without diabetes, and >50% of nontraumatic LLAs are attributable to diabetes complications (4). LLA is associated with an increased risk of lethal cardiovascular and noncardiovascular complications in patients with diabetes, resulting in a dramatic decrease in life expectancy (5–7). Pathological conditions leading to LLA include peripheral artery disease (PAD), peripheral diabetic neuropathy, impaired wound healing, and bone and soft-tissue infections. Patients undergoing LLA usually present with traditional cardiovascular risk factors (2,8), but the role of noncanonical genetic factors has seldom been investigated in people with type 1 diabetes (9).

ACE is a key regulator of the renin-angiotensin-aldosterone system (RAAS) and the kallikrein-kinin system. It converts angiotensin I into the vasoconstrictor peptide angiotensin II and inactivates the vasodilator peptide bradykinin. ACE is an ectoenzyme of vascular endothelial cells that is secreted into the circulation. Its tissue and circulating levels are genetically determined. A 287-bp Alu insertion/deletion (I/D) in intron 16 of the ACE gene (rs1799752) is dominantly or codominantly associated with cellular and plasma ACE levels. This genomic variation has been widely studied in relationship to diabetic kidney disease (DKD) in people with type 1 diabetes (10,11). We recently reported an excess risk of major kidney outcomes and all-cause death in individuals with type 1 diabetes carrying the ACE D allele (12). The association between ACE I/D and cardiovascular diseases was also examined in many populations with contrasting results (13–17). However, the relationship between ACE I/D genotype and the risk of LLA has not been investigated in individuals with diabetes. In the present work, we evaluated the associations of the ACE I/D polymorphism and plasma concentration of ACE with the risk of nontraumatic LLA in three prospective cohorts of individuals with long-standing type 1 diabetes.

We analyzed data from three French and Belgian prospective cohorts designed to investigate biochemical and genetic determinants of kidney and cardiovascular complications in people with type 1 diabetes (10,18,19). Survival Genetic Nephropathy (SURGENE) was a single-center study of all volunteers with type 1 diabetes attending the diabetes clinic at the university hospital of Angers, France (18). Participants were selected from 1989 to 1996 on the basis of a diagnosis of type 1 diabetes before the age of 40 years, with a duration of diabetes >3 years. The principal exclusion criteria were end-stage kidney disease and presence of other chronic disease. The Génétique de la Néphropathie Diabétique (GENEDIAB) study was conducted in 17 diabetes clinics in France and Belgium (see Supplementary Material). Participants were recruited from May 1994 to April 1995 on the basis of the following criteria: type 1 diabetes diagnosed before the age of 35 years and duration of diabetes >5 years, with a past or present history of preproliferative or proliferative diabetic retinopathy (10). The main exclusion criteria were terminal cancer and personal disability. The Genesis France-Belgium Study (GENESIS) was a family-based cohort that included probands with type 1 diabetes for at least 5 years (19). Participants were recruited from November 1998 to December 2000 on the basis of a diagnosis of type 1 diabetes before the age of 35 years, with initial ketosis and requirement for permanent insulin treatment within 1 year of diagnosis and past or present diagnosis of diabetic retinopathy. Study protocols were approved by the Ethics Committee of Angers University Hospital (Angers, France), and all participants from the three cohorts gave written informed consent. Characteristics of participants at baseline in each cohort were previously published (20–22) and are shown in Supplementary Table 1. Data from the three cohorts were pooled together for the current analysis.

Among 1,347 patients enrolled in the three cohorts, we excluded participants without data regarding LLA at baseline (n = 29) and those for whom ACE I/D genotyping was not available (n = 17). Hence, 1,301 participants were evaluated in the prevalence study (Supplementary Fig. 1). Participants were followed until death or the latest clinical visit up to 31 May 2019. After exclusion of patients with prevalent LLA at baseline (n = 90) and those without LLA data during follow-up (59 deaths and 145 loss to follow-up), 1,007 participants were included in the incidence study (Supplementary Fig. 1). Participants without LLA data during follow-up were not considered in the incidence study. We checked that the ACE genotype distribution was similar in participants with (II 17%, ID 48%, DD 35%) or without (II 15%, ID 50%, DD 35%) follow-up data regarding LLA. We considered all-cause death as a competing risk (see below).

LLA was defined as a minor (below-the-ankle amputation consisting of at least one ray metatarsal resection) or major (above-the-ankle amputation consisting of transtibial or transfemoral) amputation resulting from nontraumatic causes. Toe amputations under the metatarsal head were not recorded in this study. Presence of LLA at baseline and during follow-up was reported in the case report form by using a dedicated questionnaire. An independent adjudication committee validated the diagnosis of LLA on the basis of surgical reports. Data regarding the anatomic levels of amputations during follow-up were collected retrospectively. Peripheral diabetic neuropathy (absence of Achilles reflex and the loss of 10g monofilament sensation and/or the loss of vibration perception), PAD (absence of foot pulses and/or intermittent claudication), myocardial infarction, and stroke were recorded at baseline and validated in the same way. Diabetic retinopathy was staged at baseline as absent, nonproliferative, and proliferative. DKD was defined as estimated glomerular filtration rate (eGFR) (computed using the Chronic Kidney Disease Epidemiology Collaboration equation) <60 mL/min/1.73 m² and/or urinary albumin concentration (UAC) ≥30 mg/L.

ACE I/D genotypes were determined centrally by PCR as previously described (10). Plasma ACE was measured centrally at baseline in 687 participants (subsets of GENEDIAB and SURGENE participants) using an immunoradiometric method that quantifies ACE independently of its enzymatic activity (23,24). HbA_1c (high-performance liquid chromatography), lipids (colorimetric methods), UAC (nephelometry), and serum creatinine (derivation of Jaffé method with adjustment to the enzymatic method when introduced in routine practice) were measured centrally at baseline.

We investigated the prevalence of LLA at baseline (prevalence study) in the whole study and its incidence (the occurrence of the first case of LLA) during follow-up in participants without a history of LLA at baseline (incidence study). We also investigated the prevalence of LLA at the end of follow-up (baseline plus incident cases during follow-up).

Categorical variables were expressed as number (percentage), and continuous variables as mean ± SD or median (25th, 75th percentile) for those with skewed distribution. Baseline characteristics were compared using χ², ANOVA, or Wilcoxon tests. Missing data were rare (Supplementary Table 1) and removed from analyses that included the covariate (except for smoking status when it was included in the adjusted model, see below).

A logistic regression model was used to test the association between ACE I/D genotype and prevalent LLA at baseline and at the end of follow-up. Data were expressed as odds ratio (OR) with related 95% CI for XD (DD or ID) versus II genotype. The use of this genetic model was based on previous and present observations that the D allele is associated in a dominant manner with diabetes complications (10–12).

Kaplan-Meier curves were plotted to evaluate survival rates without incident LLA during follow-up by ACE genotypes and compared using the log-rank test. Cox proportional hazards regression models were fitted to estimate hazard ratio (HR), with associated 95% CI, for the risk of LLA during follow-up according to ACE genotype (XD vs. II). Analyses were adjusted for cohort membership, sex and age at baseline (model 1), plus a series of potential confounding variables at baseline, including tobacco smoking (coded as never, former, current, or unknown), duration of diabetes, HbA_1c, systolic and diastolic blood pressure, UAC, eGFR, and use of ACE inhibitors and lipid-lowering drugs (model 2). We did not consider the use of angiotensin receptor blockers in this model because it was infrequent (2%) in our cohorts at baseline. We also computed subdistribution HR, and related 95% CI, for risk of incident LLA during follow-up while accounting for the competing risk of major kidney outcome (occurrence of end-stage kidney disease or a 40% eGFR drop during follow-up) or all-cause death (further adjusting for model 2) using the Fine and Gray method. We tested the interaction between cohort membership or DKD at baseline and ACE genotype in their associations with LLA by including multiplicative interaction terms in the regression models. The Schoenfeld residuals method was used to check the proportional hazards assumption for genotype-LLA association (P = 0.16).

Harrell C-statistic, integrated discrimination improvement (IDI), and continuous net reclassification improvement (NRI) indexes were computed to evaluate the performance of ACE I/D genotype in stratifying LLA beyond model 2. We also tested the association between LLA and plasma ACE concentration (considered as a continuous variable) using a linear regression model. Plasma ACE concentration was further categorized into three equally increasing tertiles and studied in participants with versus without LLA using logistic regression and Cox analyses.

We performed a series of sensitivity analyses. First, we tested the association between ACE genotype and the incidence of minor and major LLA, considered separately when data were available (without exclusion of prevalent LLA at baseline). Second, we tested the association between ACE genotype and the prevalence of LLA at the end of follow-up after further adjustment for the underlying conditions (peripheral neuropathy, PAD, and infection) of diabetic foot ulcer (DFU) and other diabetes complications (chronic kidney disease, retinopathy stages, myocardial infarction, and stroke) further adjusting for model 2. We also adjusted for DFU treatment (peripheral revascularization, antibiotic therapy, both, or none). Third, we assessed the interaction between ACE genotype or plasma ACE and the use of ACE inhibitors in association with LLA. We also examined the association between ACE genotype and the prevalence of LLA at the end of follow-up separately in users and nonusers of ACE inhibitors. Fourth, we assessed the interactions between plasma ACE concentration and ACE I/D genotype in association with LLA. We also assessed the association between plasma ACE and LLA separately in II and XD carriers.

Statistics were performed using JMP Pro 14 (SAS Institute Inc., Cary, NC) and Stata 13 (StataCorp, https://www.stata.com) software. Two-sided P < 0.05 was considered significant.

Study population at baseline consisted of 708 (54%) men and 593 (46%) women. Age, duration of diabetes, and HbA_1c were (mean ± SD) 41 ± 13 years, 24 ± 11 years, and 8.8 ± 1.8%, respectively. The DD, ID, and II genotypes were observed in 453 (35%), 627 (48%), and 221 (17%) participants, respectively (Hardy-Weinberg equilibrium χ² = 0.03, P = 0.87). Characteristics of participants at baseline by ACE genotype (II vs. XD) are shown in Supplementary Table 2. Clinical and biological features, treatment, and history of microvascular and macrovascular complications were comparable between groups, except for history of myocardial infarction, which was more prevalent in XD than in II carriers.

A history of LLA at baseline was observed in 90 (6.9%) participants. Characteristics of participants by prevalent LLA at baseline are shown in Table 1. LLA was more prevalent in XD than in II carriers (7.4% vs. 4.5%, OR 2.07 [95% CI 1.02–4.17], P = 0.04, adjusted as in model 1). The association remained significant following adjustment for additional confounders as in model 2 (Table 2). No significant interaction was observed between ACE I/D genotype and use of ACE inhibitors (P = 0.07), cohort membership (P = 0.24), or baseline history of DKD (P = 0.99) in the associations with prevalent LLA at baseline.

Among 1,007 participants without a history of LLA at baseline, incident LLA occurred in 53 (5.3%) individuals during a median (25th, 75th percentile) duration of follow-up of 14 (6, 19) years, corresponding to 13,753 person-years and an incidence rate of 3.8 (95% CI 2.9–5.0) per 1000 person-years. Baseline characteristics of participants with or without incident LLA during follow-up are presented in Table 1. The cumulative incidence (5.9% vs. 2.2%) and the incidence rate of LLA (4.3 [95% CI 3.2–5.7] vs. 1.7 [0.6–4.5] per 1,000 person-years) were higher in XD than in II carriers (Fig. 1). The relative risk of LLA was significantly higher in XD than in II carriers (HR 3.01 [95% CI 1.22–9.97], P = 0.01, adjusted as in model 1). The association remained significant after further adjustment as in model 2 (Table 2) and when we considered major kidney outcome (subdistribution HR 3.15 [95% CI 1.020–9.81], P = 0.04) or all-cause death (3.19 [1.01–10.09], P = 0.04) as a competing risk. No interaction was observed between ACE genotype and use of ACE inhibitors (P = 0.95), cohort membership (P = 0.99), or baseline history of DKD (P = 0.99) in the associations with incident LLA.

Among participants for whom data were available, 37 (3.5%) had a minor LLA (13 metatarsal resections of one to four rays and 24 transmetatarsal resections of five rays) and 26 (2.4%) had a major LLA (16 transtibial and 10 transfemoral). The D allele was significantly associated with increased risk of minor LLA (HR 3.63 [95% CI 1.08–22.65], P = 0.03), but not with major LLA (1.88 [0.62–8.28], P = 0.29), after adjustment in model 2 (Supplementary Table 3).

At the end of follow-up, LLA was present in 143 (11%) participants. It was more prevalent in XD than in II carriers (11.9 vs. 6.3%, OR 2.48 [95% CI 1.33–4.65], P = 0.004 after adjustment as in model 2) (Table 2). The association remained significant after adjustment for the underlying conditions of DFU (peripheral neuropathy, PAD, and foot infection) and history of chronic kidney disease, diabetic retinopathy stages, myocardial infarction, and stroke, further adjusting for model 2 (3.06 [1.27–7.42], P = 0.01). The association also remained significant after further adjustment in model 2 for DFU management (peripheral revascularization, antibiotics, both, or none), further adjusting for model 2 (3.45 [1.01–11.79], P = 0.04).

No significant interaction was observed between ACE genotype and cohort membership (P = 0.29) or baseline history of DKD (P = 0.94) in the associations with prevalent LLA at the end of follow-up. The association between ACE I/D polymorphism and the prevalence of LLA at the end of follow-up was higher in nonusers of ACE inhibitors at baseline than in users (Supplementary Table 4), but no significant interaction was observed between ACE genotype and use of ACE inhibitors in the association with LLA (P = 0.08). When added to model 2, ACE genotype improved NRI (0.161 [95% CI, 0.026–0.305], P = 0.02) but not IDI (0.008 [−0.001 to 0.025], P = 0.22) or the Harrell C-statistic index (change 0.005 [−0.005 to 0.015], P = 0.33) for stratifying LLA at the end of the study (Supplementary Table 5).

The median (25th, 75th percentile) plasma ACE concentration was 446 (339, 579) ng/mL at baseline in the subset of participants for whom data were available. It was significantly higher in XD than in II carriers, as expected (470 [359, 599] vs. 356 [269, 445] ng/mL, P < 0.0001). Baseline plasma ACE concentration was not significantly different in participants with or without a prevalent or incident LLA during follow-up (Table 1). Consequently, baseline plasma ACE concentration was not different in participants who had or had not undergone LLA at the end of follow-up (462 [349, 570] vs. 443 [332, 579] ng/mL, P = 0.67). No significant association was observed among tertiles of plasma ACE concentration and LLA after adjusting for cofounders (Table 3) or when analysis was stratified by use of ACE inhibitors at baseline (Supplementary Table 3).

We observed a significant interaction between ACE I/D genotype and baseline plasma ACE concentration in the association with prevalent LLA at the end of follow-up (P for interaction = 0.007). Among II carriers, plasma ACE concentration was significantly higher in participants with prevalent LLA than in those without (mean [95% CI]: 449 [360–539] vs. 354 [286–423] ng/mL, P = 0.03, adjusted as in model 2). Such a difference in baseline plasma ACE concentration according to LLA status was not observed in XD carriers (512 [454–570] vs. 537 [488–586], P = 0.27, respectively).

In the current study, we observed associations between the D allele of the ACE I/D polymorphism and increased prevalence at baseline and incidence during follow-up of LLA in patients with long-standing type 1 diabetes. These associations were independent of a range of potentially confounding variables, including diabetes duration and glycemic control, traditional cardiovascular risk factors, and kidney function determinants. In addition, the association remained significant after adjusting for other diabetes complications, including DKD, diabetic retinopathy, myocardial infarction, and stroke, or treating major kidney outcomes or all-cause death as a competing risk. Overall, the association between ACE I/D polymorphism and LLA seems to be independent of vascular conditions. However, our findings do not allow us to draw conclusions about causality and mechanism and to rule out potential vascular conditions linking the ACE D allele to the risk of LLA in people with type 1 diabetes, especially DKD.

The association between ACE variant and LLA was driven by an increased risk of minor, but not major, amputations. This observation suggests that the ACE variant may be linked to microvascular disease rather than to a macrovascular condition. We recently reported a significant association between ACE I/D polymorphism and major kidney events (12), although no association was observed between the variant and the incidence of major cardiovascular events (unpublished data).

The addition of ACE I/D genotype to usual risk factors enhanced the NRI, but not the IDI or C-statistic, index for stratifying LLA. These findings suggest that the variant can help in reclassifying the risk of LLA, albeit modestly, but do not support its use as a strong prognostic biomarker for lower-extremity disease in people with type 1 diabetes.

To our knowledge, this report is the first of an independent association between ACE I/D genotype and risk of LLA in patients with type 1 diabetes. Few studies investigated potential genetic determinants of lower-limb complications in people with diabetes (25). Some studies evaluated the association between ACE I/D genotype and lower-limb PAD in the general population, with contrasting findings (26–29). A retrospective case-control association study reported a significant association between the D allele and PAD in 281 patients (17% with diabetes) with symptoms of intermittent claudication and an ankle-brachial index at rest of <0.90 (27). In contrast, the Health, Aging and Body Composition study, a general population prospective cohort study, did not show a consistent association between the ACE I/D polymorphism and PAD in elderly populations (28). A lack of association between the ACE I/D variant and PAD was also reported in a case-control study that included 522 Austrian patients with documented PAD (46% with diabetes) and 522 control subjects (13% with diabetes) (29). The ACE I/D polymorphism was also linked to peripheral neuropathy in people with type 2 diabetes (30). Overall, these association studies are difficult to interpret because of power issues, heterogeneity of input clinical material and mixing of patients with and without diabetes and individuals with type 1 and type 2 diabetes. Our study overcomes at least some of these limitations by addressing both prevalent LLA at baseline and incident LLA after a prolonged follow-up in extensively phenotyped cohorts of subjects with type 1 diabetes. Our study also addresses the growing need to better understand the risk factors for LLA in people with type 1 diabetes, as the pathophysiology and clinical courses of vascular complications, including LLA, differ between type 1 and type 2 diabetes (31). Therefore, our findings cannot be extrapolated to LLA in type 2 diabetes.

We did not observe a significant association between circulating ACE levels and prevalent or incident LLA in the subset of participants for whom plasma ACE concentration was measured at baseline. However, we observed an interaction between ACE genotype and plasma ACE levels in the association with prevalent LLA at the end of follow-up. Plasma ACE concentration was significantly higher in II carriers who had undergone LLA compared with those who had not, but no such association was observed among D allele carriers. We do not have a conclusive explanation for these findings, but we might speculate that the well-documented stimulating effect of chronic hyperglycemia on plasma ACE concentration (10) could mitigate the genetic effect on this parameter, especially in the D allele carriers with genetically determined higher ACE level, and blunt a potential difference in ACE concentration between people with and without LLA. In any case, because of the observational nature of our study, our findings do not allow any conclusion regarding a causal role of ACE I/D genotype and circulating levels of ACE in the pathogenesis of LLA.

The use of ACE inhibitors at baseline did not influence the association between the ACE I/D polymorphism and LLA. Nevertheless, the association between the D allele and prevalent LLA was stronger in participants who were not on ACE inhibitors at baseline than in those who used these drugs, although no significant interaction was observed between the use of ACE inhibitors and genotype in LLA risk. Our study was not designed to investigate the effect of ACE inhibitors on LLA; however, our findings may suggest some kind of pharmacological protection associated with the use of these drugs among XD carriers with regard to the risk of LLA. Despite possible differential effects of RAAS inhibitors with respect to LLA in patients with diabetes (32), ACE inhibitors are recommended as a first-line treatment in patients with PAD and hypertension (33).

ACE may exert deleterious effects in the lower-extremity circulation not only through angiotensin production but also through kinin inactivation. Indeed, bradykinin promotes the release of several vasodilatory and antithrombotic mediators, including nitric oxide, prostacyclin, and tissue plasminogen activator (34,35). Kinins also mobilize progenitor cells with vasculogenesis capacity (36). The role of bradykinin in postischemic angiogenesis has been well documented experimentally, especially in diabetes (37,38). ACE inhibitors have been shown to stimulate postischemic angiogenesis in diabetes through kinin-receptor activation (39). The association between the ACE D allele and LLA may also reflect the consequences of the vasoconstrictor and proinflammatory actions of angiotensin II. Of note, arginine vasopressin, which is stimulated by RAAS, was associated with the risk of LLA in people with diabetes (40).

The strengths of our work are the investigation of a robust and adjudicated end point in three multicenter, binational cohorts of middle-aged patients with long-standing type 1 diabetes and the collection of a comprehensive set of demographic, clinical, and biological features at baseline. Both prevalence and incidence of LLA during a long follow-up duration (median 14 years) were studied. There are also limitations of our study to acknowledge. First, our study lacks comprehensive data regarding toe amputations and DFU and its underlying conditions. Peripheral neuropathy and PAD were recorded at baseline in the GENEDIAB and GENESIS cohorts but not in SURGENE. Of note, individuals with prevalent LLA at baseline were more likely to have a history of diabetic neuropathy or PAD at baseline compared with those without these conditions. The trend was similar for incident LLA, even if we do not have comprehensive data regarding peripheral neuropathy and PAD during follow-up. Furthermore, the association between D variant and LLA remained significant after adjustment for PAD, peripheral neuropathy, foot infection, and DFU treatments. Second, we investigated a single, albeit highly informative, gene variant and did not attempt external replication of the findings in other populations. Nonetheless, we tested three different cohorts and did not observe a significant interaction between ACE I/D polymorphism and cohort membership in the association with LLA, thereby supporting intercohort replication of the results. Finally, we studied predominantly people of European descent, and our findings may not apply to people with other ethnic backgrounds.

In summary, this report is the first of an independent and consistent association between the D allele of the ACE gene I/D polymorphism and excess risk of nontraumatic LLA in patients with long-standing type 1 diabetes. The association was mainly driven by minor LLA. Additional studies are needed to investigate the role of ACE and ACE gene allelic variations in the pathophysiology of lower-limb complications in people with diabetes.

Acknowledgments. The authors thank all the patients who participated to this study as well as their physicians. The list of contributors is available in the Supplementary Material.

Funding. Y.A. was supported by a grant from the Association Diabète Risque Vasculaire.

Duality of Interest. K.M. reportspersonal fees or nonfinancial support from Novo Nordisk, Sanofi, AstraZeneca, Eli Lilly, Abbott, and Boehringer Ingelheim. P.G. reports grants or personal fees from Abbott, Amgen, AstraZeneca, Boehringer Ingelheim, Eli Lilly, Janssen, Merck Sharp & Dohme, Mundipharma, Novo Nordisk, Sanofi, and Servier. A.S. has received lecturer/advisor fees from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Janssen, Merck Sharp & Dohme, Novartis, Novo Nordisk, Sanofi, and Servier. R.R. reports research grants from Sanofi, Novo Nordisk, and Diabnext and consulting and speakers’ bureau fees (compensation donated to the nonprofit AP-HP Foundation for research) from Sanofi, Novo Nordisk, Eli Lilly, Boehringer Ingelheim, Mundipharma, Janssen, AstraZeneca, MSD, Medtronic, and Abbott. S.H. reports personal fees and nonfinancial support from AstraZeneca, grants and personal fees from Bayer, personal fees from Boehringer Ingelheim, grants from Dinno Santé, personal fees from Eli Lilly, nonfinancial support from LVL Mediphar, personal fees and nonfinancial support from MSD, personal fees from Novartis, grants from Pierre Fabre Santé, personal fees and nonfinancial support from Sanofi, personal fees and nonfinancial support from Servier, and personal fees from Valbiotis. M.M. is a consultant for Novo Nordisk Algerian subsidiary and has received personal fees from Novo Nordisk, Merck Sharp & Dohme, and Eli Lilly. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. K.M., M.M., and G.V. designed the study, researched data, and wrote the manuscript. Y.A., C.C., L.P., S.D., N.F., V.R., J.-F.G., P.G., G.C., R.R., A.S., B.B., F.A.-G., and S.H. researched data, contributed to the discussion, and reviewed/edited the manuscript. All authors approved the current version of the manuscript. K.M. and G.V. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analyses.