This systematic review and meta-analysis study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. ²⁷ The protocol was registered with PROSPERO ²⁸ under registration number CRD42024577680 and is published as a Preprint in Open science Framework (OSF) preprints. ²⁹

Studies were eligible for inclusion if they enrolled male and female participants aged 18 years or older with a confirmed diagnosis of type 2 diabetes mellitus (T2DM), and investigated angiotensin-converting enzyme (ACE) insertion/deletion (I/D) genotypes (II, ID, DD) as the primary exposure.

Diabetic nephropathy (DN) was the primary outcome, commonly defined using microalbuminuria markers such as the albumin excretion rate (AER), urine albumin-to-creatinine ratio (UACR), or albumin-to-creatinine ratio (ACR). Diagnostic thresholds for microalbuminuria were standardized across studies at 30–300 mg/day for AER and 30–300 mg/g for UACR/ACR. One study that defined DN using traditional renal function tests, including estimated glomerular filtration rate (eGFR) and serum creatinine, was also included to capture broader clinical definitions of nephropathy. This deviation from the protocol was noted but deemed valuable for completeness.

The comparator across studies was typically T2DM patients with normal UACR levels, enabling evaluation of the association between ACE genotypes and DN risk. Secondary outcomes included nephropathy-related mortality and comorbidities such as cardiovascular disease, hypertension, neuropathy, and retinopathy. Eligible studies employed observational designs (cohort, case-control, or cross-sectional) and were published between January 1990 and February 2025.

Articles with missing data on key study variables or those that did not explore the ACE I/D-DN relationship were excluded. Non-English articles whose abstracts were available in English were included in the initial screening; however, those whose full texts were not available for retrieval were eventually excluded. Abstracts, unpublished studies, and grey literature were not included in this review, and no direct contact was made with study authors for clarification or additional data. Discrepancies were resolved through discussion, ensuring consistency and minimizing bias, although formal blinding and interrater reliability statistics were not applied.

We conducted a comprehensive search on across multiple electronic databases (PubMed, Web of Science, and EMBASE) to identify studies relevant to the research objectives. Additional studies were retrieved from Google Scholar and reference lists of the included articles to ensure broad coverage.

The search strategy was designed to capture studies assessing the relationship between angiotensin-converting enzyme (ACE) insertion/deletion (I/D) genotypes and diabetic nephropathy (DN). The search included studies published from 1 ^st January 1, 1990, to 28 ^th February 28, 2025, without language restrictions. Keywords and medical subject headings (MeSH) related to “ACE I/D genotypes,” “diabetic nephropathy,” and “microalbuminuria” were combined with Boolean Operators “OR” and “AND” “AND” and truncated where applicable to optimize retrieval. The search strings used are listed in Table 1.

Two independent experienced information retrieval specialists (AAK and ML), in consultations with the lead reviewer (RK), independently conducted the database searches and verified the inclusion of additional studies from alternative sources such as Google Scholar and reference lists. Discrepancies in the search results were resolved through consensus discussion. The systematic review process was organized using Rayyan, ³⁰ a web-based platform that facilitates the filtering and assessment of papers based on predetermined requirements for inclusion and exclusion.

Two reviewers, RK and JT, independently assessed the titles and abstracts of the studies according to the inclusion criteria using Rayyan software. Full-text articles were retrieved from studies that met the eligibility requirements. Any disagreements regarding inclusion were resolved through a consensus between the two reviewers. Full-text articles were retrieved from eligible studies, and the selection process is shown in Figure 1.

Data extraction was performed independently by two authors, RK and JT, using a standardized data extraction form in Microsoft excel. The extracted data included study characteristics (author, year, population, and design, age, sex and clinical characteristics such as glycated hemoglobin levels, blood pressure, T2DM duration, body mass index), exposure details (ACE I/D genotypes), comparators (UACR measurement methods), outcomes (DN, mortality, and comorbidities), and statistical measures such as the frequency counts were also retrieved and computed in to Odds ratio with confidence intervals. Discrepancies were resolved through discussion with GNK and RA.

The data extraction captured a broad range of information including details on study design such as case control, cross-sectional or cohort, sample size and distribution between males and females, age, genotyping methods such as PCR, UACR or other DN diagnostic approaches, and outcome reporting. Genotype frequencies were entered in Excel, and odds ratio (OR) with 95% confidence intervals (CI) were calculated using R software version 4.4.2. Secondary outcomes such as comorbidities and mortality were recorded as binary variables (Yes/No) based on whether a particular study explicitly reported on them. No missing data were encountered during the meta-analysis, as only studies with complete and extractable data were included in the quantitative synthesis. Studies lacking sufficient data, as identified during visual inspection of the compiled Excel sheet, were excluded from meta-analysis and instead included in the narrative synthesis.

The risk of bias in individual studies was assessed using the ROBINS-E tool, ³¹ which evaluates bias across key domains: confounding, selection bias, classification of interventions, and outcome assessment. Confounding was assessed as part of the ROBINS-E risk of bias evaluation, where considerations were made on whether studies appropriately controlled for potential confounders such as age, sex, duration of diabetes, and comorbidities. Where studies reported such adjustments or stratifications, these were documented and factored into the overall bias assessment. Studies were assigned an overall remark of either a low, moderate or some concerns, or high risk of bias. The risk of the bias assessments as depicted by the robvis tool ³² are shown in Figure 2. The results of the bias assessment are available as Extended data. The ROBINS-E risk of bias assessment was conducted independently by CNB and SPR.

We included 25 studies in the meta-analysis to assess the association between ACE I/D genotypes and diabetic nephropathy risk. For the primary outcome, diabetic nephropathy (DN) assessed using urine albumin to creatinine ratio or renal function tests, odds ratio (OR) with 95% confidence intervals (CI) were used as the effect measure to compare the risk of DN across ACE I/D genotypes (II vs. ID and II vs. DD). This binary outcome metric was selected due to the categorical nature of the genotype data and the case or non-case definition of DN. Odds ratio OR were calculated using genotype frequency data. Secondary outcomes such as comorbidities (hypertension, cardiovascular disease) and mortality were synthesized narratively, and their presence (Yes) or absence (No) was recorded per study.

Study selection and eligibility for synthesis

Studies were deemed eligible for synthesis if they included adult patients with T2DM, assessed ACE I/D polymorphisms and reported DN outcomes using laboratory-based criteria. The review included studies that assessed DN using albumin excretion rate (AER) or microalbuminuria, albumin-to-creatinine ratio (ACR), or renal function markers such as serum creatinine and estimated glomerular filtration rate (eGFR). These criteria allowed for flexible yet clinically relevant definitions of nephropathy. Study characteristics and outcome definitions were tabulated to determine synthesis eligibility. Studies without sufficient genotype data or outcome definitions were excluded.

Data preparation and conversion

Data preparation involved standardizing genotype frequency data across included studies. Frequencies for ACE I/D genotypes were extracted and used to calculate odds ratio (OR) with corresponding 95% confidence intervals (CI). For meta-analysis, all effect sizes were transformed into log odds ratios and back-transformed for interpretation. Secondary outcomes such as comorbidities and mortality were recorded as binary variables (reported vs. not reported) in Excel spreadsheets.

Reporting study outcomes

For the primary outcome, diabetic nephropathy (DN), the standardized metric used was the odds ratio (OR) with corresponding 95% confidence intervals (CI), comparing DN prevalence across ACE I/D genotypes (II, ID, DD). OR were chosen because they are appropriate for summarizing associations in observational studies. Genotype frequencies from individual studies were extracted into Microsoft Excel, and OR were calculated using R software (version 4.4.2). These OR were log-transformed for meta-analysis to stabilize variance and ensure normality assumptions. The pooled estimates were then back-transformed to the original OR scale.

For secondary outcomes, such as mortality, cardiovascular disease, hypertension, neuropathy, and retinopathy, data were extracted into Excel as binary indicators (Yes/No) depending on whether each study reported on the specific outcome.

Statistical models

The primary effect measure used was the odds ratio (OR) for the association between ACE I/D genotypes and DN. Random-effects models were used for all meta-analyses to account for between-study heterogeneity in population, study design, and UACR diagnostic methods. Statistical significance was determined at p < 0.05. Heterogeneity was assessed using the I ² and τ ² statistics. Meta-regression was performed to explore heterogeneity sources, particularly differences in urinary albumin measurement methods (ACR vs. AER). All statistical estimates were generated using R software (version 4.4.2).

Presentation of results

Forest plots were generated to visually present pooled OR and 95% CI. Individual study estimates and weights were shown. Funnel plots and Egger’s regression tests assessed publication bias. All statistical visualizations and analyses were produced in R software. Reporting bias was examined via funnel plots and Egger’s test; no significant asymmetry was detected.

Subgroup and sensitivity analyses

To explore sources of heterogeneity, subgroup analyses were conducted based on urinary albumin measurement methods (ACR vs. AER) of the included studies. Meta-regression was used to assess the potential influence of these study-level characteristics on effect size estimates.

Sensitivity analyses were conducted using a leave-one-out approach, in which each study was sequentially removed from the meta-analysis to evaluate the robustness and stability of the pooled effect estimates.

Selective outcome reporting and qualitative assessments were done through quality appraisal of the studies conducted using the STREGA reporting guidelines. KR and DT assessed adherence, with discrepancies resolved by ENO and BBA. Studies were rated as fair, good, very good, or excellent. Excellent studies provided thorough methodological and genotyping details; very good ones had minor gaps; good studies had moderate reporting deficiencies; and fair studies lacked core STREGA elements. The full results of STREGA appraisal are provided in the Extended data.

We did not apply the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) framework, ³³ which is best suited for intervention studies. Instead, we appraised the certainty of evidence by triangulating risk of bias (via ROBINS-E), reporting quality of studies (via STREGA), consistency, precision of pooled estimates, and the direction of effects across studies.

The initial search across PubMed, EMBASE, Web of Science, Google Scholar, and reference lists yielded 968 records. After 189 duplicates were removed, 779 unique records were screened for eligibility. During the screening process, 721 records were excluded based on the title and abstract review. A total of 58 full-text articles were retrieved, all of which were successfully accessed. After full-text assessment, 12 studies were excluded due to incorrect population (n = 2), wrong outcome (n = 9), or wrong publication type (n = 1). Ultimately, 46 studies met the inclusion criteria and were included in this systematic review and meta-analysis. The complete screening and selection process is summarized in the PRISMA 2020 flowchart ³⁴ as shown in Figure 3.

After full-text review, 12 studies that initially appeared to meet inclusion criteria were excluded for the following reasons: a) Wrong publication type (n = 1): Panagiotopoulos et al., ³⁵ was a narrative review rather than a primary research study. b) Wrong outcome (n = 9): These studies, including Wong et al., ³⁶ investigated ACE polymorphisms however, interest was with macroangiopathy and not type 2 diabetes mellitus associated nephropathy; Ahluwalia et al., ³⁷ and Pai et al., ³⁸ assessed ACE polymorphisms not as insertion/deletion SNPs. Fujisawa et al., ³⁹ presented ACE I/D polymorphisms data in relation to retinopathy and myocardial infarction but not diabetic nephropathy (DN). Taha et al., ⁴⁰ assessed ACE polymorphisms in type 2 diabetes mellitus patients but did not stratify them into nephropathy cases and controls. Araz et al., ⁴¹ ACE ID, DD, II were not distributed across UACR/albuminuric groups. Canani et al., ⁴² ACE DD and ID genotypes were not reported independently. Wong et al., ⁴³ distribution of ACE I/D across different DN categories determined by UACR/Microalbuminuria/AER was not done. Draman et al., ⁴⁴ microalbuminuria/UACR was not measured. c) Wrong Population (n = 2): Pontremoli et al., ⁴⁵ focused on essential Hypertension and Al-Harbi et al., ⁴⁶ in the methods section of the full text did not specify the population as type 2 diabetes mellitus (T2DM) but unrelated adult patients. These studies were excluded to ensure consistency with our predefined Population Exposure Comparator Outcome (PECO) criteria and focus on clinically relevant DN outcomes in adults with T2DM.

We followed the Centre for Reviews and Dissemination (CRD) approach ⁴⁷ and the Synthesis Without Meta-analysis (SWiM) guidelines ⁴⁸ due to their detailed and structured framework. Forty-six (46) observational studies investigating the association between ACE I/D polymorphisms and DN in patients with T2DM were examined. The total sample size across all studies was 16,322 participants, with individual study populations ranging from 50 to 3744 participants. The combined mean of mean age in the studies was 58.11 (±4.30). The combined median of mean numbers of males was 90.5 (IQR: 62.75-171.5) and that of females was 105.5 (IQR: 57.0-183.75). Most studies were conducted in Asia (29/46), Europe (8/46), Middle East (6/46) and Africa (5/46) reflecting a diverse geographical representation. Studies were grouped based on outcome assessment methods (ACR vs. AER), study design (case-control, cross-sectional, or cohort), and genotype comparisons (II vs. DD, II vs. ID), to account for methodological and clinical heterogeneity and to enable meaningful synthesis of effect estimates across similar contexts. Most studies assessed the primary outcome of DN using either ACR ^{23, 25, 49– 70} or AER, ^{71– 91} whereas only one study used traditional renal function tests ⁹² (serum creatinine, eGFR, and urea levels). Genotyping for ACE I/D polymorphism was primarily conducted using polymerase chain reaction (PCR) methods, including conventional PCR, PCR-RFLP, high-resolution melting, and sequencing approaches. While all studies reported ACE I/D genotype distributions and DN outcomes, only five studies (10.7% of 46) explored associations with comorbidities, and one (2.2% of 46) reported on mortality. The majority identified either the D allele or DD genotype as being associated with increased DN risk, although several (40/46) reported null or population-specific findings.

The geographical distributions of the studies included in this review are shown in Figure 4 .

Due to the limited number of studies addressing ACE I/D polymorphism in relation to diabetic nephropathy and associated comorbidities, the included studies were synthesized narratively rather than statistically. The summary of findings examining ACE I/D polymorphism in relation to diabetic nephropathy and comorbidities is presented in Table 2.

From the 46 studies, only those with a low overall score on the ROBINS E assessment were included in the quantitative synthesis. We conducted a meta-analysis, meta-regression, subgroup analysis, sensitivity analysis, and publication bias assessment on 25 studies to evaluate the association between ACE I/D polymorphisms and diabetic nephropathy (DN) in the study population.

This was conducted in accordance with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines, ⁹³ and the corresponding checklist was completed to ensure comprehensive and transparent reporting. The included studies were deemed appropriate for assessing the hypothesis as they reported on ACE I/D genotypes and diabetic nephropathy outcomes in adult T2DM populations. Data were selected and coded based on sound clinical principles, prioritizing studies that clearly defined diabetic nephropathy using established albuminuria or renal function markers and reported ACE I/D genotype distributions relevant to the study objectives. Data classification and coding were performed independently by two reviewers using a standardized extraction template. Most included studies used standardized measures of albuminuria to define nephropathy, ensuring relevance to the primary outcome and alignment with the review objective. Data classification and coding were performed independently by two reviewers (RK and JT) using a standardized extraction template. Discrepancies were resolved through discussion with another reviewer; EAO, ensuring consistency and minimizing bias, although formal blinding and interrater reliability statistics were not applied.

Random-effects models using the DerSimonian–Laird method ⁹⁴ were used to estimate the pooled odds ratios (ORs) and 95% confidence intervals (CIs) across genotype comparisons to account for expected heterogeneity in study populations and methodologies across the included observational studies. Genotype II was associated with a lower risk of developing DN. The random-effects models showed a significant protective effect of the homozygous insertion (II) genotype, with pooled odds ratio (OR) of 0.70 (95% CI: 0.63–0.77) versus heterozygous insertion/deletion (ID) and 0.68 (95% CI: 0.55–0.84) versus homozygous deletion (DD). Heterogeneity across studies was moderate to substantial (I ² = 71.7%, τ ² = 0.1776, p < 0.0001), suggesting variability in the effect estimates, possibly due to methodological or population differences. A forest plot summarizing the pooled estimates and heterogeneity is shown in Figure 5.

Meta-regression was performed to explore the sources of heterogeneity in the association between ACE I/D genotypes and diabetic nephropathy (DN) risk. In the crude model (k = 25; DerSimonian–Laird estimator), the residual heterogeneity was τ ² = 0.1100 (SE = 0.0917; τ = 0.3317), with I ² = 52.3% (p = 0.0141) and R ² = 38.06%. The test for residual heterogeneity was significant (QE(12) = 25.16, p = 0.0141), while the test for moderators was not (QM(12) = 16.56, p = 0.17), indicating limited explanatory power overall. However, the form of UACR measurement was a significant individual predictor (β = 0.57, SE = 0.19, p = 0.0033; 95% CI: 0.19-0.95), suggesting that variations in measurement methods contributed meaningfully to heterogeneity in DN risk estimates. In the adjusted model, which included additional moderators, the residual heterogeneity increased (τ ² = 0.1253, SE = 0.0692; τ = 0.3540), with I ² = 63.0% (p < 0.0001) and R ² = 29.45%. The overall test for residual heterogeneity remained significant (QE(23) = 62.21, p < 0.0001). The test of moderators was significant (QM(1) = 5.79, p = 0.016), confirming that the form of UACR measurement remained a key source of heterogeneity (β = 0.40, SE = 0.17, p = 0.016; 95% CI: 0.07-0.72). The results are summarized in Table 3.

Subgroup analyses were based on the methods used to assess the primary outcome which is DN and these included; ACR, AER and RFTs. They revealed differences in the strength of associations between ACE I/D genotypes and diabetic nephropathy (DN) risk. Studies using albumin-to-creatinine ratio (ACR) showed the strongest protective effect of the II genotype, with pooled odds ratio (OR) of 0.59 (95% CI: 0.47–0.75) under the random-effects model and 0.59 (95% CI: 0.51–0.67) under the common-effect model. For studies using the albumin excretion rate (AER), the association was weaker (random-effects OR = 0.71, 95% CI: 0.55–0.92; common-effect OR = 0.77, 95% CI: 0.66–0.90). The weakest association was observed in the study using renal function tests (RFTs), with a random-effects OR of 0.68 (95% CI: 0.55– 0.84) and a common-effect OR of 0.70 (95% CI: 0.63–0.77). Heterogeneity varied across subgroups: I ² = 60.3% (p = 0.0019) for ACR and I ² = 58.3% (p = 0.0102) for AER; no heterogeneity was detected in the RFT-based subgroup. These results suggested that the method used to assess DN may influence the estimated association between ACE I/D polymorphisms with DN risk. The full summary of subgroup results is presented in Table 4.

Leave-one-out sensitivity analysis was conducted to assess the robustness of the findings and the influence of individual studies on the pooled effect estimates. One study was excluded at a time, and the overall odds ratio (OR) and 95% confidence interval (CI) were calculated. The pooled log OR ranged from -0.34 to -0.45, with minimal variation across iterations. The confidence intervals remained narrow and did not cross the null, suggesting consistent associations. The exclusion of Shaihk et al. ⁹² produced the most marked change (log OR = -0.4474; 95% CI: -0.6255 to -0.2692), followed by Kuramoto et al. ⁸¹ and Tseng et al., ⁶⁶ though none altered the overall interpretation. These results indicated that no single study had a disproportionate impact on the association between ACE I/D genotypes and diabetic nephropathy in the meta-analysis. The full results are shown in Table 5 and visualized in Figure 6.

The funnel plot shown in Figure 7 and Egger’s regression test were used to assess the publication bias. The plot shows a relatively symmetrical distribution around the pooled effect size, with no substantial asymmetry. Egger’s test (t = -0.61, df = 23, p = 0.55; bias = -0.50, SE = 0.82) indicated no significant small study effects. The residual heterogeneity variance was τ ² = 3.62, using standard error as the predictor and inverse-variance weights. Given the lack of asymmetry, a trim-and-fill analysis was not performed. These results suggest that the pooled estimates are unlikely to be affected by publication bias.

Assessment of reporting bias was informed by the “Selection of Reported Result” domain in ROBINS-E and several reporting items in STREGA. In ROBINS-E, most studies were rated as having a moderate risk for selective reporting, indicating that not all measured outcomes may have been reported. In the STREGA assessment, most studies did not report sample size justification or apply multiple testing correction, and Hardy-Weinberg Equilibrium (HWE) was tested in only a few. These findings suggest a moderate risk of bias due to missing or selectively reported results across the included studies.

The certainty of the evidence in this review was appraised qualitatively, as the GRADE approach was not applied due to the observational nature of the included genetic association studies. Instead, confidence in the findings was supported by several methodological safeguards: risk of bias was assessed using the ROBINS-E tool, guiding the inclusion of only low-risk studies in the meta-analysis; reporting quality was evaluated using the STREGA checklist to ensure transparency and reproducibility; and consistency, direction, and precision of effect estimates were examined across studies. Out of 46 studies, 25 were rated as having low risk of bias, 15 as moderate or some concerns, and 6 as high according to ROBINS-E. STREGA compliance ratings showed that 19 studies were rated as very good, 13 as fair, 9 as good, and 5 as excellent. Based on this distribution, the overall confidence in the evidence is moderate, reflecting generally sound methodology and reporting, though limitations in transparency and bias remain in a subset of studies.

Heterogeneity was assessed using I ² and τ ² statistics and explored further through subgroup and meta-regression analyses. Publication bias was evaluated through funnel plots and Egger’s regression test. Together, these assessments provide a moderate to high level of confidence in the observed association between ACE I/D genotypes and diabetic nephropathy, particularly the protective effect of the II genotype. However, the certainty of evidence for secondary outcomes, such as mortality and comorbidities, remains low due to limited data.