Undiagnosed G6PD Deficiency in Black and Asian Individuals Is Prevalent and Contributes to Health Inequalities in Type 2 Diabetes Diagnosis and Complications

Abstract

Objective: Glucose-6-phosphate dehydrogenase (G6PD) deficiency presents silently and is not routinely screened. It is associated with markedly lower HbA1c for the prevailing glucose levels. Since HbA1c is internationally recommended to diagnose and manage type 2 diabetes (T2D), we investigated the population-level impact of undiagnosed G6PD deficiency on T2D diagnosis and complications in the U.K.

Research design and methods: We used whole-exome sequencing and electronic health record data from UK Biobank (n = 467,368) and Genes & Health (n = 43,011) cohorts.

Results: In the U.K., we estimated that ∼1 in 7 Black and 1 in 63 Asian males carry G6PD deficiency alleles, compared with fewer than 1 in 10,000 White males. Despite this, less than 1 in 50 G6PD-deficient men are clinically recognized. Male G6PD carriers have considerably lower average HbA1c (0.9% [International Federation of Clinical Chemistry and Laboratory Medicine: 10.0 mmol/mol]) compared with noncarriers, while differences in average glucose were negligible. G6PD-deficient men had 1.37 (95% CI: 1.01, 1.86) higher odds of developing diabetes-related microvascular complications than noncarriers. Although risk factors were similar prior to diagnosis, male G6PD carriers diagnosed with T2D since 2011 were, on average, 4.1 years (95% CI: 0.6, 7.7) older at diagnosis compared with noncarriers. In addition, lower mean HbA1c values in G6PD carriers falsely underestimated their 10-year T2D risk.

Conclusions: Undiagnosed G6PD deficiency has significant impact on T2D diagnosis with HbA1c and associates with increased risk of diabetes complications. This has major implications for global populations using HbA1c for diagnosis and monitoring, and could contribute significantly to inequalities in diabetes outcomes.

Figure 1

Violin and box plots for HbA1c and random glucose levels by carrier status for the G6PD deficiency (a) rs1050828-T allele in Black male UK Biobank participants; (b) rs5030868-A allele in Asian male UK Biobank participants; and (c) rs5030868-A allele in South Asian male Genes & Health participants; all without diabetes. Effect sizes represent Cliff’s Delta estimates, and p-values are taken from Mann-Whitney-U tests; box plots show the median and interquartile range, and the whiskers show minimum and maximum; plots are truncated at median + 5 × (median absolute difference) with characteristics of truncated values given. ES: effect size, N: sample size.

Figure 1

Violin and box plots for HbA1c and random glucose levels by carrier status for the G6PD deficiency (a) rs1050828-T allele in Black male UK Biobank participants; (b) rs5030868-A allele in Asian male UK Biobank participants; and (c) rs5030868-A allele in South Asian male Genes & Health participants; all without diabetes. Effect sizes represent Cliff’s Delta estimates, and p-values are taken from Mann-Whitney-U tests; box plots show the median and interquartile range, and the whiskers show minimum and maximum; plots are truncated at median + 5 × (median absolute difference) with characteristics of truncated values given. ES: effect size, N: sample size.

Figure 1

Violin and box plots for HbA1c and random glucose levels by carrier status for the G6PD deficiency (a) rs1050828-T allele in Black male UK Biobank participants; (b) rs5030868-A allele in Asian male UK Biobank participants; and (c) rs5030868-A allele in South Asian male Genes & Health participants; all without diabetes. Effect sizes represent Cliff’s Delta estimates, and p-values are taken from Mann-Whitney-U tests; box plots show the median and interquartile range, and the whiskers show minimum and maximum; plots are truncated at median + 5 × (median absolute difference) with characteristics of truncated values given. ES: effect size, N: sample size.

Figure 2

(a) Forest plot and random-effects meta-analysis of odds of developing microvascular complications in hemizygote vs non-carrier males diagnosed with type 2 diabetes at any time in the UK Biobank and Genes & Health ethnic groups. Nephropathy (N_NM=39/1,519; N_H=0/20) results for UK Biobank Asian group not included as model could not be fitted; sample sizes under 5 are hidden according to cohort guidelines. UKB: UK Biobank, G&H: Genes & Health, CI: confidence interval, N_NM: no mutations sample size, N_H: hemizygote sample size. (B) Forest plot and random-effects meta-analysis of odds of developing macrovascular complications in hemizygote vs non-carrier males diagnosed with type 2 diabetes at any time in the UK Biobank and Genes & Health ethnic groups. Cardiovascular death (N_NM=110/1,519; N_H=0/20) results for UK Biobank Asian group not included as model could not be fitted. (C) Violin and box plots for age of type 2 diabetes diagnosis by carrier status of the Asahi G6PD deficiency rs1050828-T allele in Black male UK Biobank participants. Effect size represents Cliff’s Delta estimate, and p-value is taken from Mann-Whitney-U test; the mean (SD) age at type 2 diabetes diagnosis and difference of means (95% CI) are shown in the table; box plots show the median and interquartile range, and the whiskers show minimum and maximum. ES: effect size, SD: standard deviation. (D) Forest plot and random-effects meta-analysis of adjusted difference in age of type 2 diabetes diagnosis between G6PD deficiency allele carriers versus non-carriers in males diagnosed after 2011 in the UK Biobank and Genes & Health ethnic groups. Adjusted difference in age estimates correspond to the coefficients from linear regression models adjusting for self-reported sub-ethnicity and electronic health record source of age. T2D: type 2 diabetes.

Figure 2

(a) Forest plot and random-effects meta-analysis of odds of developing microvascular complications in hemizygote vs non-carrier males diagnosed with type 2 diabetes at any time in the UK Biobank and Genes & Health ethnic groups. Nephropathy (N_NM=39/1,519; N_H=0/20) results for UK Biobank Asian group not included as model could not be fitted; sample sizes under 5 are hidden according to cohort guidelines. UKB: UK Biobank, G&H: Genes & Health, CI: confidence interval, N_NM: no mutations sample size, N_H: hemizygote sample size. (B) Forest plot and random-effects meta-analysis of odds of developing macrovascular complications in hemizygote vs non-carrier males diagnosed with type 2 diabetes at any time in the UK Biobank and Genes & Health ethnic groups. Cardiovascular death (N_NM=110/1,519; N_H=0/20) results for UK Biobank Asian group not included as model could not be fitted. (C) Violin and box plots for age of type 2 diabetes diagnosis by carrier status of the Asahi G6PD deficiency rs1050828-T allele in Black male UK Biobank participants. Effect size represents Cliff’s Delta estimate, and p-value is taken from Mann-Whitney-U test; the mean (SD) age at type 2 diabetes diagnosis and difference of means (95% CI) are shown in the table; box plots show the median and interquartile range, and the whiskers show minimum and maximum. ES: effect size, SD: standard deviation. (D) Forest plot and random-effects meta-analysis of adjusted difference in age of type 2 diabetes diagnosis between G6PD deficiency allele carriers versus non-carriers in males diagnosed after 2011 in the UK Biobank and Genes & Health ethnic groups. Adjusted difference in age estimates correspond to the coefficients from linear regression models adjusting for self-reported sub-ethnicity and electronic health record source of age. T2D: type 2 diabetes.

Figure 2

(a) Forest plot and random-effects meta-analysis of odds of developing microvascular complications in hemizygote vs non-carrier males diagnosed with type 2 diabetes at any time in the UK Biobank and Genes & Health ethnic groups. Nephropathy (N_NM=39/1,519; N_H=0/20) results for UK Biobank Asian group not included as model could not be fitted; sample sizes under 5 are hidden according to cohort guidelines. UKB: UK Biobank, G&H: Genes & Health, CI: confidence interval, N_NM: no mutations sample size, N_H: hemizygote sample size. (B) Forest plot and random-effects meta-analysis of odds of developing macrovascular complications in hemizygote vs non-carrier males diagnosed with type 2 diabetes at any time in the UK Biobank and Genes & Health ethnic groups. Cardiovascular death (N_NM=110/1,519; N_H=0/20) results for UK Biobank Asian group not included as model could not be fitted. (C) Violin and box plots for age of type 2 diabetes diagnosis by carrier status of the Asahi G6PD deficiency rs1050828-T allele in Black male UK Biobank participants. Effect size represents Cliff’s Delta estimate, and p-value is taken from Mann-Whitney-U test; the mean (SD) age at type 2 diabetes diagnosis and difference of means (95% CI) are shown in the table; box plots show the median and interquartile range, and the whiskers show minimum and maximum. ES: effect size, SD: standard deviation. (D) Forest plot and random-effects meta-analysis of adjusted difference in age of type 2 diabetes diagnosis between G6PD deficiency allele carriers versus non-carriers in males diagnosed after 2011 in the UK Biobank and Genes & Health ethnic groups. Adjusted difference in age estimates correspond to the coefficients from linear regression models adjusting for self-reported sub-ethnicity and electronic health record source of age. T2D: type 2 diabetes.

Figure 2

(a) Forest plot and random-effects meta-analysis of odds of developing microvascular complications in hemizygote vs non-carrier males diagnosed with type 2 diabetes at any time in the UK Biobank and Genes & Health ethnic groups. Nephropathy (N_NM=39/1,519; N_H=0/20) results for UK Biobank Asian group not included as model could not be fitted; sample sizes under 5 are hidden according to cohort guidelines. UKB: UK Biobank, G&H: Genes & Health, CI: confidence interval, N_NM: no mutations sample size, N_H: hemizygote sample size. (B) Forest plot and random-effects meta-analysis of odds of developing macrovascular complications in hemizygote vs non-carrier males diagnosed with type 2 diabetes at any time in the UK Biobank and Genes & Health ethnic groups. Cardiovascular death (N_NM=110/1,519; N_H=0/20) results for UK Biobank Asian group not included as model could not be fitted. (C) Violin and box plots for age of type 2 diabetes diagnosis by carrier status of the Asahi G6PD deficiency rs1050828-T allele in Black male UK Biobank participants. Effect size represents Cliff’s Delta estimate, and p-value is taken from Mann-Whitney-U test; the mean (SD) age at type 2 diabetes diagnosis and difference of means (95% CI) are shown in the table; box plots show the median and interquartile range, and the whiskers show minimum and maximum. ES: effect size, SD: standard deviation. (D) Forest plot and random-effects meta-analysis of adjusted difference in age of type 2 diabetes diagnosis between G6PD deficiency allele carriers versus non-carriers in males diagnosed after 2011 in the UK Biobank and Genes & Health ethnic groups. Adjusted difference in age estimates correspond to the coefficients from linear regression models adjusting for self-reported sub-ethnicity and electronic health record source of age. T2D: type 2 diabetes.

Figure 3

(a) 10-year risk of diabetes according to QDiabetes-2018 score C for example individual with varying HbA1c levels. Individuals 1 and 2 represent example 40-year-olds at NHS Health Check appointment who share the same risk variables for QDiabetes-2018 score A, but differ in their HbA1c levels to reflect the mean HbA1c for male UK Biobank participants of Black ethnicity with no copies of the G6PD deficiency rs1050828-T allele (Individual 1) and hemizygote carriers (Individual 2). Solid grey line represents the QDiabetes-2018 score A; orange line – mean HbA1c for male UK Biobank participants of Black ethnicity matched with Individual 1 (matching based on risk variables used in QDiabetes-2018 score A) and hemizygote carriers (N=6) for the G6PD deficiency rs1050828-T allele; blue line – same for participants matched with Individual 2 and with no copies of the G6PD deficiency rs1050828-T allele (N=24); shading – 95% confidence intervals; dashed line – diagnostic threshold for prediabetes. (b) 10-year risk of diabetes according to QDiabetes-2018 scores A and C for male UK Biobank participants of Black ethnicity with high-risk QDiabetes-2018 score A (>=5.6%) diagnosed with type 2 diabetes after baseline centre visit. Carrier and non-carrier participants have been matched into four groups based on their risk variables used in QDiabetes-2018 score A (from low to very high risk). Points represent the mean risk, and error bars the 95% confidence interval.

Figure 3

(a) 10-year risk of diabetes according to QDiabetes-2018 score C for example individual with varying HbA1c levels. Individuals 1 and 2 represent example 40-year-olds at NHS Health Check appointment who share the same risk variables for QDiabetes-2018 score A, but differ in their HbA1c levels to reflect the mean HbA1c for male UK Biobank participants of Black ethnicity with no copies of the G6PD deficiency rs1050828-T allele (Individual 1) and hemizygote carriers (Individual 2). Solid grey line represents the QDiabetes-2018 score A; orange line – mean HbA1c for male UK Biobank participants of Black ethnicity matched with Individual 1 (matching based on risk variables used in QDiabetes-2018 score A) and hemizygote carriers (N=6) for the G6PD deficiency rs1050828-T allele; blue line – same for participants matched with Individual 2 and with no copies of the G6PD deficiency rs1050828-T allele (N=24); shading – 95% confidence intervals; dashed line – diagnostic threshold for prediabetes. (b) 10-year risk of diabetes according to QDiabetes-2018 scores A and C for male UK Biobank participants of Black ethnicity with high-risk QDiabetes-2018 score A (>=5.6%) diagnosed with type 2 diabetes after baseline centre visit. Carrier and non-carrier participants have been matched into four groups based on their risk variables used in QDiabetes-2018 score A (from low to very high risk). Points represent the mean risk, and error bars the 95% confidence interval.