Heterogeneity in clinically diagnosed type 1 diabetes: characterising a unique cohort with maintained C-peptide secretion in Ghana

Abstract

Aims/hypothesis: In sub-Saharan Africa, type 1 diabetes is typically diagnosed clinically, which can be challenging due to atypical diabetes presentations such as ketosis-prone type 2 diabetes or type 2 diabetes in the absence of overweight and obesity. C-peptide, a marker of residual insulin secretion capacity, is crucial for understanding these variations but understudied in the region. Here, we investigated whether C-peptide measurement and concomitant genetic, autoimmune and metabolic characterisation of individuals with clinically diagnosed type 1 diabetes confirm diabetes classification and highlight population-specific features.

Methods: In this case-control study from Ghana, we recruited 266 individuals with clinically diagnosed and insulin-treated long-term type 1 diabetes and 266 healthy control individuals. We compared clinical features, HLA class II haplotypes, autoantibodies, and inflammatory and metabolic serum profiles across control and patient groups classified by random C-peptide levels: low (<0.2 nmol/l), mid (0.2-0.6 nmol/l) and high (>0.6 nmol/l).

Results: Only 28.9% of individuals with clinically diagnosed type 1 diabetes had low C-peptide concentrations. They were the youngest and leanest group, with higher frequencies of HLA class II risk haplotypes and GAD and ZnT8 autoantibodies compared with all other groups. By contrast, 34.6% and 36.5% had mid-range or high C-peptide levels, respectively. These subgroups resembled the control group in terms of low autoantibody titres and one protective HLA class II haplotype. Ketosis at onset was most prevalent in individuals with high C-peptide. Serum proinflammatory biomarkers differed between individuals with diabetes and control participants, but not between C-peptide subgroups. Aromatic and branched-chain amino acids varied between diabetes subgroups and positively correlated with C-peptide levels.

Conclusions/interpretation: Maintained C-peptide levels in two-thirds of individuals with long-term type 1 diabetes in Ghana, combined with the absence of autoantibodies and HLA risk association, highlight the necessity for better differentiation from atypical diabetes presentations to optimise patient care and improve health outcomes in resource-limited settings.

Keywords: Autoantibodies; Autoimmunity; HLA; Ketosis-prone type 2 diabetes; Serum biomarker; Type 1 diabetes.

Fig. 1

Clinical characteristics of the diabetes subgroups. (a) Distribution of individuals with diabetes into subgroups using C-peptide values (cutoffs 0.2 and 0.6 nmol/l) plotted against the duration of diabetes. Proportions and absolute numbers in diabetes subgroups (C-peptide HI, MID and LOW) are depicted as a pie chart. (b) Age at diabetes onset in diabetes subgroups is depicted. (c, d) Bar graphs are shown for (c) BMI classification into underweight, normal, overweight and obese participants and (d) presentation at diabetes onset (DKA, ketosis, hyperglycaemia, diagnosis by screening). Symbol plots are depicted with a median line. Differentiation by colour and shape plots is indicated: HI (yellow circles), MID (blue diamonds) and LOW (red triangles). The χ² test was applied to categorical parameters, and the Kruskal–Wallis test was used for continuous variables; p values below 0.05 were considered significant. *p<0.05; **p<0.01; ***p<0.001

Fig. 2

HLA haplotype distribution. The percentages of individuals carrying at least one risk-associated or protective HLA class II haplotype in the diabetes subgroups and control group (C) are represented by bar graphs. Significant differences in HLA haplotypes were calculated using the χ² test. Only comparisons with significant p values (below 0.05) are depicted. *p<0.05; **p<0.01; ***p<0.001

Fig. 3

Type 1 diabetes autoantibody positivity. The distribution of GAD, IA2 and ZnT8 autoantibodies in the healthy control group (C) and diabetes subgroups is displayed. (a) Quantitative antibody results are depicted as symbol plots with a median line. The dotted line represents the 99th percentile of the control population (n=100 healthy control individuals). Statistical significance was determined using the Kruskal–Wallis test followed by Dunn’s post hoc test. (b) Bar graphs represent the percentage of individuals testing positive for each autoantibody. (c) The proportions of individuals testing negative or positive for one or more autoantibodies. Aab, autoantibody. p values below 0.05 were considered significant. *p<0.05; **p<0.01; ***p<0.001

Fig. 4

Inflammation-related protein and amino acid profiling. Analysis of Olink inflammatory-related proteins (IRPs) and six branched-chain and aromatic amino acids in healthy control (C; n=58) and diabetes subgroups (n=118; HI n=59, MID n=20, LOW n=39) was performed. (a) Heatmap displaying the relative expression levels of IRPs across study groups. Colours represent z scores, with red indicating higher expression and blue indicating lower expression. Hierarchical clustering was performed on rows (study groups) and columns (proteins). (b) Volcano plot comparing protein expression between healthy control participants and individuals with diabetes combined. The x-axis represents the log₂ fold change, and the y-axis represents the −log₁₀ p value, with the threshold (unadjusted p value set at 0.05) indicated with dashed lines. Red dots indicate significantly upregulated proteins, and blue dots indicate significantly downregulated proteins in the diabetes group. Grey dots represent proteins with no significant change. Only proteins that are significantly differentially expressed above the significance threshold (adjusted p value) are named and marked. (c) Heatmap illustrating the relative levels of six amino acids across healthy control and diabetes subgroups. Colours represent z scores, with red indicating higher levels and blue indicating lower levels. Hierarchical clustering was performed on both rows (amino acids) and columns (study groups). (d) Symbol plots showing the expression levels of significant amino acids between control participants and all people with diabetes. Individual points are shown with the median as a straight line. (e) Symbol plots showing the expression levels of amino acids across the diabetes subgroups. Individual points are shown with the median as a straight line. (f) Scatter plots showing the correlation between each of the six amino acid metabolites and C-peptide levels across all individuals with diabetes. A solid line in each plot represents the linear regression fit. Spearman correlation coefficients (r) and p values are provided for each correlation. Multiple testing corrections of panels (a), (b) and (c) were applied using the Benjamini–Hochberg method. The Kruskal–Wallis test was performed for group comparisons in (e), and Dunn’s correction was applied for multiple comparisons. Statistically significant differences between groups are indicated. *p<0.05; **p<0.01; ***p<0.001