Increased risk of prediabetes among virally suppressed adults with HIV in Central Kenya detected using glycated haemoglobin and fasting blood glucose

Abstract Aims As survival among people living with HIV (PLHIV) improves with universal HIV treatment, new strategies are needed to support management of co‐morbidities like type 2 diabetes (T2D). We assessed prediabetes and T2D prevalence and risk factors using haemoglobin A1c (HbA1c) among PLHIV on antiretroviral therapy (ART) in Central Kenya. Methods This cross‐sectional study, conducted at a rural and urban site, enrolled PLHIV aged ≥35 years on ART for at least 5 years. HbA1c was assayed using Cobas b 101®, a point‐of‐care device. HbA1c levels ≥6.5% were considered diagnostic of T2D. For pre‐diabetic HbA1c levels (5.7%–6.4%), participants were requested to return the following day for a fasting blood glucose (FBG) to rule out T2D. Risk factors were assessed using multivariable log‐binomial regression. Results Of the 600 completing study procedures, the prevalence of diabetes was 5% (30/600). Ten participants were known to have diabetes; thus, prevalence of newly diagnosed T2D was 3.4% (20/590). Prevalence of prediabetes (HbA1c 5.7%–6.4%) was 14.2% (84/590). Significant predictors of elevated HbA1c were increase in age (Prevalence ratio [PR]: 1.10, CI: 1.02, 1.18, p = .012), hypertension (PR: 1.43, CI: 1.07–2.3, p = .015), central adiposity (PR: 2.11, CI: 1.57–2.84, p < .001) and use of Efavirenz (PR: 2.09, CI: 1.48, 2.96, p < .001). Conclusion There is a high prevalence of prediabetes, a significant predictor of T2D, among PLHIV in Central Kenya. Point‐of‐care HbA1c may help identify PLHIV with prediabetes in a single screening visit and provide an opportunity for early intervention.

of prediabetes, T2D and associated risk factors for hyperglycaemia among ART-experienced PLHIV in a rural and urban environment in Central Kenya. Prediabetes is important to identify as it has a better predictive value for developing T2D than individual risk factors including obesity and familial history. 5,6 Outside of Africa, the incidence of T2D among PLHIV is high. In South East Asia, it ranges from 3.4 to 11.1 per 100 person years. 7,8 However, findings on prevalence of T2D among PLHIV in sub-Saharan Africa (SSA) are mixed. While earlier studies found up to fourfold increased risk for T2D among PLHIV compared to individuals without HIV, [9][10][11] a review of recent studies in SSA reported T2D prevalence closer to twofold greater among people with HIV compared to adults without HIV, with prevalence ranging from 0.5%-9.3% among PLHIV and 0.5%-3.6% among individuals without HIV. 12 These conflicting findings could be due to use of different tests and criteria for diagnosis of T2D. Heterogeneity in ART use, for example including both ART-naïve and ART-experienced patients could also explain the mixed findings. Earlier studies also suggested that treatment with protease inhibitors (PI), high viral load and low CD4 count were predictors of T2D among PLHIV. 13 With universal ART roll-out, there is a need to identify persistent risk factors for T2D among PLHIV with immune reconstitution hence a high CD4 count and viral suppression and are using newer classes of ART drugs.
Optimal screening approaches for diabetes suitable for PLHIV in resource-limited settings are critical first steps in this process.
Glycated haemoglobin (HbA1c) captures the glycaemic status over a 3-month period and does not require fasting, therefore a feasible and acceptable screening tool in this population. 14 While older studies reported that compared to fasting blood glucose (FBG) or oral glucose tolerance test (OGTT), HbA1c underestimated T2D prevalence in PLHIV populations, these studies were conducted among adults with severe HIV-associated immunosuppression. 15,16 Recent studies that include PLHIV on ART have found similar test performance comparing HbA1c to other screening tests. 17,18 Availability of consistent and standardized point-of-care (POC) HbA1c devices has increased access to HbA1c testing and POC HbA1c could be a feasible screening tool in the Kenyan population. 19,20 With this in mind, we conducted a cross-sectional study using HbA1c screening and a confirmatory blood glucose to assess the burden of prediabetes, T2D and associated risk factors for hyperglycaemia among ART-experienced PLHIV in Central Kenya.

| Study subjects and setting
During a routine clinic visit, PLHIV aged ≥35 years were recruited from the HIV Comprehensive Care Centres (CCC) at Kiambu and Kerugoya, urban and rural referral hospitals, respectively, through systematic random sampling from January through July 2018.
Eligibility criteria included having been on ART for at least 5 years and having had a clinical review within the previous 6 months.
Patients with a history of blood transfusion or anaemia in the pre- HIV-specific data including the date of HIV diagnosis, current and previous ART regimens and blood pressure (BP) readings from the two previous clinic visits were extracted by the study nurse from participants' clinical charts and confirmed from the electronic medical records. Weight, height, waist and hip circumferences were measured according to WHO international standards. 22 A blood pressure reading was obtained with the participant seated and legs uncrossed, as part of routine clinic visit procedures.

| Glucose screening
The HbA1c assay was done using For HbA1c levels ≥6.5% (≥48 mmol/mol), a random blood glucose obtained on the same day was used to confirm T2D diagnosis [ Figure 1], as per the American Diabetes Association (ADA) guidelines where a diagnosis needs to be confirmed using a different test. For HbA1c levels between 5.7% and 6.4% (39-47 mmol/ mol), participants were requested to return while fasting the following day for a fasting blood glucose using Accu-Check ® by Roche©. Some patients who also needed a fasting lipid profile (details not presented here) also got a fasting blood glucose regardless of their HbA1c.

| Definitions of diabetes and hypertension
Hypertension was defined by two parameters: (1) two readings where systolic BP ≥140 mm Hg and/or diastolic BP ≥90 mm Hg, 24 using the enrolment BP measurement and a second BP reading from the clinical chart or (2) reported use of antihypertensive medications. Prediabetes was defined as a HbA1c between 5.7 and 6.4% (39-47 mmol/mol). Diabetes was defined by three parameters: (1) a HbA1c ≥6.5% and random blood glucose ≥11.1 mmol/L or (2) a HbA1c between 5.7 and 6.4% and fasting blood glucose ≥7.0 mmol/L or (3) reported use of insulin or oral hypoglycaemics. 25,26

| Statistical analysis
The sample size was based on an a priori estimated prevalence of diabetes among the PLHIV population of 16%-25%. 11,27 Using a normal approximation to the binomial distribution, we needed a sample size of 588 to observe an estimate of 25% prevalence of T2D with a precision interval of ±3.5%.
Due to lower-than-expected numbers of participants with diabetes in the sample, prediabetes and diabetes were combined to create a binary outcome for the regression analysis. Multivariable log-binomial regression was used to assess factors associated with prediabetes and diabetes. Covariates considered include age, sex, family history of T2D, time since HIV diagnosis, duration on ART, individual ART drugs, alcohol use, hypertension, body mass index (BMI) and waist circumference. Following bivariable analysis, covariates with a p-value of ≤.05 were included in the multivariable model.

| Cohort demographics and prevalence of prediabetes and T2D
Of the 600 participants who completed all study procedures, 383 (3.7%) had more than 1000 copies/ml, the cut-off for virologic treatment failure. A majority (93%) were on a combination of two nucleoside reverse transcriptase inhibitors (NRTI); tenofovir [TDF] or zidovudine [AZT], lamivudine 28 ), and a non-nucleoside reverse transcriptase inhibitor (NNRTI) efavirenz 28 or nevirapine. Only 7% were using a protease inhibitor. One hundred and thirty-six (45.6%) participants from the rural clinic were hypertensive compared to 79 (26.2%) from the urban clinic (p-value: <.001). Of note, among hypertensive patients on antihypertensives, none of their medication information was recorded in the electronic medical system.

| Factors associated with prediabetes and diabetes.
Factors associated with a diagnosis of prediabetes or diabetes are shown in Table 2  Among the 486 participants who had a normal HbA1c, 75 of them returned the following day for a fasting lipid profile and also got a fasting blood glucose; hence, a total of 165 people had fasting blood glucose [ Figure 2]. Using FBG as the outcome, the prevalence of hyperglycaemia (>5.6 mmol/L) was 44% (73/165) and the direction of association between the risk factors and elevated FBG remained the same, although statistical significance was lost due to the smaller sample size. (55%), provider-driven screening (e.g., as part of clinical work-up) was higher (39%) in the urban clinic compared to 31% in the rural clinic (p = .039).

| DISCUSS ION
After excluding those with known T2D, overall prevalence of newly diagnosed diabetes using HbA1c was low (3.4%) in this virally sup- have been reported. 7 These differences could be due to lower prevalence of other risk factors, like alcohol and smoking, in our study population relative to PLHIV in these regions or to other environmental or genetic differences. 30,39 The association of hyperglycaemia with Efavirenz is also consistent with what has previously been reported in the literature. 9,11 This is particularly significant in Kenya, where a greater proportion of patients are on TDF and EFV as their first-line regimen, yet diabetes screening is emphasized only among the small proportion on PIcontaining regimens. Dolutegravir, an integrase inhibitor which has been associated with weight gain and hyperglycaemia, 40,41 is being rolled out nationally to replace EFV as part of the first-line regimen.
Should the hyperglycaemic effect of both drugs be cumulative, a

| CON CLUS ION
In summary, we believe our findings add support to calls for universal HbA1c screening using POC screening methods as part of comprehensive HIV care services. Our cohort reflects the growing population of older PLHIV who have been on ART for long, have high CD4 counts and have achieved viral suppression. With time and without intervention, it is expected that a significant proportion of those with prediabetes will develop overt diabetes and ultimately its costly complications. In addition to counselling regarding positive lifestyle changes for these PLHIV, longitudinal follow-up of PLHIV with prediabetes will be important in understanding the risk of progression to T2D and to better inform screening and intervention strategies for diabetes among PLHIV in sub-Saharan Africa.

ACK N OWLED G EM ENTS
We would like to acknowledge all the patients and healthcare workers from Kiambu and Kerugoya County hospitals who partici-

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.