Reference interval for glycated hemoglobin (HbA1c) in non-pregnant, non-diabetic women in Ethiopians: a cross-sectional study

¹Department of Medical Physiology, Addis Ababa University, Addis Ababa, Ethiopia

^&Corresponding author
Kassahun Tekle Takiso, Department of Medical Physiology, Addis Ababa University, Addis Ababa, Ethiopia

Introduction    

Glycated hemoglobin (HbA1c) is a critical blood marker that reflects a person´s average blood glucose levels over the preceding two to three months. It is a cornerstone for diagnosing diabetes mellitus and monitoring long-term glycemic control in patients [1,2]. To interpret any diagnostic test like HbA1c, clinicians rely on a reference interval (RI) - the range of values found in a healthy population. This interval is a fundamental tool for distinguishing healthy from disease [3]. International guidelines, such as those from the International Federation of Clinical Chemistry (IFCC), strongly recommend that each laboratory or region establish its own RIs tailored to its specific population [4,5]. This is because RIs are not universal; they can vary significantly due to differences in genetics, ethnicity, diet, race, age, and environmental factors [6-10].

Despite this recommendation, most clinical laboratories find it challenging to establish local RIs due to the high cost, the complexity of recruiting a sufficient number of healthy reference individuals, and the extensive testing required [11]. Consequently, they often rely on manufacturer-based reference intervals (MBRIs), which are convenient but have major limitations. MBRIs may lack generalizability and can be inaccurate for populations not represented in the manufacturer´s original reference study. In contrast, laboratory-based reference intervals (LBRIs), though resource-intensive to develop, provide higher-quality, population-specific data that lead to more accurate diagnosis and better clinical decisions [12,13].

The need for population-specific RIs is particularly evident for HbA1c. Numerous studies have identified ethnic and racial variation in HbA1c levels. For example, at similar blood glucose levels, non-Hispanic Black individuals have been shown to have higher HbA1c levels than non-Hispanic White individuals [14]. Similarly, New Zealand uses a different HbA1c threshold for diagnosing gestational diabetes for women of Asian ethnicity compared to European ethnicity [15], underscoring that a single standard is not appropriate for all. In Ethiopia, there is a lack of locally established reference intervals for HbA1c. Current clinical practice relies on intervals derived from Western populations, which may not be applicable to Ethiopians. This gap increases the risk of misdiagnosis, including both overdiagnosis, leading to unnecessary treatment, and underdiagnosis, resulting in a missed opportunity for timely intervention. Therefore, this study aimed to establish a robust, country-specific reference interval for HbA1c in healthy, non-pregnant, non-diabetic Ethiopian women and to investigate the correlation between HbA1c levels and demographic and clinical variables.

Methods     

Study design, setting, and period: an institution-and laboratory-based cross-sectional study was conducted from December 4, 2024, to February 28, 2025, at Janmeda and Teklehaymanot Health Centers in Addis Ababa, Ethiopia. A total of 129 non-pregnant, non-diabetic women aged 18 to 48 years were included in the study. Participant recruitment and sampling were conducted at these two government health centers.

Ethical consideration: the study protocol was reviewed and approved by the Institutional Research Ethics Review Committee (IRERC) of the College of Health Sciences, Addis Ababa University (protocol number 014/24/physio). The study was conducted in strict accordance with the ethical principles of the Declaration of Helsinki [16]. Before any data collection, written informed consent was obtained from every participant after the nature and purpose of the study had been fully explained. To ensure confidentiality, all participant data were de-identified using unique study codes. Completed questionnaires and consent forms were stored separately in a locked cabinet, and access to all study materials was restricted to the principal investigator and supervising authors.

Study population: the source population comprised non-pregnant, non-diabetic women aged 18 years and above attending routine medical cheek-ups at the selected health centers in Addis Ababa, Ethiopia.

Sampling technique: participants were recruited using a systematic random sampling technique. The sampling frame was obtained from the health center records. A random starting point was chosen by drawing a number between 1 and 9, and every k-th woman was subsequently invited until the required sample size (minimum 120) was achieved, ensuring adequate statistical power for a non-parametric test.

Sample size: the sample size was determined based on recommendations from the International Federation of Clinical Chemistry (IFCC), which advises a minimum of 120 individuals for establishing reference intervals using non-parametric methods [17,18]. Our study enrolled 129 participants, thereby meeting and exceeding this requirement. Parametric methods were employed following data normality; the sample size is also more than sufficient to invoke the Central Limit Theorem. This theorem ensures the robustness of parametric test with large sample size (a common threshold being *n* ≥ 30), further validating our analytical approach [19,20].

Data collection and personnel

Data collection team: data were collected by four trained medical professionals-two midwife nurses and two medical laboratory technologists- who were employed at the respective health centers. The entire data collection process was supervised daily by the study authors.

Training of data collectors: the data collectors underwent a one-day training session on standardized study procedures. The training covered protocols for participant recruitment based on inclusion criteria, obtaining informed consent, and assigning a unique identification number (ID) to each participant. It also included instructions on administering the questionnaire and ensuring all responses were correctly linked to the participant´s ID. Additionally, the training addressed laboratory procedures, including collecting blood samples into EDTA and SST tubes labeled with the participant IDs, performing immediate point-of-care hemoglobin and random blood sugar tests, and preparing EDTA samples for daily transport to the Black Lion Hospital Laboratory.

Data collection

Participant recruitment: the study enrolled non-pregnant, non-diabetic women aged ≥ 18 years who attended routine medical check-ups at the selected health centers. Eligible participants met the following inclusion criteria: non-pregnant women (confirmed by a negative urine hCG rapid test), age ≥ 18 years, with a random blood glucose (RBS) level < 100 mg/dL and hemoglobin concentration >11 g/dL, in accordance with WHO Standards [21]. Women were excluded if they had a history of diabetes, alcoholism, pregnancy, anemia, or any chronic systemic disease (e.g., hypertension, splenectomy, or hemolytic, hepatic, immunological, renal, or cardiac disease).

Data collection procedures

Sociodemographic and clinical characteristics: after obtaining informed consent, data were collected using a semi-structured questionnaire adapted from similar sources [22-24]. The questionnaire was translated into Amharic via a forward-backward translation process and pilot-tested for clarity and cultural appropriateness. Minor modifications were made to phrasing based on this feedback. The final tool collected data on socio-demographics, clinical characteristics, lifestyle, and health-related factors.

Laboratory procedures: biological samples were collected from each participant. A total of 5 mL of whole blood was drawn: 3 mL was dispensed into a K2EDTA tube for hemoglobin and HbA1c measurement, and 2 mL into a serum separate tube (SST) for random blood glucose measurement and serological testing. In addition, approximately 4 mL of urine was collected in a sterile container for pregnancy testing. After phlebotomy, standardized protocols were followed to ensure sample integrity. At the health centers, EDTA whole blood samples were used immediately for hemoglobin and blood group analysis within 40 minutes of collection. Concurrently, SST blood samples were allowed to clot at room temperature for 30 minutes, and then centrifuged at 3000 x g for 10 minutes. The separated serum was immediately analyzed on-site for RBS and other serological tests, while urine hCG testing was also performed on-site. The remaining EDTA whole blood samples were transported to the Black Lion Hospital Laboratory for HbA1c analysis. Black Lion Hospital, the largest national referral hospital in Ethiopia, also serves as a primary training institution for health professionals [25]. HbA1c measurement was performed within 4 hours of collection, ensuring the entire process was completed within a strict 4-8 hour window.

Biochemical analysis

All biochemical analyses were performed following standard operating protocols: HbA1c was quantified using a cobas C 311 analyzer (Roche Diagnostics, Switzerland) via a turbidimetric inhibition immunoassay (TINIA). The analyzer was calibrated according to the IFCC reference system, and two levels of quality controls (Precicontrol Norm and Path) were run daily to ensure accuracy. Plasma glucose was measured using a BS-200 clinical chemistry analyzer (Mindray, China) with the hexokinase method. Two levels of controls (low and high) were analyzed prior to sample testing. Hemoglobin concentration was determined using a Sysmex XS-500i hematology analyzer (Sysmex Corporation, Japan) employing the modified, cyanide-free, sodium lauryl sulfate (SLS)-hemoglobin method.

Statistical analysis: data were analyzed using SPSS version 27 (IBM Corp., USA). Potential outliers in all continuous variables were identified and excluded using Tukey´s statistical methods [26]. The normality of continuous variables, including HbA1c, was assessed using the Shapiro-Wilk test and complemented by visual inspection of histograms and Q-Q plots. For variables that satisfied the assumption of normality, reference intervals were calculated parametrically as the mean ± 2 x standard deviation (SD). Continuous variables that did not follow a normal distribution were summarized using the median with range (minimum-maximum). The relationship between HbA1c and demographic and clinical characteristics was assessed using Pearson´s correlation coefficient, selected because: (1) HbA1c demonstrated normal distribution after outlier removal, and (2) Pearson´s correlation is robust to mild deviations from normality with adequate sample size.

Results     

Out of 210 non-pregnant women assessed, 174 were found eligible, and 129 met the inclusion criteria, while 45 were excluded for various reasons: 21 had random plasma sugar levels exceeding 100mg/dl, 4 were diagnosed with anemia, 5 tested positive for hCG, 4 were positive for HBsAg, 1 was positive for HCV, and 5 had other medical conditions. The distribution of HbA1c values was assessed before calculating the reference interval. The Shapiro-Wilk test indicated that HbA1c closely followed a normal distribution (W = 0.981, P = 0.071). Visual inspection of the normal Q-Q plot (Figure 1) further supported this finding, as the observed data points aligned closely with the expected diagonal reference line. Based on the conformation of normality, the reference interval for HbA1c was estimated using the parametric method (mean ± 2 x SD). Continuous variables that did not meet normality assumptions were summarized using the median with range (minimum-maximum).

Demographic and clinical characteristics: the baseline demographic and clinical characteristics of the 129 non-pregnant women included in the study are summarized in Table 1. Participants were young to middle-aged women, with a median age of 30 years (range: 18-48 years) and a median body mass index (BMI) of 20.3kg/m² (range: 18.7 - 25.6kg/m²). The median hemoglobin (Hb) level was 14.6 g/dL (range: 12-16.4 g/dL). Median systolic and diastolic blood pressures were 116 mmHg (range: 90 -135) and 78 mmHg (range: 70 - 88 mmHg), respectively. The median Random Plasma Glucose (RPG) level was 5.1mmol/L (range: 4.3- 5.5mmol/L). Blood group “O” was the most prevalent (41.9%), followed by groups “A” (30.2%) and “B” (23.3%). All participants tested negative for hCG, hepatitis B, and hepatitis C, and none reported a family history of diabetes mellitus (DM) or gestational diabetes mellitus (GDM).

Reference intervals for HbA1c: the reference intervals for HbA1c were established in a cohort of 129, non-diabetic non-pregnant women. The 95% reference interval was 32-44 mmol/mol (IFCC standard) and 5.1-6.2% (DCCT standard). The mean HbA1c values were 38 (SD 3.1) mmol/mol and 5.6 (SD 0.28)%, respectively, both centrally located within the calculated reference intervals (Table 2).

Correlation of HbA1c with demographic and clinical variables: the relationship between HbA1c (normally distributed) and key demographic and clinical variables was assessed using Pearson´s Correlation Coefficient. Results are presented in Table 3. Age showed a statistically significant, weak positive correlation with HbA1c (r = 0.285, p = 0.007). No significant correlations were observed between HbA1c and random blood glucose (r = 0.010, p =0.914) or weight (r = 0.048, p = 0.588)

Discussion     

Age, ethnicity, genetic background, and geographical location significantly influence HbA1c values, as highlighted by numerous studies. As no HbA1c reference intervals had previously been developed in Ethiopia, this study aims to establish a reference interval for HbA1c levels in non-diabetic, non-pregnant women in Ethiopia. Currently, clinical practices in Ethiopia rely on manufacturer-provided reference intervals, which may not accurately reflect the unique characteristics of the local population. Developing a population-specific reference interval can enhance the precision of diabetic management in Ethiopia.

In this study, the reference interval for HbA1c in non-diabetic, non-pregnant women was 5.1% to 6.2%, with a mean of 5.6%. HbA1c values were positively associated with age. While the lower limit of HbA1c value (5.1%) was higher than in some previous studies, the upper limit (6.2%) was aligned with findings from Italy (4.8% to 6.2%) [27], China (4.7% to 6.3%) [28], and Nordic Caucasians (4.1% to 6.4%) [29]. The upper limit in the current study is also consistent with the value (≤ 6.5%) established by the American Diabetes Association (ADA), which further considers a value below 6.0% to be optimal [30]. A study conducted at Althagelvin Area Hospital in Londonderry reported an upper limit of 6.5% [31], which is nearly similar to our result. The mean HbA1c value in our study (5.6%) was also comparable to a study conducted in Sudan, which reported a mean of 5.57% among non-pregnant women [32].

However, the reference interval reported in Sudan was wider (4.7% to 6.7%) [33] than in our study (5.1% to 6.2%). Further studies are required to determine the reason for such variations, as differences in dietary habits, genetic predisposition, environmental influences, and other conditions could contribute to inconsistent findings. In contrast, the mean HbA1c value (5.6%) was higher than in another study reported (3.43%) [34]. This represents a striking contrast, highlighting potential differences in glycemic control and overall metabolic health between these populations. The reference intervals in the current study (5.1% to 6.2%), with the lower limit of 5.1% this value were also higher than in the study conducted in South Asia (4% to 5.7%), Pakistan (3.51% to 4.87%), and Japan (3.51% to 4.87%) [23,35,36]. On a similar note, a study conducted among non-diabetic, non-pregnant women in the United States noted an HbA1c value of 5.26±0.44% [37].

Consistent with the previous studies [38-41], we observed a positive correlation between age and HbA1c values. This indicates that advancing age is associated with higher HbA1c levels, possibly due to physiological changes, alterations in glucose tolerance, red blood cell lifespan, metabolic control, and insulin resistance. However, other studies did not find a positive relationships [34,42], suggesting the need for further research to clarify these discripancies.

Limitations: while this study provides robust reference intervals derived from a well-characterized cohort, the sampling framework-restricted to a single urban center-may limit the generalizability of our findings to Ethiopia´s diverse populations or analogous geographical settings. Regional variation in demographic, nutritional, or environmental factors could influence the extrapolation of these reference values.

Conclusion     

This study establishes a locally derived reference interval for HbA1c among non-diabetic, non-pregnant Ethiopian women in Addis Ababa, determined to be 5.1%-6.2%. Notably, the lower limit of this interval is higher than the 5.0-5.1% commonly cited international guidelines, highlighting that global standards may not be directly applicable to this population. We found no significant correlation between HbA1c and weight, or random blood sugar (RBS). In contrast, age showed a weak but significant positive association with HbA1c, suggesting it may need consideration in clinical assessment. Overall, these findings underscore the clinical importance of using population-specific HbA1c reference intervals for the accurate diagnosis and monitoring of dysglycemia in Ethiopia.

Recommendation: based on the findings of this study, we recommend the following: 1) Clinical adoption: Implement the established HbA1c reference interval (5.1% - 6.2%) in Ethiopia clinical practice to improve the accuracy of dysglycemia diagnosis in non-pregnant women. 2) Guideline update: Incorporate this locally derived reference interval into national diabetes management guidelines and laboratory standards. 3) Future research: Validate and expand this work across other Ethiopian regions and demographic groups, including men, pregnant women, and different age cohorts.

Competing interests     

The authors declare no competing interests.

Authors' contributions     

Kassahun Tekle Takiso: conceptualization, methodology, investigation, data curation, formal analysis, supervision, and the writing of the original draft, as well as review and editing. Abebeye Aragaw Lemine: writing, revising, and editing the manuscript, project administration and supervision. Aster Abebe Tsegaye: data analysis and statistical evaluation, verifying the accuracy and reliability of the research findings. Mekoya Mengistu Dabulo: methodology and manuscript revision. Abenet Desalegn W/Senbet: methodology and investigation, data collection, data interpretation. All the authors have read and agreed to the final manuscript.

Acknowledgments     

We would like to express our sincere gratitude to Mrs. Laura Goosen Miler for her generous support in providing the cobas c 311 reagents, calibrator, and control, which greatly facilitated this research despite geographical constraints.

Tables and figure     

Table 1: baseline demographic and clinical characteristics of non-pregnant women recruited at Janmeda and Teklehaymanot Health Centers, Addis Ababa, Ethiopia from December 4, 2024 to February 28, 2025 (n=129)

Table 2: reference intervals and mean values for HbA1c among study participants, recruited at Janmeda and Teklehaymanot Health Centers, Addis Ababa, Ethiopia from December 4, 2024 to February 28, 2025 (n=129)

Table 3: person correlation analysis of HbA1c with demographic and clinical variables among study participants, recruited at Janmeda and Teklehaymanot Health Centers, Addis Ababa, Ethiopia from December 4, 2024 to February 28, 2025 (n=129)

Figure 1: quantile-quantile (Q-Q) plot of HbA1c values in the study population of Ethiopian women; findings from the study conducted at Janmeda and Teklehaymanot Health Centers, Addis Ababa, Ethiopia, 2025

References