We here provide evidence that there is a difference in the relation between circulating FPG level and HbA1c with age and in younger women vs younger men, with implications for the diabetes diagnostic threshold. Specifically, we identified, using matched HbA1c and FPG data from two cohorts of primary care non-diabetes patients, that (a) both HbA1c and FPG increase with age, (b) FPG is consistently lower in women than men across all age groups, and (c) HbA1c is lower in women aged < 45 years than men aged < 45 years but not in the 45–55 and ≥ 55 years age groups.

The key message from this study is that our previously identified difference in HbA1c levels in younger women compared to younger men, but not older women and men [1], is not reflected in the FPG levels taken at the same time. This adds weight to our original argument that this may reflect non-physiological, not pathological, factors such as menstruation in pre-menopausal women. While some may argue that this is self-evident and has indeed been identified elsewhere [16,17,18,19,20], it raises important questions regarding the diagnostic thresholds used for both FPG and, particularly, HbA1c that have hitherto not been addressed.

At face value, these findings might suggest that age- and sex-specific reference ranges should be used for the diagnosis (and potential monitoring) of diabetes mellitus. However, it is important to differentiate age- and sex-related differences in HbA1c and FPG levels arising from the interaction between physiological (e.g. differences between the sexes in glucose metabolism and its regulation; including the effect of pregnancy), pathological (e.g. decreasing pancreatic function with age) and chemical factors (e.g. factors influencing erythrocyte turnover and hence haemoglobin glycation levels).

Changes in HbA1c and FPG with Age

The increasing concentrations of FPG and HbA1c with age have been previously documented [21,22,23] and are suggested to reflect decreasing tissue insulin sensitivity and reduced insulin secretion by the pancreas with age [24]. This increase is reflected in the increasing prevalence of diabetes with age [25]. While some have suggested that age-dependent reference ranges should be used [26], we would suggest that this would not be warranted as it is more likely to reflect age-related deterioration of β-cell function, potentially associated with changes in visceral and subcutaneous fat [27, 28], though the molecular mechanisms behind these changes are complex and not fully understood [29]. While it is true that there are non-physiological changes in HbA1c with age, including changes in erythrocyte turnover [26], leading to calls from some to introduce age-specific reference ranges, we propose that age-related differences are likely to reflect important pathological changes and therefore do not warrant age-specific reference ranges.

Differences in FPG Between Men and Women

The finding of lower FPG levels in women than men across all age groups is well known and reflects complex sex-related differences in glucose homeostasis, including interrelated factors such as oestrogen levels, insulin sensitivity, muscle and adipose tissue mass [30]. It is well established that oestrogen acts as an insulin-sensitising agent [31, 32]. It might be expected therefore that this relation between FPG in men and women might change at the menopause. Our data illustrate that this is not the case, with the lower FPG levels being consistently 0.1–0.2 mmol/L lower in women across all ages. It is likely that this consistent difference is a reflection of both physiological and pathological effects, so our view is that there is currently insufficient evidence to recommend sex-specific reference ranges in FPG.

Differences in HbA1c Between Men and Women

Our findings on the age- and sex-related changes in HbA1c are consistent with observations from our previous work and those of other groups, who reported that HbA1c was lower in pre-menopausal women than men of the same age [17,18,19,20]. We and others have proposed that these observations may reflect the effect of menstruation in younger women, leading to a shorter mean erythrocyte half-life and hence shorter exposure of haemoglobin to circulating glucose compared with individuals who do not menstruate [33]. The nature of this study as a retrospective laboratory-data-based project precludes any further investigation at this stage in relation to studies of red cell turnover. However, we would argue that the lack of a similar pattern with FPG suggests that this observation is less likely to be due to sex- and age-specific physiological changes such as the effect of oestrogen in the pre-menopausal period [20]. The change in the ratio between HbA1c and FPG with age in women makes this contrast between FPG and HbA1c patterns particularly evident (Fig. 1c). We contest, therefore, that there is a case for sex-specific reference ranges for HbA1c in pre-menopausal women.

Implications of Sex-Specific Reference Ranges for HbA1c in Pre-menopausal Women

Our observations raise the question as to whether the ‘set point’ for HbA1c may be lower in women than in men with the result that female levels of HbA1c that would be considered ‘normal’ for men are actually a reflection of developing metabolic stress in women. These differences between men and women, together with those due to culture, education, lifestyle and ethnicity, will influence the progression to overt diabetes over time.

Given that the diagnosis of T2D is now largely based on HbA1c, the lack of use of sex-specific reference ranges and a diagnostic cut point for type 2 diabetes for pre-menopausal women could potentially lead to the underdiagnosis of T2D in younger women and missed opportunities for intervention and timely cardiovascular risk reduction intervention [7, 9, 10], which is of relevance given the higher mortality rates in women vs men with T2D [8]. This might contribute to our understanding of why women with diabetes are diagnosed later in life [2] and have a higher mortality rate than men with T2D [8, 34].

We accept that while the reported differences in HbA1c are small, at a population level in relation to screening, many thousands of women would still be affected by a change in the threshold for the diagnosis of T2D.

It might be argued that the 1 mmol/mol difference in HbA1c level in younger women is small and would not warrant such a major change in policy regarding diabetes diagnosis and monitoring. However, as we discussed in our previous paper, the implications at a population level of not implementing this change are significant [1].

Limitations

Limitations of our study include that it only involved two UK laboratories. Nevertheless, a large number of parallel samples in relation to paired HbA1c and FPG collected on the same day were analysed. Furthermore, we have previously shown that the pattern of HbA1c with age and gender was consistent across seven laboratories when analysed separately [1]. This was also the case when the sites considered in this study were analysed separately (data not shown). We therefore believe that our findings are representative of the UK population. Furthermore, similar patterns have been seen in other populations [12,13,14,15,16].

The between-sex difference could be exaggerated by the large sample size. However, we would emphasise that the aim of this paper is not to change the guidance but rather to raise a question about whether the reference range for younger women is correct, with the suggestion that this matter merits further evaluation.

As we examined laboratory data here, not primary care data, information about the full medical history, including a family history of diabetes and physical characteristics, is not available. Such an evaluation would require access to primary care records. A strength of this study is that because we have used laboratory data, we can be certain that all the tests done for HbA1c and fasting glucose were done in the context of screening for type 2 diabetes rather than the monitoring of already diagnosed type 2 diabetes and were done on the same day.

Our dataset included matched FPG and HbA1c samples based on primary care patients with a single HbA1c sample over a 3-year period. While this would minimise the likelihood of including people with diabetes or being tested for gestational diabetes, it is possible that some such patients would have been included. However, the proportion of these is likely to be small and therefore they are unlikely to change the overall conclusions of the study. Similarly, while the sampling may include some pregnant women and people with other conditions affecting HbA1c and/or FPG levels (such as anaemia), these are unlikely to affect the patterns seen to any significant degree. An important limitation of the study regarding diabetes diagnostic thresholds is the absence of most of the confounding variables that may affect the results.