Which risk factors in T2D increase infection risk?

27th April 2024, A/Prof Chee L Khoo

glucose variability

We know that suboptimal HbA1c increases the risk of micro and macrovascular complications in people with diabetes. We also know that suboptimal HbA1c increases infection risks in these people. There is increasing evidence that glucose variability is a potent predictor of complications. There are two ways to look at glucose variability – day to day variation as seen on continuous glucose monitoring and variability in HbA1c measured over years. Perhaps, this is what we are missing in our jigsaw puzzle. Why do some patients with T2D have more complications that others even though their HbA1c isn’t that bad.

One of the problems in trying to tease out the effect of glucose variability on complications is most of the patients with suboptimal HbA1c are the ones with fluctuating glucose levels and is the higher complication rates from the high HbA1c or from the variability. We also don’t know whether the higher risks apply to those with lower HbA1c.

Carey I et al. explored associations between HbA1c variability and serious infections, and how these vary by HbA1c level, age, sex and ethnicity. Only patients who have had at least 4 HbA1c recorded during 2011-2014 qualified for the study. He used primary care database in the UK, Clinical Practice Research Datalink (CPRD).

HbA1c variability score (HVS) counts how frequently HbA1c rises or decreases across a series of successive measurements made over time and summarised as a percentage. One criticism is whether an absolute visit-to-visit change of 0.5 % is equivalent for patients with very different baseline HbA1c levels, and whether the threshold for significant fluctuations should instead be based on a relative change. In this study, for their main analyses, they used relative changes of +/-10 percent from the previous HbA1c but they used absolute change for their sensitivity analysis. For infection outcomes, they used linked hospital data.

Patients with a history of hypoglycaemia by 2015 were excluded. They only included HbA1c measures if they were at least 90 days apart and excluded patients who had a prior infection hospitalisation during 2011–14. They only included patients first diagnosed in the last 5 years (2010–14) and patients whose HbA1c measurements during 2011–14 were all between 6-8%.


The average HbA1c level was 7.5%. More than 1-in-3 patients had a HbA1c variability score (HVS) of ≥ 50, meaning that at least half of their measurements during 2011–14 were varying by ≥ 10 % relative to the previous measurement. Being male, of Black ethnicity, living in more deprived areas, having a higher BMI recorded and being prescribed more anti-diabetic medications and/or requiring insulin were also associated with higher mean average HbA1c levels and HVS.

During the 5-year follow-up, 88,929 people with T2D (21.6 %) were hospitalised for an infection or had a hospital acquired infection (an annual rate 67.2 per 1,000 persons). Increasing HbA1c levels and variability were both separately associated with infections requiring hospitalisation, a 20 % greater risk (IRR > 1.2) was observed with modest variability (HVS ≥ 20, 73 % of T2D patients) whereas a greater risk was only seen at higher average levels (≥64 mmol/mol, 27 % of T2D patients). Attributable risk fractions (assuming a causal relationship and all patients achieve the reference category) were 19.1 % for variability and 10.0 % for average level.

When they looked at those with less severe diabetes, variability overall was still more influential than average level in predicting infections resulting in a hospitalisation. When they looked at infection risk with average HbA1c, the positive association between variability and infections is observed at all HbA1c levels except at the highest category (≥86 mmol/mol). At lower average HbA1c levels the association between infection risk and HVS was far stronger. For example, in those with lower HbA1c (6-6.5%), high HVS were 2X more likely to have infections compared with those with similar HbA1c but with low HVS. On the other hand, with those with highest levels of HbA1c, HVS only increased the infection risk by 1.5 X.


Individual risks associated with HbA1c variability were more marked in women and older patients, but largely absent among the Black ethnic group. The associations between HVS and infection at HbA1c levels below 9.0% are consistently observed in both men and women, but for age the trends with HVS appear stronger among older patients with T2D. All ethnicities show positive associations with HVS at HbA1c levels below 75 mmol/mol except for Black ethnicity, where there is an absence of a consistent positive association with variability.

The largest association was with sepsis, where 1-in-5 infections could be attributed to an HVS ≥ 20; similar strong associations were seen for bone and joint, genitourinary and lower respiratory tract infections. High average HbA1c level contributed most to bone and joint infection (AF = 12.9 %). For each infection type, attributable risk fractions for variability were higher than those estimated for average levels.

In summary, high HbA1c variability seems to have a greater impact on diabetic complications (especially infection risk) than high HbA1c. This is particularly the case in those with lower HbA1c levels. Thus, just looking at a patient’s HbA1c alone is insufficient to gauge their complications risks. Visit to visit variation of the HbA1c is a potential risk marker. SGLT2 inhibitors have been shown to reduce glucose variability in the post-hoc analysis of the EMPA-REG outcome trials (2). Whether the reduction in mortality in the trial is attributable to reduction in glucose variability is unknown.

Of course, with the advent of continuous glucose monitoring (CGM), we can also get a report from the Ambulatory Glucose Profile (AGP) about glucose variability.


  1. Iain M Carey, Julia A Critchley, Umar A R Chaudhry, Derek G Cook, Stephen DeWilde, Elizabeth S Limb, Liza Bowen, Stephen Woolford, Peter H Whincup, Naveed Sattar, Arshia Panahloo, Tess Harris, Effects of long-term HbA1c variability on serious infection risks in patients with type 2 diabetes and the influence of age, sex and ethnicity: A cohort study of primary care data.  Diabetes Research and Clinical Practice, Volume 211, 2024, 111641
  2. Ceriello, A.P. Ofstad, I. Zwiener, et al. Empagliflozin reduced long-term HbA1c variability and cardiovascular death: insights from the EMPA-REG OUTCOME trial Cardiovasc Diabetol, 19 (1) (2020)