Does glucose self-testing help in type 2 diabetes without insulin?

For people with type 2 diabetes who do not take insulin, testing blood sugar at home with a finger-prick meter has repeatedly failed to improve long-term glucose control in randomised trials [s1] [s3]. The question is worth taking seriously because self-monitoring is widely done, costs money, and involves a needle several times a day for a result that, on the trial evidence, does not move the number that matters.

That number is HbA1c, the share of haemoglobin that has become glycated, used as a measure of average blood sugar over the preceding months. It is what treatment aims to control, and it is the outcome the trials below were built to measure. Self-monitoring — the finger-prick spot check — is a different thing from HbA1c and from the continuous sensors now marketed to people without diabetes; the case for it rests on the idea that seeing a reading prompts a change that eventually lowers the average.

What the trials found

The DiGEM trial, published in the BMJ in 2007, randomised 453 patients with non-insulin-treated type 2 diabetes across 48 general practices in Oxfordshire and South Yorkshire [s1]. Their mean age was 65.7 years, the diabetes had been present for a median of three years, and the mean HbA1c at entry was 7.5% — reasonably well controlled to begin with [s1]. One group received usual care with HbA1c checked every three months (n=152); a second added self-monitoring with advice to contact their doctor to interpret the results (n=150); a third added self-monitoring with extra training in how to act on the readings (n=151) [s1].

At 12 months the differences in HbA1c between the three groups were not statistically significant (P=0.12) [s1]. The unadjusted change from baseline differed from usual care by −0.14 percentage points (95% confidence interval −0.35 to 0.07) in the less intensive monitoring group and by −0.17 percentage points (−0.37 to 0.03) in the more intensive group — small effects whose confidence intervals both cross zero [s1]. The authors concluded the evidence was not convincing of a benefit from self-monitoring, with or without training in how to use it [s1].

A decade later the MONITOR trial reached the same place by a different route. Conducted in 15 primary care practices in central North Carolina, it randomised 450 people with non-insulin-treated type 2 diabetes and an HbA1c above 6.5% but below 9.5% to one of three approaches: no self-monitoring, once-daily self-monitoring, or once-daily self-monitoring with automatic tailored feedback messages delivered through the meter [s3]. Of those, 418 (92.9%) completed the final visit [s3]. At 52 weeks there were no significant differences in HbA1c across the groups (P=0.74): self-monitoring with messaging differed from no monitoring by −0.09 percentage points (95% confidence interval −0.31 to 0.14), and plain self-monitoring by −0.05 (−0.27 to 0.17) [s3]. Quality of life did not differ either, and there was no signal of more hypoglycaemia, more health care use, or earlier insulin initiation [s3]. Even adding a "smart" meter that told patients what their reading meant changed nothing [s3].

What the pooled evidence says

A Cochrane review brought together 12 randomised trials in 3,259 people with non-insulin-treated type 2 diabetes, with follow-up ranging from six to 12 months [s2]. Pooling the studies in patients whose diabetes had lasted a year or more, self-monitoring produced a statistically significant fall in HbA1c of 0.3 percentage points at up to six months (95% confidence interval −0.4 to −0.1; nine trials, 2,324 participants) [s2]. By 12 months that effect had shrunk to 0.1 percentage points and was no longer significant (95% confidence interval −0.3 to 0.04; two trials, 493 participants) [s2]. On patient satisfaction, general well-being and health-related quality of life, the review found no effect [s2].

The pattern across all three is consistent: a small early dip that does not hold, and no improvement in how people feel. A drop of a few tenths of a percentage point in HbA1c is at the edge of what is clinically meaningful, and here it fades.

Why the intuition misleads

Self-monitoring plainly helps people who adjust insulin doses, where a reading directly informs the next injection — that is not in dispute, and the trials above deliberately excluded such patients [s2]. The gap is between that setting and the person on tablets or lifestyle measures alone, for whom a number on a screen carries no built-in action. Without a decision that hangs on the reading, the information does not translate into a lower average, and repeated testing can add cost and anxiety for no measurable return.

What it means for a reader

For non-insulin-treated type 2 diabetes, the trial evidence does not support routine daily finger-prick testing as a way to improve control [s1] [s3]. This is a statement about a population average, not a rule for an individual: some people and clinicians use occasional structured testing to troubleshoot a specific problem, check for lows, or learn how a particular meal or illness moves their glucose, and that targeted use is a different question from testing every day out of habit. As with validated home blood-pressure monitoring and with what an A1c can and cannot tell you, the value of a home measurement depends entirely on whether a decision actually turns on it. Whether to monitor, and how often, is a clinical judgement — and the same caution applies to glucose sensors sold to people without diabetes, where the case is weaker still.