What Does 0.29% Reduction Mean in Diabetics?

The recent largely expanded Cochrane Systematic Review on periodontal treatment for glycemic control in diabetics by Simpson et al. 2015 calculates, in one of numerous meta-analyses, a weighted mean 0.29% reduction (95% confidence interval 0.10% lower to 0.48% lower) of HbA1c  3-4 months after nonsurgical periodontal therapy. Fourteen clinical trials were considered comprising about 1499 patients with both diabetes and periodontitis. The quality of evidence was considered low. Quality was downgraded twice, first for risk of bias, mainly due to lack of blinding; and then due to moderate heterogeneity (I2 = 0.53%). The observed small reduction of 0.29% was not sustained afterwards [1]. At 6-month follow-up (the meta-analysis considered 5 studies with 826 patients), the weighted mean reduction was 0.02% (95% CI 0.20% lower to 0.16% higher).

What does a weighted mean 0.29% reduction of HbA1c levels mean in diabetics? Simpson et al. claim that,

the epidemiological analysis of UKPDS 1998 data indicated that for every percentage point decrease in HbA1c, there was a 35% reduction in the risk of microvascular complications, which appeared to be linear; however we acknowledge that a linear relationship may not exist at lower levels.

The UKPDS compared intensive diabetic care, including sulphonylurea medication or insulin supplementation, with a standard diet protocol. 3867 newly diagnosed patients were randomized. Over 10 years, HbA1c averaged 7% in the intensive care group as compared to 7.9% in the control group [2]. Most of the observed risk reduction in the any diabetes-related aggregate endpoint was due to a 25% (95% CI 7%-40%) risk reduction in microvascular endpoints.

Simpson et al. (2005) emphasize that glycemic control is only one component of management of diabetes and

smoking cessation, weight loss, physical activity and management of dyslipidaemia and hypertension, where appropriate, are particularly important to reduce risk of macrovascula disease. Consequently, the effect of the modest reduction (0.2% [sic!]) in HbA1c seems unlikely to result in a major population-level effect, particularly as effectiveness of periodontal treatment is only estimated to 3-4 months post-treatment within this review.

There was no evidence that periodontal treatment results in a signifcant effect at 6 months.

As regards assessed quality of evidence (low), one might argue whether downgrading due to lack of blinding (diabetic patients were aware whether they underwent periodontal treatment or not), which was considered as main source of bias, was really justified. Also, downgrading because of moderate heterogeneity might be reconsidered. Simpson et al. (2015) compare their results with only three systematic reviews up to 2013 [3]. They notice that point estimates were slightly higher in previous meta-analyses but conclude that,

[as] would be expected with an increased number of included studies and participants, the treatment effect estimate range has narrowed from previous reviews.

Inclusion of the study by Engebretson et al. (2013) , the largest study so far, might be criticized by an armada of editors of our major periodontal and dental journals, presidents and past presidents of our main scientific societies and numerous further “opinion leaders” who had pointed to problems with recruitment of study participants and questionable periodontal outcomes after possibly insufficient periodontal treatment and suggested that the study should be excluded from future systematic reviews [4]. Simpson et al. write,

Having considered these concerns, we believe that the inclusion of such patients [patients with HbA1c range being close to optimal levels, patients above and below protocol-specified HbA1c thresholds and high body mass index] reflects the breadth of population likely to be seen in clinical practice, and the inclusion of the trial enhances estimation of the true effect of periodontal treatment for glycaemic control in diabetic patients. A further criticism of Engebretson 2013 made by Borgnakke 2014 is that the trial’s periodontal outcomes indicate that the accepted standard of care was not met by the periodontal therapy provided to trial participants; however, in meta-analysis to derive treatment effect estimates for periodontal indices, Engebretson 2013’s outcomes were consistent at both time points for all reported outcomes with the other included studies. Consequently, we are satisfied that Engebretson 2013’s clinical conduct is not of sufficient concern to warrant post-hoc sensitivity analyses excluding its contribution, and have confidence in its findings being consistent with those of other included studies.



[1] My own random effects meta-analysis in June 2014 yielded a largely similar weighted mean HbA1c reduction of 0.28% (95% CI -0.45%; -0.10%). Likewise, moderate heterogeneity was observed.

[2] Despite intensively scanning the document, I couldn’t find the “35% for every percentage point” inference claimed by Simpson et al. (2015).  Given that risk reduction in relation to HbA1c reduction is linear, Simpson et al. (2015) might have conceded that it was not clear whether a 0.3% HbA1c reduction may be translated in about 8% risk reduction of microvascular complications (after 10 years).

[3] Since then another six systematic reviews  can easily be identified in PubMed when searching for (diabetes) AND (periodontal therapy).  I have written about the surge of questionable systematic reviews about the topic earlier this year.

[4] I have documented the “Engebretson scandal” intensively on this blog, see, for example, here, here and here.


22 November 2015 @ 9:05 am.

Last modified November 22, 2015.



Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s