Effect of non-surgical periodontal treatment on HbA1c: a meta-analysis of randomized controlled trials

Authors


Address for correspondence:

Dr Amy Liew Kia Cheen

Faculty of Dentistry

Universiti Kebangsaan Malaysia

Jalan Raja Muda Abdul Aziz

50300 Kuala Lumpur

Malaysia

Email: amyliew@dental.ukm.my

Abstract

Background

A meta-analysis of randomized clinical trials (RCTs) was conducted to evaluate whether non-surgical periodontal treatment can reduce the HbA1c% level in type 2 diabetic patients.

Recent accumulation of RCTs necessitates updating of the findings of previous reviews.

Methods

A search of the literature on English publications was conducted in Cochrane CENTRAL, Medline and EMBASE (until 31 March 2012). An RCT was selected if the study population was type 2 diabetic patients (≥16 years old) diagnosed with periodontitis, and compared HbA1c% change with or without non-surgical periodontal treatment for at least three months of the study duration. Weighted mean differences for pooled data and antibiotic use strata were calculated. Heterogeneity and publication bias were explored.

Results

A total of 358 articles were identified but only six were suitable. Compared to the control group, the pooled analysis (n = 422) showed −0.41% (95% CI: −0.73% to −0.09%, p = 0.013) absolute difference in HbA1c% with treatment. Studies without adjunctive antibiotic had HbA1c% change of −0.64% (95% CI: −1.06% to −0.23%, p = 0.002), but we could not conclude on the effect of adjunctive antibiotic use (p = 0.734). Publication bias was significant with Egger's test (p = 0.014) but not with Begg's test (p = 0.06).

Conclusions

The meta-analysis suggested that non-surgical periodontal treatment was associated with a reduction in HbA1c%.

Abbreviations and acronyms
HbA1c

glycated haemoglobin

MD

mean difference

RCT

randomized clinical trials

SE

standard error

Introduction

In recent years, the relationship between periodontal disease and systemic diseases such as diabetes mellitus has attracted the attention of researchers worldwide. Many reviews suggesting a two-way relationship between periodontitis and diabetes mellitus have been published.[1-5] Poor glycaemic control is associated with increased risk of developing periodontal disease,[6, 7] more extensive and severe destruction of periodontal tissue[6, 7] and alveolar bone loss.[8] Hence, it is commonly regarded as one of the risk factors of periodontal disease. Improved glycaemic control may alleviate periodontal disease.[9] In turn, control of periodontal disease was postulated to enhance glycaemic control in patients with type 2 diabetes.[9]

Periodontitis is diagnosed when there is gingival inflammation, connective tissue and alveolar bone loss, and apical migration of junctional epithelium.[11] Clinical signs include increased probing pocket depth or clinical attachment loss. Interventions for periodontitis range from oral hygiene instruction, scaling, root planing, antibiotics, chlorhexidine to surgical treatment, or a combination of these.

Chronic hyperglycaemia resulting from defects in insulin secretion and/or insulin action is typical for diabetic mellitus.[12] Possible long-term outcomes of this disease includes pathological and functional damages to various organs. Diabetes mellitus has a high prevalence worldwide, with an age-standardized adult prevalence of 9.8% in males and 9.2% in females in 2008.[13] Blood glucose control is crucial to prevent complications.[14-19] Glycated haemoglobin (HbA1c) is often used in clinical settings to monitor average blood glucose levels over the past few months, hence high HbA1c indicates poor glycaemic control.[20]

In 2010, a Cochrane review estimated a 0.4% reduction of HbA1c associated with periodontal treatment. However, the review also reported low quality of evidence and suggested more studies to be done.[21] Since then, several new controlled trials have emerged in publications.[22-26]

Therefore, it is timely to conduct a systematic review and meta-analysis to aggregate more evidence and enhance the knowledge of the potential effect of non-surgical periodontal treatment in glycaemic control. Answering this fundamental question will enable physicians and dentists to make evidence-based recommendations to diabetic patients.

Methods

Search strategy

Three databases, Cochrane CENTRAL, MEDLINE (via PubMed) and EMBASE, were systematically searched using the search terms as listed online (Supplementary Appendix 1) with the assistance of a reference librarian, from the inception of all databases to 31 March 2012. Systematic hand-searching was abandoned as the initial hunt could not identify relevant trials which went unrecorded in electronic databases. However, the reference lists of the retrieved publications were carefully cross-checked to avoid omission of relevant trials.

Study selection criteria

To be included in the meta-analysis, the studies had to meet the following criteria: (1) study design: randomized controlled clinical trial (RCT); (2) population: diabetic patients (type 2 diabetes mellitus), at least 16 years old, diagnosed with periodontitis; (3) intervention: non-surgical periodontal treatment including mechanical instrumentation, ultrasonic debridement, supragingival irrigation, subgingival irrigation; with or without adjunctive use of local drug delivery and systemic antibiotics; a minimal follow-up period of three months; (4) comparator: control group with no periodontal treatment or delayed treatment; (5) outcome: mean change in HbA1c level; or pre- and post-treatment HbA1c levels; and (6) published in the English language.

The Cochrane Handbook for Systematic Reviews of Interventions (Version 5.1.0) was used as a guideline for the selection process.[27] Four independent reviewers (AL, NP, YC and JY) evaluated all retrieved articles. Disagreements were resolved by consensus. The reasons for exclusion are listed in Fig. 1. As the databases were not mutually exclusive, replication of entries was removed.

Figure 1.

QUORUM flowchart.

After excluding articles ineligible for the study, the relevance of all articles from the titles and abstracts was evaluated. Full-text articles were reviewed while articles which were not in accordance with the pre-specified selection criteria were excluded.

Methodological study quality assessment

The selected studies were assessed on the criteria of a randomized controlled trial, which were: (1) random sequence generation; (2) allocation concealment; (3) addressing incomplete outcome data; (4) blinding; and (5) intention-to-treat analysis.

Data extraction

Four authors (AL, NP, YL and JY) independently extracted data from each RCT in pre-designed electronic forms. Any discrepancy was resolved by consensus. From all relevant studies, the key points including demographic characteristic of the population, study quality, intervention in the control and treatment groups, study duration and design, pre- and post-treatment and absolute change of HbA1c levels in each treatment arm were recorded.

When not reported, the absolute difference of HbA1c percentage was calculated by extracting the mean change of HbA1c in the control group and the mean change of HbA1c in the intervention group.

If the standard deviation was not reported, standard error (SE) was imputed first by the following formulas using the available t-value or t-value computed from p-value[27]:

display math

in which SE is the standard error, MD is the mean difference and t is the t-value. Using the SE, the standard deviation can be obtained by:[27]

display math

where NI and NC denote sample size for the intervention group and control group, respectively. If the standard deviation was not reported, it was imputed by the following formulas:[27]

display math

in which Corr is the correlation coefficient and SD is the standard deviation. Correlation coefficient was set at 0.5, consistent with a previous review.[28]

Statistical analysis

All numerical data for meta-analysis were entered in STATA 11.0 (Texas, USA). A meta-analysis was conducted, comparing the intervention group which received periodontal treatment and the control group with no periodontal treatment.

For a particular multi-arm study, the intervention group fulfilling the inclusion criteria for pair-wise comparison would be considered.[27] Thus, only the intervention group and control group contrasting non-surgical periodontal treatment (with/without adjunctive antibiotic) versus no treatment/delayed treatment/usual care will be used for data analysis.

Weighted mean difference was estimated using a random-effect model (DerSimonian-Laird method) from the selected RCTs. In addition, the presence of significant heterogeneity was evaluated by chi-square test for homogeneity (Cochran Q-statistic and p-value), as well as the I-squared statistic and 95% CI of the I2. Stratification was done for studies with and without antibiotic use. Publication bias was assessed by Begg's test, Egger's test and funnel plot.

Results

The literature search from the three databases yielded a total of 358 articles, after removal of duplications. The titles and abstracts of these articles were screened for relevance, and the reasons for exclusion are listed in Fig. 1. Of these articles, only 26 were qualified for full-text reading. In the end, only six RCTs fulfilled the pre-specified inclusion criteria.

Singh et al. mentioned that the subjects in their study were randomized,[29] but the randomization process such as sequence generation and allocation concealment methods were not described. Hence, the study was excluded from this meta-analysis. No additional studies were identified by cross-referencing the bibliographies of relevant and included articles.

Study characteristics

The characteristics of the six studies are summarized in Table 1. All studies were reported as randomized controlled trials.[22, 23, 30-33] The mean age per study ranged from 53.8 to 63.2 years. Although smoking is an important risk factor of both periodontal disease[34] and diabetes mellitus,[35] only Jones et al. reported the percentage of smokers in each arm.[32] All studies described study populations having type 2 diabetes mellitus and were diagnosed with chronic periodontitis.[22, 23, 30-33]

Table 1. Study characteristics
StudyYearCountryNumber (n)InterventionsMean age–years (SD) Male%Mean baseline HbA1c% (SD)Follow-up time (months)Study design and details
TotalCTx1Tx2CTx1Tx2CTx1Tx2
Kiran et al.302005Turkey442222 

Control: delayed treatment

Tx1: oral hygiene instruction/advices, scaling, root planing/curettage/debridement

52.82 (12.27)

36

55.95 (11.21)

45

 7.0 (0.72)7.31 (0.74) 3

RCT

Single centre

Blinding of assessor

Promsudthi et al.312005Thailand522527 

Control: no treatment

Tx1: oral hygiene instruction/advices, scaling, root planing/curettage/ debridement, removal of supra and subgingival calculus

61.64 (5.81)

40.74

61.11 (5.83)

32

 9.17 (1.02)8.98 (0.88) 3

?RCT

Single center

Jones et al.322007US1548074 

Control: delayed treatment

Tx1: scaling, root planing/curettage/ debridement, antibiotic, chlorhexidine

60

94

59

100

 10.22 (1.29)9.9 (1.28) 4

RCT

Multi-centre, veterans

intention-to-treat analysis.

Blinding of assessor

C:19% smoker, Tx1:29% smoker

Katagiri et al.332009Japan491732 

Control: delayed treatment

Tx1: Oral hygiene instruction/advices, root planing/curettage/ debridement, antibiotic

59 (4.8)

54.55

60.3 (9.9)

65.63

 6.9 (0.9)7.2 (0.9) 6

RCT

Multi-centre

Chen et al.232012China134444545

Control: no treatment

Tx1: local treatment (e.g. caries, restoration, root canal tx, extraction), scaling, root planing/curettage/ debridement

Tx2: supragingival prophylaxis

63.2 (8.51)

41.5

59.86 (9.48)

54.8

57.91 (11.35)

60.5

7.25 (1.49)7.31 (1.23)7.29 (1.55)6

RCT

Single centre

Moeintaghavi et al.222012Iran401822 

Control: delayed treatment

Tx1: scaling, root planing/curettage/ debridement

N/A

61.1

N/A

40.9

 8.72 (2.22)8.15 (1.18) 3

RCT

Single centre

Blinding of assessor

Scaling and root planing/curettage/debridement were prescribed as the basic non-surgical periodontal intervention for the treatment group in all studies.[22, 23, 30-33] In Chen et al., Group 2 (n = 45) was not included for meta-analysis because only supragingival prophylaxis was given at three-month follow-up.[23] Variation existed with regards to adjunctive therapy, including the prescription of chlorhexidine and antibiotic.

Methodological study quality assessment

The studies varied in terms of quality and were inadequate in many ways (Table 2). In particular, reporting of the allocation concealment, assessment of loss-to-follow-up and blinding was limited in these studies. Only Jones et al. reported conducting intention-to-treat analysis.[32] Blinding was not reported in many studies.[23, 31-33]

Table 2. Quality measure of included studies in the meta-analysis
 Adequate sequence generation?Allocation concealment?Incomplete outcome data addressed?Blinded clinical operator?Masking of laboratory assessment?Masking of periodontal outcome assessor?Intention-to-treat analysis done?
  1. + adequate, ? unclear, - inadequate

Kiran + ? + + + + ?
Promsudthi + + ? ? ? ? -
Jones + + + + + + +
Katagiri + + - ? ? ? ?
Chen + + + - ? ? -
Moeintaghavi + ? + + ? + -

Pooled analysis

A significant difference in HbA1c (%) change was evident in the pooled analysis (Fig. 2) between the treatment and control groups (p = 0.037), with an effect size of −0.41% (95% CI: −0.73% to −0.09%).

Figure 2.

Pooled analysis.

Exploration of heterogeneity

Heterogeneity among studies was significant (χ2 = 11.85, p = 0.037) when evaluated by the Q statistic. I2 was estimated at 57.8%, which suggested moderate heterogeneity.

Since antibiotic use was identified a priori as a potential source of heterogeneity, subgroup analysis was done accordingly. Non-surgical periodontal treatment without adjunctive antibiotic showed a significant improvement (p = 0.002) of HbA1c level [−0.64% (95% CI: −1.06% to −0.23%)]. Also, the heterogeneity among studies was moderate (χ2= 5.35, p = 0.148, I2 = 43.9%). However, this improvement was not significant (p = 0.734) with antibiotic use.

Publication bias

Publication bias existed as the funnel plot (Fig. 3) displayed an asymmetrical distribution. In Egger's test, the intercept value was negative (−4.54). The publication bias was statistically significant in Egger's test (p = 0.014), therefore the null hypothesis of no small-study effects was rejected. However, Begg's test was marginally significant (p = 0.06).

Figure 3.

Funnel plot. Assymetrical distribution, suggesting small study effect.

Discussion

The selected studies showed reasonable comparability in terms of population characteristics as well as baseline HbA1c levels within the study. Population comparability between studies was also satisfactory, except Jones et al. which had a high percentage of males as the study was conducted among veterans.[32] The research quality was far from ideal. For instance, per protocol analysis was done in Promsudthi et al.,[31] Chen et al.[22] and Moeintaghavi et al.[23] This could jeopardize the purpose of randomization.

Furthermore, the reporting of study methods was not comprehensive, making it difficult to inspect for heterogeneity due to study quality. The duration of treatment ranged from three to six months; treatment and reassessment cycles were also differential. Although all studies provided mechanical instrumentation as intervention, Jones et al. prescribed additional systemic doxycycline 100 mg by mouth daily for 14 days,[32] whereas Katagiri et al. provided topical minocycline 10 mg for all periodontal pockets at the end of every visit.[33]

Despite this, it was evident that non-surgical periodontal treatment is beneficial in glycaemic control, as there was an overall 0.41% reduction in HbA1c level with intervention. This was similar to the findings of the two reviews in 2010 by Simpson et al. and Teeuw et al., which reported a 0.40% reduction of HbA1c compared to the control group.[21, 28] However, caution is warranted in interpreting these data as the heterogeneity among studies was significant (p = 0.037).

When stratified by antibiotic use, the analysis confirmed that mechanical instrumentation alone could indeed improve glycaemic control, reducing 0.64% unit of HbA1c level compared to no treatment. The subgroup analysis of non-surgical periodontal treatment without antibiotic showed greater effect size, but this could be due to the exaggeration driven by small studies (n < 100).

In contrast, studies using adjunctive antibiotic, whether systemic or topical, suggested that the overall effect was non-significant. Interestingly, recent studies have shown otherwise. Sub-antimicrobial level was found to be beneficial for periodontal control[36] and glycaemic control[37] in diabetic patients. Biological plausibility and pharmacological mechanism could not be established at this point. Also, only two trials used adjunctive pharmacological therapy. Therefore, more research is needed in this area.

Despite the promising outcome, our study showed there was a significant publication bias resembling ‘small-study effects’. Small studies with findings favouring the intervention of interest were more likely to be published in journals, and vice versa.[38] In this meta-analysis, four out of six studies had a small sample size (n < 100). However, language restrictions in the study selection could have omitted published negative findings in other languages. Alternatively, it was likely that the current attention on oral systemic health association had encouraged researchers and reviewers on publishing concurring results.

The percentage reduction of HbA1c associated with non-surgical periodontal treatment was modest. Still, its clinical importance could not be denied. Since type 2 diabetes mellitus is a multifactorial disease, glycaemic control requires effective management of these risk factors.[39, 40] Hence, despite the small effect size, treatment of periodontal diseases contributes to the overall picture of better general health outcome.

However, monitoring and controlling glycated haemoglobin for type 2 diabetes mellitus should be individualized for each diabetic patient, as disease duration, pre-existing macrovascular conditions, hypoglycaemic unawareness, comorbidities and frailty varied among patients,[41] and there was limited evidence of improved mortality with HbA1c reduction.[42, 43] Recommendations of healthy eating habits, healthy lifestyle, physical activities and weight management remained as essential.[44]

Also, HbA1c should not be the sole indicator of glycaemic control, as confounders such as haemolytic anaemia, red-cell survival, extracellular-intracellular glucose balance and non-glycaemic genetic determinants of haemoglobin glycation may hinder accurate diagnosis.[37] Future research about the association between periodontal treatment and glycaemic control should utilize a combination of conventional measures of blood glucose (e.g. fasting plasma glucose[22, 23, 30, 31] and post-prandial glucose[30]) to improve validity.

The strengths of this meta-analysis were the restriction to only randomized controlled trials and the inclusion of additional studies since the publication of the two reviews in 2010. Effort was made to systematically and thoroughly search the literature. The findings suggested the need to better understand the interaction of antibiotic and non-surgical periodontal treatment, as well as the significant presence of publication bias.

However, limitations in the robustness of the results were recognized, including the scarcity of RCTs, small sample size within the study, as well as variation in intervention within and across studies. This meta-analysis also did not include non-English publications, which might be systematically different.

Thus, recommendations for future research examining this association are to: (1) have adequate randomization; (2) reduce within study heterogeneity by giving standardized treatment without topical or systemic antibiotic use, as well as standardized control of delayed treatment; (3) have a follow-up duration of at least three months; (4) examine fasting blood glucose level concurrently; (5) perform intention-to-treat analysis; (6) have thorough reporting of methods and measured outcomes.

More importantly, biological plausibility of the effect of non-surgical periodontal treatment and change in glycated haemoglobin level should be explored. In view of the lack of evidence, we also urge researchers to embark on studies comparing scaling and root planing only (control) versus scaling, root planing and adjunctive antibiotic (treatment). Researchers should be aware of the presence of publication bias in this area, and should not be discouraged to publish negative findings.

Finally, it is hoped that with more high-quality research, evidence-based guidelines could be formulated in the future, to guide physicians and dentists in the management of periodontal disease in diabetic patients.

Conclusions

This meta-analysis found there was a small beneficial effect of non-surgical periodontal treatment in glycaemic control. However, the effect of adjunctive antibiotic usage in this association remained inconclusive to date.

Ancillary