Treatment of periodontal disease for glycaemic control in people with diabetes (Review)

Transcription

1 Treatment of periodontal disease for glycaemic control in people with diabetes (Review) Simpson TC, Needleman I, Wild SH, Moles DR, Mills EJ This is a reprint of a Cochrane review, prepared and maintained by The Cochrane Collaboration and published in The Cochrane Library 2010, Issue 5

2 T A B L E O F C O N T E N T S HEADER ABSTRACT PLAIN LANGUAGE SUMMARY SUMMARY OF FINDINGS FOR THE MAIN COMPARISON BACKGROUND OBJECTIVES METHODS Figure Figure RESULTS Figure Figure Figure Figure DISCUSSION AUTHORS CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES CHARACTERISTICS OF STUDIES DATA AND ANALYSES Analysis 1.1. Comparison 1 Scaling, root planing and oral hygiene (+/- adjunctive antibiotics) versus no/usual treatment, Outcome 1 HbA1c after 3/4 months Analysis 1.2. Comparison 1 Scaling, root planing and oral hygiene (+/- adjunctive antibiotics) versus no/usual treatment, Outcome 2 CAL after 3/4 months Analysis 1.3. Comparison 1 Scaling, root planing and oral hygiene (+/- adjunctive antibiotics) versus no/usual treatment, Outcome 3 PD after 3/4 months Analysis 1.4. Comparison 1 Scaling, root planing and oral hygiene (+/- adjunctive antibiotics) versus no/usual treatment, Outcome 4 BOP after 3/4 months APPENDICES HISTORY CONTRIBUTIONS OF AUTHORS DECLARATIONS OF INTEREST SOURCES OF SUPPORT INDEX TERMS i

3 [Intervention Review] Treatment of periodontal disease for glycaemic control in people with diabetes Terry C Simpson 1, Ian Needleman 2, Sarah H Wild 3, David R Moles 4, Edward J Mills 5 1 Edinburgh Dental Institute, University of Edinburgh, Edinburgh, UK. 2 Unit of Periodontology and International Centre for Evidence- Based Oral Healthcare, UCL Eastman Dental Institute, London, UK. 3 Centre for Public Health and Primary Care Research, Public Health Sciences, University of Edinburgh, Edinburgh, UK. 4 Oral Health Services Research, Peninsula Dental School, Plymouth, UK. 5 Faculty of Health Sciences, University of Ottawa, Ottawa, Canada Contact address: Terry C Simpson, Edinburgh Dental Institute, University of Edinburgh, Lauriston Place, Edinburgh, Scotland, EH3 8HA, UK. t.c.simpson@oriel.oxon.org. Editorial group: Cochrane Oral Health Group. Publication status and date: New, published in Issue 5, Review content assessed as up-to-date: 23 March Citation: Simpson TC, Needleman I, Wild SH, Moles DR, Mills EJ. Treatment of periodontal disease for glycaemic control in people with diabetes. Cochrane Database of Systematic Reviews 2010, Issue 5. Art. No.: CD DOI: / CD pub2. Background A B S T R A C T Glycaemic control is a key issue in the care of people with diabetes mellitus (DM). Some studies have suggested a bidirectional relationship between glycaemic control and periodontal disease. Objectives To investigate the relationship between periodontal therapy and glycaemic control in people with diabetes and to identify the appropriate future strategy for this question. Search strategy A comprehensive approach was adopted employing handsearching; searching of electronic databases including the Cochrane Oral Health Group s Trials Register, CENTRAL, MEDLINE, EMBASE, CINAHL, ZETOC, ISI Web of Knowledge and LILACS; contact with appropriate non-english language healthcare professionals; authors and organisations. The final date for searching for studies was 24th March Selection criteria This review studied randomised controlled trials of people with Type 1 or 2 diabetes mellitus (DM) with a diagnosis of periodontitis. Suitable interventions included mechanical periodontal therapy with or without adjunctives and oral hygiene education. Data collection and analysis The titles and abstracts of 690 papers were examined by two review authors independently. Ultimately, seven studies were included and 19 excluded after full text scrutiny. All trials were assessed for risk of bias. 1

4 Main results Three studies had results pooled into a meta-analysis. The effect for the mean percentage difference in HbA1c for scaling/root planing and oral hygiene (+/- antibiotic therapy) versus no treatment/usual treatment after 3/4 months was -0.40% (95% confidence interval (CI) fixed effect -0.78% to -0.01%), representing a statistically significant reduction in HbA1c (P = 0.04) for scaling/root planing. One study was assessed as being at low risk of bias with the other two at moderate to high risk of bias. A subgroup analysis examined studies without adjunctive antibiotics -0.80% (one study: 95% CI -1.73% to 0.13%; P = 0.09), with adjunctive antibiotics in the test group -0.36% (one study: 95% CI -0.83% to 0.11%; P = 0.14), and with antibiotics in both test and control groups after 3/4 months -0.15% (one study: 95% CI -1.04% to 0.74%; P = 0.74). Authors conclusions There is some evidence of improvement in metabolic control in people with diabetes, after treating periodontal disease. There are few studies available and individually these lacked the power to detect a significant effect. Most of the participants in the study had poorly controlled Type 2 DM with little data from randomised trials on the effects on people with Type 1 DM. Improving periodontal health is an important objective in itself. However, in order to understand the potential of this treatment to improve glycaemic control among people with diabetes, larger, carefully conducted and reported studies are needed. P L A I N L A N G U A G E S U M M A R Y Treatment of periodontal disease for glycaemic control in people with diabetes Long-term control of blood sugar levels is considered to be of critical importance in preventing complications associated with diabetes. Research evidence has suggested that treating gum disease in people with diabetes may assist in lowering blood sugar levels. The aim of this review was to investigate the relationship of treating gum disease on blood sugar control in people with diabetes and suggest a future strategy for research and medical/dental practice. The evidence gathered suggested that there may be small but significant improvement in blood sugar control from treating pre-existing gum disease in people with Type 2 diabetes mellitus. 2

5 S U M M A R Y O F F I N D I N G S F O R T H E M A I N C O M P A R I S O N [Explanation] Scaling, root planing and oral hygiene(+/- adjunctive antibiotics) compared to no treatment/usual treatment for glycaemic control in people with diabetes Patient or population: patients with glycaemic control in people with diabetes Settings: Intervention: scaling, root planing and oral hygiene(+/- adjunctive antibiotics) Comparison: no treatment/usual treatment Outcomes Illustrative comparative risks*(95% CI) Relative effect (95% CI) Assumed risk no treatment/usual treatment HbA1cafter3/4months The mean hba1c after 3/ 4 months in the control groups was 8.82 Corresponding risk scaling, root planing and oral hygiene (+/- adjunctive antibiotics) ThemeanHbA1cafter3/ 4 months in the intervention groups was 0.4 lower (0.78 to 0.01 lower) No of Participants (studies) 244 (3) Quality of the evidence (GRADE) low 1,2 Comments CALafter3/4months The mean cal after 3/ 4 months in the control groups was 3.54 mm PDafter3/4months The mean pd after 3/ 4 months in the control groups was 2.95 mm The mean CAL after 3/4 months in the intervention groups was 0.07 lower (0.64 to 0.49 lower) The mean PD after 3/4 months in the intervention groups was 0.39 lower (0.64 to 0.15 lower) 90 (2) 90 (2) low 1 low 1 3

6 BOP after 3/4 months The mean bop after 3/ 4 months in the control groups was percentage of sites The mean BOP after 3/4 months in the intervention groups was lower ( to lower) *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention(and its 95% CI). CI: Confidence interval; 90 (2) low 1 GRADE Working Group grades of evidance High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Lowquality:Furtherresearchisverylikelytohaveanimportantimpactonourconfidenceintheestimateofeffectandislikelytochangetheestimate. Very low quality: We are very uncertain about the estimate. 1 Onestudyhadunclearsequencegeneration,allocationconcealmentandincompleteoutcomedata. 2 Effectofantidiabeticmedicationnotaccountedforbyallstudies. 4

7 B A C K G R O U N D Glycaemic control is a key issue in the care of people with diabetes mellitus (DM). Prolonged hyperglycaemia is associated with complications such as retinopathy, peripheral neuropathy, macrovascular disease (coronary heart and cerebrovascular disease), foot disease (arising from a combination of vascular and neuropathic disease) and renal failure. The United Kingdom Prospective Diabetes Study (UKPDS) and the Diabetes Control and Complications trial in the USA have demonstrated that intensive treatment of hyperglycaemia can reduce the risk of long-term complications (DCCT 1993; UKPDS 1998; Stratton 2000). Each 1% reduction in the haemoglobin A1c (HbA1c) in the United Kingdom Prospective Diabetes Survey was associated with a relative risk reduction of 21% for any diabetes-related endpoint, 21% for diabetes-related deaths, 14% for myocardial infarction and 37% for microvascular complications (Stratton 2000). As part of this process, blood glucose levels may be monitored daily by the patient but also by regular haematological tests in a clinical laboratory. The HbA1c level is commonly measured to assess blood glucose levels over a period of approximately 6 to 8 weeks preceding the test. This test is recognised as a good indicator of glycaemic control (Bunn 1981) and higher levels are associated with increased risk of diabetes-related complications. Periodontitis is defined as inflammation and destruction of the underlying supporting tissues of the teeth. The categorisation of the condition is by its behaviour as aggressive or chronic (Armitage 1999). There is no subclass specific to DM, as it is recognised that diabetes may modify all forms of periodontal disease (Milward 2003). Disease severity is graded by measurement of the clinical attachment loss or pocket depth. Periodontal treatment includes a number of components of care. In many cases oral hygiene instruction will be used to educate and motivate patients to control for themselves the accumulation of the causative factor, dental plaque. In addition, mechanical debridement (different forms of scaling) by a dentist or hygienist is often required to removed both plaque and plaque deposits that have mineralised and hardened (calculus). These deposits can form both above and below the gum edge. With more advanced forms of disease, surgery can be needed to lift the gum away from the tooth facilitating access to clean away the deposits when located below the gum line. Some of these measures require several patient visits. Antiseptics and antibiotics have also been used as adjuncts to scaling, although without evidence of a clear benefit. Poorly controlled diabetes is also a well-recognised risk factor for developing periodontal disease (Seppälä 1993; Papapanou 1996) with epidemiological evidence that people with both Type 1 and Type 2 diabetes experience more periodontal disease, and, of greater severity than the general population (Firlati 1997; Sandberg 2000). In the last few years evidence has been published suggesting a bi-directional relationship between glycaemic levels and periodontal disease (Grossi 1998; Stewart 2001; Taylor 2001). In other words, the chronic inflammation and infection that results from periodontal disease could also have an adverse effect on glycaemic control in people with diabetes which in turn could lead to worsening gum disease. It has been estimated that the total surface area of inflamed and ulcerated epithelium of the periodontal tissues in an individual with periodontitis is at least equivalent to the surface area of the palm of the hand (Page 1998). Authoritative studies on diabetes such as DCCT 1993; UKPDS 1998 and Stratton 2000 did not collect data on periodontal disease or oral health in general. The cost to the health budget of managing people with diabetes is substantial. In the UK, current estimates suggest 5% of total National Health Service (NHS) resources and up to 10% of hospital in-patient resources are used for the care of people with diabetes (DoH 2001). We will include any evidence of cost implications of treatment, where these are available from the studies. Any improvement in glycaemic control resulting from regular and appropriate periodontal treatment has the potential to make a significant impact on the quality of life for patients with diabetes. If the hypothesis of a direct effect of periodontal therapy on glycaemic control is accepted, the implications are profound. Periodontal disease, which is prevalent in most populations, could be an additional confounder in studies of the effect of glycaemic control. Simple treatment by dentists and auxiliary workers could have a marked effect in improving glycaemic control among people with diabetes. O B J E C T I V E S The null hypothesis is: Periodontal therapy has no effect on glycaemic control in people with diabetes. The main objectives are: To investigate the effect of periodontal therapy on glycaemic control in people with diabetes To identify whether further research is required in this area and if so, to identify the important research questions and appropriate study designs. A further objective is: To investigate the various combination of therapies used in treating periodontal disease in people with diabetes. 5

8 M E T H O D S Criteria for considering studies for this review Types of studies Randomised controlled trials (RCTs) were eligible for inclusion. Trials were only included if the patients were followed up for no less than 90 days after completion of the treatment course. Types of participants People with diabetes mellitus and periodontal disease who are at least 16 years of age were included. We analysed participants as having a diagnosis of Type 1 diabetes mellitus if, at the beginning of the study, the individual s disease classification was juvenile-onset diabetes, type I or insulin-dependent diabetes mellitus (IDDM). If an individual was described as having adult-onset, type II or non-insulin dependent diabetes mellitus (NIDDM) we analysed the data as for that of a participant with Type 2 diabetes mellitus. For periodontitis, we accepted an author s statement that subjects were selected on the basis of a diagnosis of chronic or adult periodontitis and aimed to investigate the effect of adequacy of diagnosis on the outcomes (see Study quality ). Studies were included regardless of the general medical health of the subjects. No restriction was placed on setting - primary care, hospital or community were all considered. Studies would have been excluded had more than 10% of the study sample had been diagnosed with gestational diabetes (diabetes associated with pregnancy). Types of interventions Interventions included: oral hygiene instruction education or support sessions to improve self-help or selfawareness of oral hygiene mechanical debridement- scaling, subgingival curettage, flap surgery or gingivectomy antiseptic mouthrinses, gels or dentifrices antimicrobial therapy, either locally applied or systemically administered other drug therapy with a possible benefit of improving the periodontal condition of the participant. Interventions were compared with no treatment or placebo, for the purposes of testing our hypothesis. Comparisons between different combinations of those interventions listed above was allowed to allow exploration of the effects of adjunctive treatment. Types of outcome measures A number of different blood products have been identified as indicators of blood glucose levels and therefore, a possible prognostic marker. The glycated (glycosylated) haemoglobin assay (HbA1c) gained widespread acceptance during the 1980s as the laboratory test of choice and is still widely used. HbA1c has been measured using a number of differing methods with several internationally adopted standards. These include the Diabetes Control and Complications Trial (DCCT) or the International Federation of Clinical Chemistry (IFCC) standard tests. The latter consistently gives lower values (non-diabetic reference range is about 3% to 5% IFCC and 4% to 6% DCCT with good control in diabetic groups as 5% IFCC and 7% DCCT. Treatment alteration becomes a requirement with values >6% IFCC and >8% DCCT) (Florkowski 2003). Methods and reference ranges were noted where given and would have been subjected to sensitivity analysis had the information been available. More recently other assays have been used, such as fructosamine (glycated albumin) which is a glycated serum protein, measured in µmol/l. It was anticipated that some studies may have used this measure either in addition to or as an alternative to HbA1c. Its use is as a measure of blood sugar control over a matter of 2 or 3 weeks rather than months (Braatvedt 1997). Measures of glycaemic control may, therefore, not be comparable between studies, but the focus of this review was internal comparisons. Some studies measured blood glucose levels (such as plasma glucose fasting levels) in papers scrutinised in this review. However, we did not feel that it was appropriate to use this as an outcome measure. Whilst blood glucose is useful for management on a daily basis (particularly in Type 1 diabetes mellitus), it can be very variable and heavily influenced by many factors like diet, exercise etc. HbA1c gives a better measure of long-term control and is shown to be more strongly associated with complications of diabetes than blood glucose (Goldstein 2004). Primary outcomes The absolute percentage change from baseline in HbA1c - from pre-treatment for periodontal condition to at least 90 days post-treatment. The percentage change from baseline in fructosamine - from pre-treatment for periodontal condition to at least 90 days post-treatment. Secondary outcomes Changes in periodontal attachment levels. Gingival inflammation. Bleeding. Oral hygiene. Any adverse effects of treatment. Quality of life indicators. Cost implications. 6

9 Diabetic complications. Search methods for identification of studies Electronic searching For the identification of studies included or considered for this review, detailed search strategies were developed for each database searched. These were based on the search strategy developed for MEDLINE (Appendix 1) but revised appropriately for each database to take account of differences in controlled vocabulary and syntax rules. The MEDLINE search strategy combined the subject search with the Cochrane Highly Sensitive Search Strategy for identifying reports of randomised controlled trials (as published in Box 6.4.c in the Cochrane Handbook for Systematic Reviews of Interventions version updated September 2009) (Higgins 2009). The EM- BASE search was carried out with the addition of a search filter developed by the Cochrane Oral Health Group. Details of the searches are shown in Appendix 1. Databases searched The Cochrane Oral Health Group s Trials Register (24th March 2010) The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2010, Issue 3) MEDLINE via OVID (1950 to 24th March 2010) EMBASE via OVID (1980 to 24th March 2010) CINAHL via EBSCO (1982 to 24th March 2010) LILACS (1982 to 24th March 2010) ISI Web of Knowledge (conference proceedings) (1990 to 24th March 2010) ZETOC (conference proceedings) (1993 to 24th March 2010). Handsearching We confirmed from the Cochrane Oral Health Group which journals have already been handsearched as part of their handsearching programme. Known papers previously published on the topic were scrutinized for potentially relevant references. The following journals were handsearched by the authors: -Annals of Periodontology (1996 to 2003) -Periodontology 2000 (1993 to 2003). We did not handsearch any medical or specialist journals relating to diabetes. Searching other resources We attempted to contact known authorities, as identified by the Cochrane Oral Health Group, in the following languages for information about publications, which might contain relevant trials: Japanese, Chinese, German, French and Spanish. In addition to this any papers identified by any of the database searches, which were not in any aforementioned languages were translated and considered for inclusion. We attempted contact in three languages (Spanish, French and English) with organisations serving as information sources for both patients and clinicians on the topics of diabetes mellitus and periodontal disease to request information on the existence of trials known to the organisations. A comprehensive list of worldwide dental organisations was found at the American Dental Association website at This exercise was repeated for organisations related to diabetes using the contact details provided at the following: the Diabetes Care Optimization Through Information Technology (DOIT) website at and the Diabetes Resource website at diabetes_organizations. (The DOIT website is currently unavailable.) In addition, we wrote to pharmaceutical and manufacturers involved in the care of people with diabetes as listed on the World Wide Web at diabetes_companies.php#meter. Contact with authors of relevant studies was sought for clarification regarding their own studies and for information regarding other studies of which they are aware. There were no language restrictions. Data collection and analysis Study eligibility All titles (and abstracts if available) were screened by two review authors (Terry Simpson (TS) and Edward J Mills (EJM)). Only clearly irrelevant records were excluded. Following this, the full text of potentially relevant studies were retrieved and examined. Two review authors (TS and EJM) recorded details about the studies independently. Where studies were not agreed for inclusion another review author (Ian Needleman (IN)) acted as arbiter. The review authors were not blinded, as this has been shown to be of dubious value, but adds a significant amount of time to completion of the process (Berlin 1997). A detailed log of study eligibility and reasons for exclusion was maintained. Both review authors were assessed for inter-examiner and intra-examiner agreement reported as Cohen s kappa (Cohen 1960) and these data are presented in Appendix 3. Data collection We collected the data on a pre-determined and piloted form (Appendix 2). The following is not an exhaustive list but the following characteristics of each study considered were recorded on the data extraction form. 7

10 General characteristics - year of study, language of original publication, country of origin, funding. Trial design - sample size, method of allocation, blinding and comparative group characteristics. Population studied - ethnic groups, setting, social class, whether Type1 or Type 2 diabetes mellitus (DM) or both, duration of diabetes, duration of diabetic control, presence of other stated medical conditions, type of periodontal disease (gingivitis only, chronic/adult periodontitis, aggressive/earlyonset periodontitis), smoking habits, alcohol consumption, drug therapy. Nature of the intervention - oral hygiene, self-administered measures, type of periodontal therapy and antimicrobial/ antiseptics employed. Compliance. Primary outcomes - Hba1C at baseline, during therapy and post-therapy. Fructosamine at baseline, during therapy and posttherapy. Details of the test method was noted along with reference values for the test where these were available. Where a test corresponds to DCCT or IFCC standards this would also have been recorded, if presented. Secondary outcomes - changes in clinical attachment level (CAL), probing pocket depth (PPD), bleeding on probing (BOP), gingival index (GI), plaque index (PI) and the presence/ absence of causative bacteria. Also diabetic complications and changes in antidiabetic therapy. Numerical data were extracted independently by two review authors (TS and EJM) into data tables and Review Manager (RevMan) software. The data inputted into RevMan was verified by a third review author (David R Moles (DRM)). Study quality Studies were analysed for the following to assess validity as a threshold for inclusion, which is described as one of the options in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2009) on the following individual quality criteria: Adequate sequence generation: Yes, No, Unclear Allocation concealment: Yes, No, Unclear Incomplete outcome data addressed: Yes, No, Unclear Free of selective outcome reporting: Yes, No, Unclear Free of other biases: Yes, No, Unclear Blinding of clinical operator: Yes, No, Unclear Masking of (primary outcome) assessor: Yes, No, Unclear Masking of (secondary outcome) assessor: Yes, No, Unclear. Yes indicates a low risk of bias, No indicates high risk of bias and Unclear indicates either lack of information or uncertainty over the potential for bias. A risk of bias table was completed for each included study (Characteristics of included studies). Results are presented graphically by study (Figure 1) and by domain over all studies (Figure 2). Figure 1. Methodological quality graph: review authors judgements about each methodological quality item presented as percentages across all included studies. 8

11 Figure 2. Methodological quality summary: review authors judgements about each methodological quality item for each included study. 9

12 Diagnostic assessment This was assessed as: Diabetes mellitus diagnosis: criteria for diagnosis clearly defined: adequate, inadequate, unclear Periodontal disease diagnosis: criteria for diagnosis clearly defined: adequate, inadequate, unclear. Adequate: At least two sites with probing depth of ³5 mm with ³2 mm loss of clinical attachment and/or alveolar bone loss of more than 4 mm. Inadequate: Less than two sites with probing depth of ³5 mm with ³2 mm loss of clinical attachment and/or alveolar bone loss of more than 4 mm. Uncertain/Unclear: No criteria given. The results were summarised in tabular form and are presented in the review (Additional Table 1). We did not exclude any study on the basis of uncertain, unclear or inadequate criteria. Table 1. Diagnostic assessment Study Grossi Rocha Al-Mubarak Rodrigues Comments Diabetes - WHO criteria for designation as having DM. Periodontal - mean PPD around 3.5 to 3.7 mm for all groups. CAL in the range 4.5 to 5 mm at baseline. No periodontal inclusion criteria stated. Diabetes - diagnosis of DM for at least 5 years. Periodontal - PPD >3 mm in at least 1 tooth. Diabetes - all had the disease for >1 year. Attending hospital for the condition. Periodontal - 4 teeth in 2 quadrants with > or equal to 5 mm but less than 8 mm PPD. No statement on CAL or alveolar bone loss. Diabetes - diagnosed with Type 2 DM. Periodontal - 1 site and 2 teeth with >5 mm PPD and >6 mm CAL. Kiran Diabetes - participants under treatment for Type 2 diabetes with HbA1c in the range 6%-8%. Periodontal - CAL averaged 3.19 mm in the treatment group and 2.92 mm in the control group at baseline. Jones Diabetes -statement that inclusion depended on a repeat HbA1c of > or equal to 8.5%. Periodontal - population did not have very high severity of periodontal disease. Only 13.9% in early treatment group (13.6% of usual care group) and had PD of >3 mm at baseline. Yun Diabetes - newly diagnosed Type 2 DM. Periodontal - PPD > or equal to 5 mm but <8 mm in 1 site in 4 teeth or 2 different quadrants. No indication of CAL or alveolar bone loss. Study authors inclusion criteria for diabetes and periodontal disease. CAL = clinical attachment level; DM = diabetes mellitus; PD = pocket depth; PPD = probing pocket depth. Data synthesis Data were collated in evidence tables, grouped according to research design and assessed for possible meta-analysis on the basis of homogeneity of main characteristics. We recorded and analysed the results for patients with Type 1 and Type 2 DM. separately on data extraction forms. Where this was not possible, due to the trial presenting the results as one, the results and analysis were conducted as a combined group. Meta-analysis was conducted where studies were judged similar enough. Random-effects meta-analyses were used as appropriate 10

13 to combine quantitative data, where there were at least four studies. All the data analysed was continuous. Pooled outcomes were expressed as mean differences (MD) with their associated 95% confidence intervals. Statistical heterogeneity was assessed by calculation of the Q statistic with P value set at P < This was quantified by the calculation of the I 2 statistic for heterogeneity. If there were sufficient studies publication bias would have been assessed by graphical methods and via the Begg and Mazumdar (Begg 1994) adjusted rank correlation test and the Egger et al regression asymmetry test (Egger 1997). If there were sufficient studies sensitivity analyses and meta-analysis regression (using STATA software) would have been used to explore, quantify and control for sources of heterogeneity between studies for the following quality criteria and prognostic factors: study quality periodontal disease severity (initial probing depth) DM control - through categorisation of patients into good, fair and poor (mean HbA1c 7%, between 7% and 8.5% or ³8.5% on the DCCT or equivalent scale) DM duration smoking habits alcohol consumption general health status plaque control socio-economic class drug therapy. The only subgroup analysis we were able to undertake involved the use of antibiotics. Due to insufficient data we were unable to analyse the following subgroups: Type 1 and Type 2 DM smokers and non-smokers duration since diagnosis metabolic control - classified as good, moderate and poor by age by sex presence of other medical conditions. R E S U L T S Description of studies See: Characteristics of included studies; Characteristics of excluded studies; Characteristics of studies awaiting classification; Characteristics of ongoing studies. We examined the titles and abstracts of 690 studies, letters and conference abstracts. These were obtained through the electronic search strategy, handsearch, contact with authors and other sources. Of these, 648 were excluded as not being randomised controlled trials (RCTs) or having a relevant primary outcome or after more information became available. The full text of 26 studies were scrutinised in more detail: 19 were excluded due to inappropriate study design, leaving seven included trials. 16 studies are currently under review and will require further investigation for inclusion in future revisions of this review. Additionally, one further paper awaiting classification (Calbacho 2005) was found in the bibliography of a recent review (Darre 2008). Included studies Characteristics of included studies. Seven studies met our criteria for inclusion in the review however only three studies addressed the primary hypothesis comparing intervention with no treatment/usual treatment (Kiran 2005; Jones 2007; Yun 2007). All seven trials used a parallel design. One was a multicentre trial (Jones 2007) and used a two by two design with a two-stage randomisation process. This paper only reported on the study after the first allocation to groups (0-4 months), although the study was due to run for 12 months from baseline data gathering. This trial randomised the largest number of participants (193). Other trials varied in size between 30 and 125 participants. Four trials recruited their patients from diabetes or endocrinology hospital units (Al-Mubarak 2002; Rodrigues 2003; Kiran 2005; Jones 2007). One trial targeted patients by hospital record in a population (Pima Indians of Arizona, USA) with a very high incidence of Type 2 diabetes mellitus (DM) (Grossi 1997). One study (Yun 2007) recruited from patients referred by their physician to a university periodontal department. The remaining study drew its participants from home visits and public announcements (Rocha 2001). Age ranges varied with the widest range 18 to 80 years old (Al- Mubarak 2002) and the narrowest 50 to 60 years old (Rocha 2001). Four trials included only people with Type 2 DM (Grossi 1997; Rocha 2001; Rodrigues 2003; Kiran 2005). Two others (Al- Mubarak 2002; Jones 2007) enlisted predominantly patients with Type 2 DM, but also had some participants with Type 1 DM. Most studies had a near-equal mix of males and females in each group, but Jones 2007 had nearly all male participants. Conversely one study (Grossi 1997) recruited predominantly female participants with at least one third males allocated to each arm. The latter information was supplied by the study author (Grossi). Each study indicated exclusions based on the medical condition of each participant. Three studies excluded patients taking antibiotics within a defined prior period (Al-Mubarak 2002; Rodrigues 2003; Yun 2007). Some studies gave extensive amounts of lifestyle data including body mass index (BMI), smoking, alcohol intake etc. (Jones 2007) whilst others gave no such information (Grossi 1997; Rodrigues 2003). There are several methods of assessing glycated haemoglobin. Cation-exchange chromatography tests were performed in one trial (Rocha 2001). The HbA1c test employed was not specified in several studies, although some specified the laboratory used (Rocha 11

14 2001; Rodrigues 2003; Jones 2007). One trial (Grossi 1997) specified both the test and laboratory used (high performance liquid chromatography). One study (Yun 2007) used high performance liquid chromatography but with no further details provided. No reference range was quoted in any study. Another study (Kiran 2005) also conducted further metabolic tests for triglycerides, total cholesterol LDL-cholesterol, HDL-cholesterol and microalbuminurea levels. With the exception of one trial, all studies used a two-arm design of intervention versus control group. The exception (Grossi 1997) deployed five different groups. A combination of antibiotic therapy (doxycycline), irrigant (either chlorhexidine gluconate, povidone-iodine) and water or placebo was used in addition to mechanical therapy (scaling and root planing) plus oral hygiene instruction. Two studies tested mechanical therapy versus no treatment (Kiran 2005; Yun 2007). Yun 2007 combined antibiotic therapy (doxycycline) with the mechanical debridement and antibiotic therapy was also given in the control group. Rodrigues 2003 assigned groups to mechanical therapy plus or minus adjunctive therapy (amoxicillin/clavulanic acid). In Al-Mubarak 2002 both groups had mechanical therapy and oral hygiene with the addition of oral irrigation for the test group. Jones 2007 tested mechanical therapy and antibiotics (doxycycline) with antiseptic mouthrinse (chlorhexidine) against usual therapy (including attending own dentist for individual s normal therapy or no treatment). Rocha 2001 examined the effect of alendronate against placebo with both groups receiving mechanical therapy. A table describing changes in antidiabetic therapy has been included (Additional Table 2) Table 2. Changes in hypoglycaemic medication pre- and during intervention Author Pre-intervention During intervention Al-Mubarak Subjects were controlled by oral hypoglycaemic agents, insulin and diet in combination or separately and had been on the same type and dose of diabetic medication for the past 6 months. Patterns of treatment not compared between groups. No comment. Rocha Not described. No comment. Jones Pattern of treatment similar in both groups (x2 = 50.89, P = 0.64) for proportions receiving insulin, insulin and oral hypoglycaemics, oral hypoglycaemics alone. Participants in the usual care group were twice as likely (20% versus 11%, P < 0.12) to increase insulin from baseline to 4 months and less likely to decrease insulin (1% versus 6%, P < 0.21). Grossi Randomisation stratified by insulin use. Dose and type of medication monitored. Most changes were to other oral agents (not described in detail). 2 people in each group were started on insulin. Similar results found when people whose treatment had been changed were excluded (but actual data not given). Rodrigues People treated with insulin were excluded. No comment. Kiran Not described. No changes made in treatment during study. Yun Well-matched initially for oral hypoglycaemic medication and proportion prescribed diet control. No comment. Excluded studies Characteristics of excluded studies. In the studies that we excluded after examination of the fulltext papers, the most common design used was that of the interrupted time series study ( before and after treatment study). A trial (Campus 2006) was excluded because it had no relevant primary outcome as was Williams 1960 (which monitored the effect on in- 12

15 sulin requirement subsequent to periodontal treatment on insulin dependent diabetic patients). One study (Wolf 1977) did not have a control group and no stated primary outcome of interest. The following used HbA1c as a primary outcome, but had no control group: Sastrowijoto 1990; Miller 1992; Safkan-Seppala 1992; Iwamoto 2001; Campus 2006). Smith 1996 was excluded because the test group was matched to a healthy (and historical) control group. In addition, Westfelt 1996; Tervonen 1997; Christgau 1998; Sandberg 2000 and Faria-Almeida 2006 matched diabetic patients to healthy controls with both groups receiving a similar intervention. Seppala 1994 was a cohort study with poorly-controlled and well-controlled groups of patients with diabetes, both groups received the intervention. Almas 2003 had no randomisation and measured fasting blood glucose as the only outcome relating to metabolism. Martorelli De Lima 2004 used a splitmouth design, but had no relevant primary outcome. Attempts to contact the study authors to confirm that no unpublished HbA1c was available proved unsuccessful. Stewart 2001 did have parallel groups with a primary outcome, however no randomisation appeared to have been performed and the treatment groups were matched to historical controls. Aldridge 1995 comprised two studies both of which were randomised satisfactorily, but the follow-up period was less than the 3 months specified in the protocol in order to meet the requirement for inclusion. One study author (Watts) supplied detailed information about the methods employed but also confirmed no data had been collected beyond 2 months. Both primary outcomes were investigated (fructosamine and HbA1c). Promsudthi 2005 utilised a parallel design with allocation to periodontal treatment including antibiotic therapy in the intervention group and no periodontal treatment in the control group. Originally, this study was included by the review authors. However, with further discussion, concerns arose regarding the methods of randomisation. The authors had placed patients refusing to participate into the control group. After consultation with the Cochrane Oral Health Group editorial team we decided to exclude this study on the grounds that it was not a properly randomised trial. Ongoing studies Characteristics of ongoing studies. From the clinical trial registers (found on the World Wide Web at National Research Register and ClinicalTrials.gov), it was possible to identify two unpublished studies. The first trial (Tonetti 2007 (National Research Register identifier N found on the World Wide Web at National Research Register)) is a pilot study proposing to evaluate the efficacy of Agent A (minocycline HCl 1 mg (Arestin Oropharma)) in decreasing glycosylated haemoglobin levels in subjects who present with Type 2 diabetes and periodontitis. The project is listed as completed on the National Research Register although contact with the authors to verify this proved unsuccessful. The other study (Taylor 2007 (National Library for Medicine identifier NCT (NLM) - found on the World Wide Web at ClinicalTrials.gov) is proposed as a pilot study with 45 participants. Two groups were due to receive ultrasonic scaling with local anaesthesia (as needed), local antimicrobial treatment with povidone-iodine irrigation, and an oral systemic antibiotic (either doxycycline or metronidazole) at the initial treatment visit. The third group ( controls ) were intended to receive a placebo and supra-gingival oral prophylaxis and ultrasonic removal of supra-gingival calculus with water irrigation at the initial treatment visit. At the 9-month follow-up visit, controls receive subgingival ultrasonic scaling with povidone-iodine irrigation. Contact with the principal researcher (Taylor) in this study established that data gathering is complete, but the data were not ready for analysis at the point of publication of this review. One further study Hagiwara 2007 was found in the reference section of Rodrigues 2003 and published as a conference abstract. The principal author of this paper was contacted by and indicated that he hoped to publish primary outcome data at a later date. Initial indications suggest that the study design may not meet the criteria for inclusion in this review. Studies awaiting classification Characteristics of studies awaiting classification. Seventeen studies are currently under review and will require further investigation for inclusion in future updates of this review (Calbacho 2005; Chee 2006; Mansouri 2006; Schara 2006; Sonoki 2006; Navarro-Sanchez 2007; da Cruz 2008; Llambés 2008; Madden 2008; O Connell 2008; Al-Zahrani 2009; Dag 2009; Katagiri 2009; Lee 2009; Santos 2009; Spangler 2009; Vergnes 2009). As a result of scrutinising the abstracts of papers identified by the electronic search strategy, one study was found to have published secondary outcome data, but no primary outcome (Llambés 2005). After contacting the lead author it was established that primary outcome data related to this study will be published at a later date, but were not available for release at that time. The same author has now published data from a trial on people with Type 1 DM (Llambés 2008) and this study (which is probably from the same trial) is currently awaiting classification. Risk of bias in included studies Where information on methodological quality was not available in the text of the published report of the seven included studies, we attempted contact with the study authors. Only in the case of Al-Mubarak 2002 was no contact necessary due to sufficient information being available in the published report. Initial contact with the principal author in Grossi 1997 was successful, but later attempts at contact by did not meet with a reply. No replies were received from authors of Rocha Authors from the following studies - Rodrigues 2003 (Taba), Kiran 2005 (Unsal), Jones 2007 (Jones) Yun 2007 (Firkova) supplied additional 13

16 information and this added to our assessment of the papers and is indicated in the text below. Randomisation was used in all the included studies. Adequate descriptions were provided in three studies (Al-Mubarak 2002; Rodrigues 2003 (details supplied by the author); Jones 2007 (clarified in correspondence with the principal author)). Adequate description of sequence generation was unavailable in four trials (Grossi 1997; Rocha 2001; Kiran 2005; Yun 2007). Concealment of allocation was adequate in two studies (Al- Mubarak 2002; Jones 2007). No description of concealment was available in Grossi 1997; Rocha 2001; Rodrigues 2003; Kiran 2005; Yun The principal author in Jones 2007 described the method of allocation in correspondence. All included studies had primary outcomes analysed by laboratories remote from the clinical setting. Authors in Kiran 2005 and Jones 2007 confirmed that the primary outcomes assessor was masked, with no other author supplying similar details. All studies employed a single outcome assessor for periodontal parameters. Masking of the outcomes assessor for periodontal therapy was present in five studies (Grossi 1997; Rocha 2001; Al-Mubarak 2002; Kiran 2005 (details supplied by an author); Jones 2007 (details supplied by the principal author)). In the other two studies it was unclear if there was blinding (Rodrigues 2003; Yun 2007). Blinding of the clinical operator was present in three studies (Al-Mubarak 2002; Rodrigues 2003; Kiran 2005 (details supplied by an author)). It is unclear if the individuals carrying out the periodontal treatment were blinded in Grossi 1997; Rocha 2001; Jones 2007 or Yun A clear description of losses and withdrawals was given in two studies (Al-Mubarak 2002; Jones 2007). A partial description (four of 12 losses are attributed to changes in the method of diabetes control) is given in Grossi 1997, and with either no reported losses or description of these losses in: Rocha 2001; Rodrigues 2003; Kiran 2005; Yun Risk of bias assessment Risk of bias was assessed for each study undertaken independently by two review authors (Terry Simpson (TS) and Sarah Wild (SW)). Two rounds of reviewing the criteria for the domains for the risk of bias assessments were made independently by two review authors; with discussion before the second review. Kappa scores were calculated for four domains for both rounds, and an overall kappa. Details are provided in Appendix 4. Of the three studies (Kiran 2005; Jones 2007; Yun 2007) which addressed the primary hypothesis, only one (Jones 2007) could be considered to be at low risk of bias. Of the remaining four studies (Grossi 1997; Rocha 2001; Al-Mubarak 2002; Rodrigues 2003) only Al-Mubarak 2002 could be considered to be at low risk of bias. These results are presented in Figure 1 and Figure 2. Effects of interventions See: Summary of findings for the main comparison Scaling/root planing and oral hygiene (+/- antibiotic therapy) versus no treatment/usual treatment HbA1c after 3/4 months (Comparison 1 Outcome 1.1) Figure 3. Three studies all involved either usual treatment (Jones 2007) or no treatment (Kiran 2005 and Yun 2007) in the control group and in our judgement were sufficiently similar to have their results pooled into a meta-analysis. (Jones 2007 included patients who were allowed to visit their own dentist ( usual treatment ) with the two other studies having either no treatment or delayed treatment. Jones 2007 also presented data for adjusted (n = 154) and non-adjusted analyses (n = 165) of the changes in HbA1c. The adjusted score is included in the meta-analysis in accordance with guidance given in the Cochrane Handbook for Systematic Reviews of Interventions Section (Higgins 2009). The results from the unadjusted data were nearly identical.the effect for the mean percentage difference reduced slightly to -0.39% (95% confidence interval (CI) fixed effect -0.76% to -0.01%), n = 255, representing a statistically significant reduction in HbA1c (P = 0.04) for scaling/ root planing. Heterogeneity tests showed provided no evidence of a different treatment effect between the three studies (Chi 2 for heterogeneity = 1.05, degrees of freedom (df) = 2, P = 0.59, I 2 = 0%). The unadjusted data took account of all randomised patients less those excluded prior to starting the study through death, illness etc. The adjusted figure includes 22 people who failed to complete the study (n = 132).) 14

17 Figure 3. Forest plot of comparison: 1 Scaling, root planing and oral hygiene (+/- adjunctive antibiotics) versus no/usual treatment, outcome: 1.1 HbA1c after 3/4 months. The effect for the mean percentage difference was -0.40% (95% CI fixed effect -0.78% to -0.01%), n = 244, representing a statistically significant reduction in HbA1c (P = 0.04) for scaling/root planing. Heterogeneity tests showed provided no evidence of a different treatment effect between the three studies (Chi 2 for heterogeneity = 1.04, df = 2, P = 0.60, I 2 = 0%). Jones 2007 also reported changes of insulin requirement by 15% over the 4-month interval for both groups. The results for this comparison are presented as three subgroups based on whether an antibiotic was not used (Kiran 2005), or was used adjunctively in the test group (Jones 2007), or was used in both test and control groups (Yun 2007). The comparison was not significant for either subgroup (P = 0.09; P = 0.14; P = 0.74). Clinical attachment level (CAL) after 3/4 months (Comparison 1 Outcome 1.2) Figure 4. We considered all three studies where they provided data however one study (Jones 2007) did not report CAL, so the results were based on the other two studies (Kiran 2005; Yun 2007). The overall result for CAL showed an improvement in CAL of mm (95% CI mm to 0.49 mm), n = 90. This was not statistically significant in favour of scaling/root planing (P = 0.98). (Chi 2 for heterogeneity = 0, df = 1, P = 0.98, I 2 = 0% with little evidence of heterogeneity). 15