Is diabetes mellitus associated with reduced vertical and horizontal ridge augmentation in individuals receiving dental implant treatment and GBR?

Population: Individuals aged 18 years or older requiring the placement of one or more implants in sites with horizontal and/or vertical ridge deficiencies necessitating GBR (one-stage or staged approaches) in either jaw. No restriction was imposed regarding techniques or employed biomaterials.

Exposure: Type 1 or type 2 diabetic individuals (based on study criteria);

Comparator: Individuals without a previous diabetes history;

Outcomes: The primary outcomes focused on changes in vertical (height) and/or horizontal (width) dimension of the ridge, assessed clinically or through cone-beam computed tomography (CBCT) using linear or volumetric measurements.

Time: At least four months of follow-up post-augmentation/implantation procedures;

Study: Either non-randomized or randomized controlled trials (N-RCT and RCT).

We conducted a comprehensive and systematic search for relevant studies published up to 20 September 2023. The US National Library of Medicine (MEDLINE-PubMed), EMBASE, Cochrane Central Register of Controlled Trials (CENTRAL), and SCOPUS databases were investigated. We also searched Google Scholar (the first 300 most relevant hits) for grey literature. Additionally, the reference lists of the included studies were hand-searched to identify additional relevant studies. No language and date restrictions were imposed.

To construct our search strategy, we first consulted some systematic reviews addressing similar research questions to retrieve important search terms (11–13). Then, we applied a structured search strategy using a combination of controlled vocabulary (MeSH) and free text terms to retrieve relevant papers. The search strategies for the different databases are displayed in Appendix S1.

We performed a literature search using a two-step process by two independent reviewers (LMO and FBZ). Firstly, we screened record titles and abstracts retrieved from the searches. The Mendeley Desktop 1.19.4 (England) was used to manage the references and identify duplicates. Then, the retrieved records were classified as “include”, “exclude” or “uncertain”. Following this initial screening, we thoroughly read the full texts of papers identified as relevant. Studies satisfying all eligibility criteria were included and underwent data extraction. Divergences between the reviewers were resolved by discussion. In case of persistent disagreement, the judgment of an additional reviewer (RPA) was considered decisive.

The same reviewers (LMO and FBZ) independently collected data from eligible studies using a pilot-tested Excel form (Microsoft, Redmond, Washington, United States). The collected data included study identification (first author, year of publication, and location); study design; sources of funding; duration; participants' characteristics (sex, age, smoking, and periodontal status); diabetes information (diagnosis); augmentation site and assessment method; surgical protocol; and baseline and follow-up measurements of vertical and horizontal ridge dimension.

The risk of bias was assessed independently by the same two reviewers (LMO and FBZ) using the Risk of Bias in Non-Randomized Studies of Interventions (ROBINS-I) tool (20). Each study was graded according to seven domains: (1) confounding (especially sex, age, smoking, anatomical characteristics of the site of augmentation and [glycated hemoglobin (HbA1c) levels during the follow-up]; (2) selection of participants; (3) classification of interventions (considering potential misclassification due to the method of measuring diabetes); (4) deviations from intended interventions (exposures); (5) handling of missing data; (6) measurement of outcomes; and (7) selection of the reported result. Within each domain, we judged the risk of bias as “low” (comparable to a well-performed randomized clinical trial), “moderate” (sound for a non-randomized study), “serious” (indicating important concerns), “critical” (the study is too problematic to provide useful evidence) or “no information”. We thus rated the overall risk of bias for each study based on the most serious risk-of-bias judgment across any of the seven domains (overall risk of bias is serious if at least one domain is rated serious). Any disagreements were resolved through discussion. An additional reviewer's (RPA) judgment was considered decisive if a disagreement persisted. No automation tool was used. We created the summary risk-of-bias plot using the Risk-of-Bias VISualization tool (robvis).

First, we summarized the data in evidence tables to outline the study characteristics. Thereafter, we conducted pair-wise comparisons using data from included studies to compare vertical and horizontal bone gain in diabetic and non-diabetic individuals. When change scores were not reported, we calculated by subtracting the final from baseline scores. We estimated the standard deviation (SD) changes using the formula SD_delta = √[(SD_baseline)²+ (SD_final)² – (2*r*SD_baseline*SD_final)], in which we assumed r = 0.5 (16).

The included trials varied in protocols of dental implant placement and bone gain assessments. Thus, we used a random-effects model with the DerSimonian and Laird variance estimator (21) to determine the Hedges'g as standardized mean difference (SMD) and 95% confidence intervals (CI) as the summary estimate. All tests were two-tailed, with a significance level set at 0.05, except the Cochran's Q test (significance level at 0.1). We used the I² statistic to assess heterogeneity, where values higher than 75% indicated high heterogeneity (16).

For outcomes with three or more studies, we estimated prediction intervals for pooled SMD to provide a range of expected effects. Additionally, sensitivity analyses were performed to identify any individual study significantly affecting the pooled result using the “leave one out” approach. All analyses were conducted using the Stata software, version 14.0 (Stata Corporation; College Station, TX, USA). However, due to the low number of included studies (<10), we were unable to explore heterogeneity with meta-regression and to quantify publication bias through statistical evaluation (Egger statistics).

We assessed the certainty using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach (22). We created a summary-of-findings table including the evidence profile (number of studies, assessed domains, and interpretation). Estimates of vertical and horizontal bone gain according to each comparison group were reported as SMD with 95% CI, along with the overall GRADE score.

The assessment of GRADE domains involved the levels of concern “serious”, “very serious” or “no concerns” for each domain. The key domains considered were risk of bias (according to ROBINS-I), inconsistency (evaluating heterogeneity in observed effects across studies and its potential explanations), imprecision (assessing whether confidence intervals led to conflicting interpretations), indirectness (differences between the characteristics of included studies and the research focused question), publication bias, and upgrading domains (according to the research focused question, only “large effect size” was considered).

The included studies provided data of 42 individuals with type 2 diabetes and 45 individuals without. Current smokers were excluded in all three included trials. Specifically, two studies (18, 19) provided information on HbA1c levels of unexposed groups, while one study (17) provided details on diabetes duration. Diabetics were well controlled in two studies (18, 19) and moderately or well controlled in one (17). Additionally, one study clearly outlined the augmentation site (17). All patients across the studies received prophylactic antibiotic coverage and underwent similar regenerative procedures involving a 50% autologous bone and 50% synthetic bone substitute, associated with resorbable membranes. Notably, staged GBR was performed in two of the studies (17, 18).

The summary of findings using the evidence profile format is presented in Table 1.

The meta-analysis for changes in the horizontal ridge dimension contributed to data from all three N-RCTs (17–19). The pooled standardized mean difference calculated using a random-effects model showed no statistically significant difference between individuals with and without diabetes (SMD = −0.41, 95% CI: −0.92–0.10; I² = 28.3%). Our prediction intervals suggested that in approximately 95% of all populations, the true effect size will fall within the range from −4.92 to 4.10, as illustrated in Figure 2. Sensitivity analysis revealed a key study (17), which was particularly influential in overall results. Removing this study from the analysis could lead to a statistically significant change in results, potentially favoring the unexposed group (Figure 3). Regarding the meta-analysis for changes in the vertical ridge dimension, the combined results from two studies (18, 19) did not show a statistically significant difference between exposed and unexposed individuals (SMD = 0.06, 95% CI: −0.43 to 0.56; I² = 0.0%), as illustrated in Figure 4.

All three studies exhibited a serious risk of bias. The lack of information regarding the HbA1c levels in the unexposed group (17), specific details about the site and anatomical characteristics of the site of augmentation (18, 19), and HbA1c levels during the follow-up period indicated unmeasured confounding (17–19). Additionally, using a single assessment to categorize HbA1c levels posed a risk of misclassifying hyperglycemia, as possible variations over time were not accounted for. The possibility of contamination of the unexposed group with non-diabetic hyperglycemic subjects cannot be ruled out. Lastly, the absence of blinding among outcome assessors led to our assessment of serious bias in outcome measurement (17–19). A summary of the results is presented in Figure 5.

The summary of findings using the evidence profile format is presented in Table 2.

Considering the limited evidence addressing our focused question, it is imperative to conduct further investigations evaluating the impact of diabetes on GBR outcomes. Thus, future studies should improve on both reporting and methodological aspects. Considering that the defect anatomy predicts GBR outcomes (8), providing a comprehensive description of the augmentation site and detailed measurements is crucial. Additionally, continuous monitoring of HbA1c levels throughout the follow-up period is essential to evaluate the impact of metabolic control on outcomes. By embracing these best-practice methodologies and potentially integrating others, prospective research endeavors hold much promise when it comes to elevating the level of certainty in evidence, thereby contributing significantly to the delineation of clinical recommendations.