Probiotics for managing caries and periodontitis: Systematic review and meta-analysis

Abstract

Objectives

Probiotics might be beneficial to prevent or treat caries, gingivitis or periodontitis. We aimed to appraise trials assessing probiotics for managing caries and periodontal disease.

Data

We included randomized controlled trials comparing the efficacy of probiotics versus (placebo) control with regards to Streptococcus mutans [SM], lactobacilli [LB], periodontal pathogens numbers, gingivitis, oral hygiene, caries incidence/experience increment, or periodontitis. Meta-analysis and trial-sequential-analysis were performed.

Sources

Three electronic databases (Medline, Embase, Central) were screened.

Study selection

50 studies (3247 participants) were included. Studies were mainly performed in children and used lactobacilli (45); bifidobacteria (12) or other genus (3). Probiotics significantly increased the chance of reducing SM (OR: 2.20, 95% CI: 1.23/3.92) or LB (OR: 2.84; 1.34/6.03) <10 4 CFU/ml. Such reduction was confirmed for SM counts (standardized mean differences: −1.18, 95% CI: −1.64/-0.72), but not LB (SMD: 0.33; 0.15/0.52). For periodontal pathogens, no significant difference was found. Probiotics significantly reduced bleeding-on-probing (SMD: −1.15; −1.68/-0.62) and gingival index (SMD: −0.86; −1.52/-0.20), but not plaque index (SMD: −0.34; −0.89/0.21). Caries incidence was not significantly reduced (OR: 0.60; 0.35/1.04), neither was caries experience (SMD: −0.26; −0.55/0.03) or CAL (SMD: −0.46; −0.84/0.08). In contrast, probing-pocket depths (SMD: −0.86; −1.55/-0.17) were significantly reduced. Data was quantitatively insufficient for conclusive findings, and risk of bias was high.

Conclusion

Current evidence is insufficient for recommending probiotics for managing dental caries, but s upportive towards managing gingivitis or periodontitis. Future studies should only record bacterial numbers alongside accepted disease markers or indicators.

Clinical significance

Probiotic therapy could be used for managing periodontal diseases. For caries, further studies should ascertain both efficacy and safety.

Introduction

Probiotics are defined as “live microorganisms which, when administered in adequate amounts, confer a health benefit on the host” . Currently, antibacterial effects (e.g. via co-aggregation, toxic by-products or competition for substrates), stabilization of the flora and modulation of the host’s immune system are thought to provide these benefits. A range of bacteria (most of them being acidogenic like lactobacilli, streptococci or bifidobacteria) have been suggested to exert one or more of these effects .

Probiotics might be beneficial to prevent or treat oral diseases like caries, gingivitis or periodontitis, which are associated with a shift in the bacterial biofilm composition and activity as well as subsequent host reactions. Potential effects of probiotic species on cariogenic or periodontal pathogens have been abundantly demonstrated in vitro . Clinically, alterations of surrogate markers like bacterial numbers have been used to argue for the benefits of probiotic therapy. Only few studies, however, have used indicators of the diseases themselves (increment of newly developed caries lesions or caries experience; probing-pocket depths or clinical attachment loss) to prove the efficacy of probiotics for preventing or treating caries or periodontitis . Moreover, some studies even claim probiotics to not have beneficial, but potentially harmful effects .

Recent systematic reviews in the field have either qualitatively summarized selected studies on either caries or periodontitis , or meta-analyzed available surrogate markers . No study so far has attempted to comprehensively display the available evidence from randomized controlled trials on effects and efficacy of probiotics on both caries and periodontal disease using both qualitative and quantitative synthesis. Moreover, no study investigated potential causes for heterogeneity between studies, i.e. assessed the role of potential effect modifying variables. The present study aimed to systematically review and synthesize available randomized controlled studies investigating effects of probiotics on oral caries or periodontal disease (gingivitis and periodontitis). The results of this review should be useful to guide clinical decision-making and further research in the field.

Materials and methods

This review follows international guidelines for performing and reporting systematic reviews . The study protocol was registered after the screening stage (PROSPERO CRD42015026138). We deviated from this original protocol by only assessing outcomes at the last recorded visit, not separately after the intervention and after conclusion of follow-up. This was done, as not at all studies had a follow-up period. Moreover, periodontal pathogen numbers were analysed separately for each species (not pooled as planned). The following review question was addressed: In humans, what effects do probiotics exert on caries or periodontal disease, assessed via bacterial numbers, gingival or periodontal health, or dental caries incidence and increment, compared with placebo or alternative treatments?

Eligibility criteria

We included randomized controlled trials (RCTs) published in 1967 or later, without any language restrictions, which reported on dentate humans who consumed oral probiotics, regardless of the way of consumption or the probiotic species. Inclusion criteria were chosen that broad, as researchers and clinicians will be interested in the (comparative) efficacy of all available bacteria, not only a specific species or strain. The control intervention could have been placebo or alternative treatments (chlorhexidine, xylitol). No further specification was used to be as sensitive as possible. However, only studies allowing to determine the additional effect of probiotics were included (e.g. studies comparing probiotics plus xylitol against only xylitol were included, while those comparing probiotics plus xylitol against placebo were not). One of the following outcomes needed to be assessed: Bacterial numbers ( Streptococcus mutans [SM], lactobacilli [LB], Aggregatibacter actinomycementcomitans [AA], Porphyromonas gingivalis [PG], Prevotella intermedia [PrI]); oral hygiene and gingival health (Gingiva Index [GI], Plaque Index [PI], Bleeding on Probing [BOP]), caries (caries or caries experience prevalence, i.e. DMFT/dmft > 0 or DT/dt > 0; caries experience or its increment); periodontitis (probing pocket depths [PPD], clinical attachment loss [CAL]).

Search strategy and study selection

Identification of studies was based on a search strategy for each electronic database (Cochrane Central Register of Controlled Trials, Medline via PubMed, Embase); the search was carried out on September 29th 2014 and last updated June 1st 2015. Screening procedures used a three-pronged approach without controlled vocabularies (MeSH) being used ( Fig. 1 ). Cross-referencing from retrieved full-text studies was used to identify further articles. Neither authors nor journals were blinded to reviewers. Title and abstract of identified studies were screened independently by two calibrated reviewers (FS, DG). Calibration with regards to possible eligibility and inclusion of studies was performed on a subset of 20 studies prior to searching all databases. Consensus was obtained by discussion or consulting a third reviewer (SP).

Fig. 1
Flow of the search.

Data collection

Data from eligible studies was independently extracted by two reviewers (FS, DG) using piloted electronic spreadsheets. Data was recorded according to guidelines outlined by the Cochrane Collaboration . If data was missing, we contacted authors via e-mails, and sent reminders after 2 weeks.

Data items

The following items were recorded: Year, type and setting of study; age, size and recruitment of sample; test and control interventions including probiotic species and any pre-treatment (mechanical or chemical disinfection), vehicle, total dose (daily dose multiplied with consumption time in colony-forming units [CFU]), frequency and length of consumption; possible wash-out period in cross-over trials; follow-up; drop-out and sample size at follow-up. Outcomes were recorded as follow: bacterial numbers on ordinal or continuous scale, periodontal health and caries (experience)-increment on continuous scale, caries (experience) prevalence (e.g. DT > 0, DMFT > 0) dichotomously. Outcome data was extracted from the last follow-up visit (i.e. either after the intervention if patients were not followed further, or after conclusion of follow-up).

For ordinal records of bacterial numbers, we assessed how many patients experienced a reduction below the threshold of 10 4 CFU/ml. The latter was varied in sensitivity analyses, as we wanted to avoid to arbitrarily set a threshold. Moreover, such analyses all served to assess the effect of threshold choice. Note that none of these thresholds has a specific clinical relevance. For all other outcomes, post-intervention parameters in test and control group were compared.

Risk of bias in individual studies

Selection bias (sequence generation, allocation concealment), performance and detection bias (blinding of participants, operators, examiners), attrition bias (loss-to-follow-up and missing values or participants) and reporting bias (selective reporting, unclear withdrawals, missing outcomes) were recorded, assessed and classified according to Cochrane guidelines . For cross-over studies, separate aspects (e.g. wash-out period and risk of carry-over) were additionally recorded.

Summary measures and synthesis

The unit of analysis for meta-analysis was the patient. Comparisons were only made between groups measuring the additional effect of probiotics. In studies reporting on more than two interventions, groups were combined if possible to avoid unit-of-analysis conflicts. If studies employed a factorial design (e.g. probiotic with or without fluoride, placebo with or without fluoride), comparisons between comparable groups were separately entered into meta-analysis as study subgroups. If required, control groups were divided for meta-analysis (this was only possible for count outcomes, not continuous data). Alternatively, only those groups with the largest difference between test and control were entered. Individuals in cross-over trials were treated as independent, i.e. results from the first and second period were treated as if they came from different groups of patients. Note that this ignores any within-patient correlation, for which paired analyses are recommended. This approach was chosen as paired data were not reported by most trials, and insufficient detail available to reconstruct paired estimates. The chosen approach does not introduce bias, but is overly conservative, i.e. leads to under-weighting of studies and confidence intervals being wider than when paired data were used . Given the limited number of cross-over per all trials and the relatively large numbers of recorded outcomes, we avoided separate reporting of parallel and cross-over trials.

Meta-analysis was performed using random-effect model via Comprehensive Meta-Analysis 2.2.64 (Biostat, Englewood, NJ, USA), with Odds Ratios (OR) or Standardized Mean Differences (SMD) and 95% confidence intervals (95% CI) being calculated as effect estimates. Meta-regression was performed for comparisons with min. 10 studies included to assess the impact of pre-treatment disinfection (yes/no), total dose (in CFU), mono versus mixed species probiotic therapy, and total intervention and follow-up period (in months) on effect estimates. Missing co-variables for meta-regression were imputed by means, the effect of which was tested by sensitivity analysis. For meta-regression, the unrestricted maximum-likelihood method was used, and Bonferroni adjustments applied to correct for multiple testing . For each outcome variable, subgroups of different bacterial genus were additionally displayed.

Consistency measures and risk of bias across studies

Heterogeneity was assessed quantitatively using Cochran’s Q and I 2 -statistics . Funnel plot analysis and Egger test were performed to assess small study effects or publication bias for analyses with five or more trials being present . OR or SMD were adjusted (ORa, SMDa) to check the impact of possible publication bias .

Evidence grading and trial sequential analysis

Evidence for estimates was graded using Grade Profiler 3.6 according to GRADE guidelines (Atkins et al., 2004). In addition, trial sequential analysis (TSA) was performed to assess if quantitative findings were robust, and to calculate the required information size (RIS), i.e. the cumulative sample size needed to yield significant differences between probiotic and control therapy . RIS was calculated based on type I error risk of α=0.05 and a type II error risk of β = 0.20 (equivalent to a power of 0.80). For dichotomous outcomes, the control event proportion (i.e. event incidence in control group) and the relative risk reduction (RRR) were used to estimate RIS, while empirical estimates of effect and variance were used for continuous outcomes. In this review, RRR was based on an a priori defined worthwhile interventional effect of 20%, while smaller effects might well be relevant, but would increase RIS even further. RIS was adjusted for the diversity in the meta-analysis (DARIS). To assess if differences yielded by conventional meta-analysis are robust, TSA additionally estimates trial sequential monitoring boundaries (TSMB), i.e. statistical thresholds for significance which are adapted depending on the so far reached sample size. Effect estimates supported by only few small trials are thus handled stricter than those supported by large samples. Further details regarding the applied method to calculate TSMB have been reported elsewhere . TSA was performed with TSA 0.9 (Copenhagen Trial Unit, Copenhagen, Denmark) .

Materials and methods

This review follows international guidelines for performing and reporting systematic reviews . The study protocol was registered after the screening stage (PROSPERO CRD42015026138). We deviated from this original protocol by only assessing outcomes at the last recorded visit, not separately after the intervention and after conclusion of follow-up. This was done, as not at all studies had a follow-up period. Moreover, periodontal pathogen numbers were analysed separately for each species (not pooled as planned). The following review question was addressed: In humans, what effects do probiotics exert on caries or periodontal disease, assessed via bacterial numbers, gingival or periodontal health, or dental caries incidence and increment, compared with placebo or alternative treatments?

Eligibility criteria

We included randomized controlled trials (RCTs) published in 1967 or later, without any language restrictions, which reported on dentate humans who consumed oral probiotics, regardless of the way of consumption or the probiotic species. Inclusion criteria were chosen that broad, as researchers and clinicians will be interested in the (comparative) efficacy of all available bacteria, not only a specific species or strain. The control intervention could have been placebo or alternative treatments (chlorhexidine, xylitol). No further specification was used to be as sensitive as possible. However, only studies allowing to determine the additional effect of probiotics were included (e.g. studies comparing probiotics plus xylitol against only xylitol were included, while those comparing probiotics plus xylitol against placebo were not). One of the following outcomes needed to be assessed: Bacterial numbers ( Streptococcus mutans [SM], lactobacilli [LB], Aggregatibacter actinomycementcomitans [AA], Porphyromonas gingivalis [PG], Prevotella intermedia [PrI]); oral hygiene and gingival health (Gingiva Index [GI], Plaque Index [PI], Bleeding on Probing [BOP]), caries (caries or caries experience prevalence, i.e. DMFT/dmft > 0 or DT/dt > 0; caries experience or its increment); periodontitis (probing pocket depths [PPD], clinical attachment loss [CAL]).

Search strategy and study selection

Identification of studies was based on a search strategy for each electronic database (Cochrane Central Register of Controlled Trials, Medline via PubMed, Embase); the search was carried out on September 29th 2014 and last updated June 1st 2015. Screening procedures used a three-pronged approach without controlled vocabularies (MeSH) being used ( Fig. 1 ). Cross-referencing from retrieved full-text studies was used to identify further articles. Neither authors nor journals were blinded to reviewers. Title and abstract of identified studies were screened independently by two calibrated reviewers (FS, DG). Calibration with regards to possible eligibility and inclusion of studies was performed on a subset of 20 studies prior to searching all databases. Consensus was obtained by discussion or consulting a third reviewer (SP).

Fig. 1
Flow of the search.

Data collection

Data from eligible studies was independently extracted by two reviewers (FS, DG) using piloted electronic spreadsheets. Data was recorded according to guidelines outlined by the Cochrane Collaboration . If data was missing, we contacted authors via e-mails, and sent reminders after 2 weeks.

Data items

The following items were recorded: Year, type and setting of study; age, size and recruitment of sample; test and control interventions including probiotic species and any pre-treatment (mechanical or chemical disinfection), vehicle, total dose (daily dose multiplied with consumption time in colony-forming units [CFU]), frequency and length of consumption; possible wash-out period in cross-over trials; follow-up; drop-out and sample size at follow-up. Outcomes were recorded as follow: bacterial numbers on ordinal or continuous scale, periodontal health and caries (experience)-increment on continuous scale, caries (experience) prevalence (e.g. DT > 0, DMFT > 0) dichotomously. Outcome data was extracted from the last follow-up visit (i.e. either after the intervention if patients were not followed further, or after conclusion of follow-up).

For ordinal records of bacterial numbers, we assessed how many patients experienced a reduction below the threshold of 10 4 CFU/ml. The latter was varied in sensitivity analyses, as we wanted to avoid to arbitrarily set a threshold. Moreover, such analyses all served to assess the effect of threshold choice. Note that none of these thresholds has a specific clinical relevance. For all other outcomes, post-intervention parameters in test and control group were compared.

Risk of bias in individual studies

Selection bias (sequence generation, allocation concealment), performance and detection bias (blinding of participants, operators, examiners), attrition bias (loss-to-follow-up and missing values or participants) and reporting bias (selective reporting, unclear withdrawals, missing outcomes) were recorded, assessed and classified according to Cochrane guidelines . For cross-over studies, separate aspects (e.g. wash-out period and risk of carry-over) were additionally recorded.

Summary measures and synthesis

The unit of analysis for meta-analysis was the patient. Comparisons were only made between groups measuring the additional effect of probiotics. In studies reporting on more than two interventions, groups were combined if possible to avoid unit-of-analysis conflicts. If studies employed a factorial design (e.g. probiotic with or without fluoride, placebo with or without fluoride), comparisons between comparable groups were separately entered into meta-analysis as study subgroups. If required, control groups were divided for meta-analysis (this was only possible for count outcomes, not continuous data). Alternatively, only those groups with the largest difference between test and control were entered. Individuals in cross-over trials were treated as independent, i.e. results from the first and second period were treated as if they came from different groups of patients. Note that this ignores any within-patient correlation, for which paired analyses are recommended. This approach was chosen as paired data were not reported by most trials, and insufficient detail available to reconstruct paired estimates. The chosen approach does not introduce bias, but is overly conservative, i.e. leads to under-weighting of studies and confidence intervals being wider than when paired data were used . Given the limited number of cross-over per all trials and the relatively large numbers of recorded outcomes, we avoided separate reporting of parallel and cross-over trials.

Meta-analysis was performed using random-effect model via Comprehensive Meta-Analysis 2.2.64 (Biostat, Englewood, NJ, USA), with Odds Ratios (OR) or Standardized Mean Differences (SMD) and 95% confidence intervals (95% CI) being calculated as effect estimates. Meta-regression was performed for comparisons with min. 10 studies included to assess the impact of pre-treatment disinfection (yes/no), total dose (in CFU), mono versus mixed species probiotic therapy, and total intervention and follow-up period (in months) on effect estimates. Missing co-variables for meta-regression were imputed by means, the effect of which was tested by sensitivity analysis. For meta-regression, the unrestricted maximum-likelihood method was used, and Bonferroni adjustments applied to correct for multiple testing . For each outcome variable, subgroups of different bacterial genus were additionally displayed.

Consistency measures and risk of bias across studies

Heterogeneity was assessed quantitatively using Cochran’s Q and I 2 -statistics . Funnel plot analysis and Egger test were performed to assess small study effects or publication bias for analyses with five or more trials being present . OR or SMD were adjusted (ORa, SMDa) to check the impact of possible publication bias .

Evidence grading and trial sequential analysis

Evidence for estimates was graded using Grade Profiler 3.6 according to GRADE guidelines (Atkins et al., 2004). In addition, trial sequential analysis (TSA) was performed to assess if quantitative findings were robust, and to calculate the required information size (RIS), i.e. the cumulative sample size needed to yield significant differences between probiotic and control therapy . RIS was calculated based on type I error risk of α=0.05 and a type II error risk of β = 0.20 (equivalent to a power of 0.80). For dichotomous outcomes, the control event proportion (i.e. event incidence in control group) and the relative risk reduction (RRR) were used to estimate RIS, while empirical estimates of effect and variance were used for continuous outcomes. In this review, RRR was based on an a priori defined worthwhile interventional effect of 20%, while smaller effects might well be relevant, but would increase RIS even further. RIS was adjusted for the diversity in the meta-analysis (DARIS). To assess if differences yielded by conventional meta-analysis are robust, TSA additionally estimates trial sequential monitoring boundaries (TSMB), i.e. statistical thresholds for significance which are adapted depending on the so far reached sample size. Effect estimates supported by only few small trials are thus handled stricter than those supported by large samples. Further details regarding the applied method to calculate TSMB have been reported elsewhere . TSA was performed with TSA 0.9 (Copenhagen Trial Unit, Copenhagen, Denmark) .

Only gold members can continue reading. Log In or Register to continue

Stay updated, free dental videos. Join our Telegram channel

Jun 19, 2018 | Posted by in General Dentistry | Comments Off on Probiotics for managing caries and periodontitis: Systematic review and meta-analysis

VIDEdental - Online dental courses

Get VIDEdental app for watching clinical videos