Vous êtes sur la page 1sur 10
Journal of Dentistry 43 (2015) 1298 – 1307 Contents lists available at ScienceDirect Journal of Dentistry

Contents lists available at ScienceDirect

Journal of Dentistry

homepage: www.intl.else vie rhealth.com/journals/jden Review article Antibacterial effects of cavity lining: A

Review article

Antibacterial effects of cavity lining: A systematic review and network meta-analysis

Falk Schwendicke a , * , Yu-Kang Tu b , Le-Yin Hsu b , Gerd Göstemeyer a

a Department of Operative and Preventive Dentistry, Charité Universitätsmedizin Berlin, A ßmannshauser Str. 4-6, 14197 Berlin, Germany b Institute of Epidemiology & Preventive Medicine, College of Public Health, National Taiwan University, Taipei, Taiwan

of Public Health, National Taiwan University, Taipei, Taiwan A R T I C L E I

A R T I C L E

I N F O

Article history:

Received 21 March 2015 Received in revised form 26 June 2015 Accepted 1 July 2015

Keywords:

Bacteria

Bayesian

Calcium hydroxide

Dental

Dental caries

Mineral Trioxide

Aggregate

A B S T R A C T

Objectives: Cavity liners are frequently used prior placing a restoration, with one main aim being to reduce the number of remaining bacteria. We systematically appraised studies comparing antibacterial effects of different liners against each other or no liner. Data studies: reporting the number of sterile cavities before/after lining or sealing, or the reduction in bacterial numbers (colony-forming-units) in two or more treatment groups were included. Treatments were categorized as: no/placebo liner, calcium hydroxide, mineral trioxide aggregate, antibiotic/ disinfectant, calcium phosphates, zinc oxide eugenol, black copper cement, and glass ionomer cement liners. Pairwise and network meta-analyses were performed. Study selection: From 113 identi ed studies, 14 (500 treated lesions) were included. Risk of bias was high or unclear. Based on 11 studies, network meta-analysis found mineral trioxide lining to yield the greatest probability of achieving sterile cavities after a lining/sealing period (73%), followed by antibiotic/ disinfectant (8%) and zinc oxide eugenol (7%). Only six studies assessed bacterial reduction after lining/ sealing, and zinc oxide eugenol was found to have the highest probability of achieving a bacterial reduction. In both analyses, not providing any lining was found to have low antibacterial effects. Conclusion: Within the limitations of this review and the included studies, certain liners seem more suitable to achieve sterile cavities or reduce bacterial numbers than others. Given the paucity of data and the unclear impact of remaining bacteria on clinical outcomes, further recommendations for speci c cavity treatments prior a restoration are not possible. Clinical signi cance: There is insuf cient evidence to generally recommend cavity lining or the use of any speci c liner based on their antibacterial effects. Dentists might continue to use liners, but should be aware that such use is not strongly supported by clinical studies. ã 2015 Elsevier Ltd. All rights reserved.

1. Introduction

For treating pulpo-proximal dentin after excavation of deeper caries lesions, the use of cavity liners has been recommended for decades [1] . As liners are thought to induce the development of reactionary dentin [2,3] , reduce post-operative pulpal in am- mation [4] , or isolate the pulp from chemical irritants like hydroxyethyl methacrylate [5] , they are commonly used for pulp protection. However, clinical studies do not necessarily support these arguments, with only few trials evaluating for example the

* Corresponding author at: Charité Centre for Dental Medicine, Department for Operative and Preventive Dentistry, Aßmannshauser Str. 4 6, 14197 Berlin, Germany. Fax: +49 30 450 7562 556. E-mail address: falk.schwendicke@charite.de (F. Schwendicke).

0300-5712/ ã 2015 Elsevier Ltd. All rights reserved.

risk of post-operative hypersensitivity or the need for endodontic follow-up treatments in teeth with versus without cavity lining [6 8] . A second reason why the use of liners has been advocated was their remineralizing effects, especially when selective (incomplete) or stepwise excavation was performed prior to restoration. Some studies demonstrated such remineralization induced by cavity liners [9,10] , whilst others indicate that such mineral gain in the residual caries lesion might well be mediated by the pulp, and does not seem to require the application of a liner prior to restoration [11,12] . Last, lining materials are used as they might reduce bacterial numbers, i.e. acting as cavity disinfection:

This has been especially postulated for the most widely used material, calcium hydroxide, whose alkaline pH is supposed to exert strong antibacterial effects [13 15] . Such effects have also been shown for other lining materials, for example glass ionomer cements [16,17] , whilst it remains unclear which material is most

F. Schwendicke et al. / Journal of Dentistry 43 (2015) 1298 1307

1299

suitable for reducing the bacterial load prior to restoration. More importantly, a changed understanding of the pathology of caries and the pathogenic effects of cariogenic bacteria has raised doubts regarding the need for such cavity disinfection via liners: a growing number of studies show that the sealing of the remaining bacteria via adhesive restorations has signi cant antibacterial effects [18] , and the relevance of an additional treatment of the cavity needs to be questioned [11,19 23] . Nevertheless, the majority of practical dentists rely on the use of liners, with one of the most cited reasons being the fear of remaining viable bacteria harming the pulp [24,25] . Given the discussed uncertainties, this study aimed at systematically appraising clinical trials investigating the antibacterial effects of different liners in comparison with each other or with restorations without any liner. Given the variety of available lining materials, we applied network meta-analysis for comparing the bacterial reductions in different treatment groups. Network meta-analysis allows indirect comparison and provides ranking of treatments according to their probability of having bene cial (or harmful) effects, in this case of having antibacterial effects. Based on this review, the most potent antibacterial treatment should be identi ed. The results of this review could then either be contrasted with clinical outcomes [26,27] to inform practitioners, or could be used for guiding future research regarding dental lining materials.

2. Materials and methods

This review followed international guidelines for performing and reporting systematic reviews and pairwise or network meta- analysis [28 30] , and assessed if, in human patients with cavities resulting from caries removal, cavity lining compared with another cavity liner or no liner has different effects on the remaining number of bacteria (sterility of the cavity, reduction of bacterial numbers within the dentin). Assessing how antibacterial effects translate into clinical outcomes (pain, clinical success) was beyond

the scope of this review. Similarly, we did not assess the effects of different antibacterial restoration materials, but focussed on the comparison of different liners or lining versus no lining.

2.1. Participants and intervention

Clinical trials investigating children or adults with primary caries lesions receiving operative treatment involving caries removal and restoration were included, with minimum two treatment groups comparing the antibacterial effects of different cavity treatment (i.e. different liners, or liner versus restoration without liner). We did not separate primary from permanent or anterior from posterior teeth, and did not de ne the depth or the location or extension of the lesion a priori. Treatment groups should differ only with regards to the lining/restorative material, whilst the treatment procedure like excavation should have been identical to limit potential biases introduced by including non-randomized trials.

2.1.1. Selection criteria

Whilst both randomized and non-randomized trials were included, only studies which allocated treatments independently from the cavity depth or the baseline bacterial load were eligible for inclusion to avoid selection bias by indication. Studies should have reported on the bacterial reduction associated with different cavity treatments, i.e. should have evaluated the bacterial load after excavation and after a certain effect period. Bacterial reduction should have been determined by assessing bacterial numbers, e.g. via cultivation or polymerase chain reaction, in dentin samples.

2.1.2. Search strategy

Three electronic database (Medline via PubMed, Embase, Cochrane Central Register of Controlled Trials) were searched at September 23rd 2014. Screening procedures used a three-pronged approach ( Fig. 1 ), and cross-referencing from the bibliographies of

Fig. 1 ), and cross-referencing from the bibliographies of Fig. 1. Screening process. Different domains (boxes)

Fig. 1. Screening process. Different domains (boxes) of the search sequence were combined using the Boolean operator AND (example for medline search).

1300

F. Schwendicke et al. / Journal of Dentistry 43 (2015) 1298 1307

full-text studies was used for identifying further articles. Grey literature was searched electronically (www.opengrey.eu), and ongoing trials screened using clinicaltrials.gov. No restrictions with regard to language or publication date were applied. Neither authors nor journals were blinded. Title and abstract of identi ed studies were screened by two calibrated reviewers (FS, GG) for eligibility. Consensus was obtained by discussion. Inclusion of studies was independently decided by both reviewers; consensus was again achieved by discussion.

2.1.3. Data extraction and evaluation

Duplicative data extraction was performed independently by two calibrated reviewers (FS, GG) using a jointly developed spreadsheet which had been piloted on ve studies a priori. Disagreement was solved by discussion. Data was recorded according to guidelines outlined by the Cochrane Collaboration [31] . We recorded setting, study type, source of funding, risk of bias, description of the sample, performed excavation (selective, stepwise, complete), treatment of the cavity oor, restoration, attrition, follow-up, and outcomes. Outcomes were recorded either dichotomously (number of positive samples) or continuously (reduction of colony forming units [CFUs]).

Networks constructed by plotting different treatments (as nodes) and comparisons (as edges) were inspected for geometry and asymmetry [32,33]. Random-effects pairwise meta-analysis were performed using STATA, with OR or SMD as effect sizes. STATA command metan was used to obtain the forest plot, and command netfunnel and networkplot were used to obtain the network funnel plots and the network plots. Network meta-analyses were performed using Bayesian random-effects models and a Markov Chain Monte Carlo simulation using Bayesian software package WinBugs and GeMTC 0.6 [34] implemented in R 3.0.3 (R foundation, Vienna, Austria). For the number of positive bacterial samples after different lining treatments, the Bayesian network meta-analysis rst used the binomial likelihood to model the data:

r j ; k Binomial ð p j ;k ; n j ; k Þ

where r j,k is the number of events in treatment group k of the jth trial, n j,k is the number of observations and P i,k the corresponding probability of the event for treatment k . Then it was modeled on the logit scale as

log it ð p j ;k Þ ¼ m j þ d j ; b ; k I ð k A Þ ; b > k

2.1.4. Risk of bias assessment

Selection bias (sequence generation, allocation concealment), performance and detection bias (blinding of operators, or participants and personnel), bias due to incomplete data, and reporting bias (selective reporting, unclear withdrawals, missing outcomes) were recorded, assessed and classi ed according to Cochrane guidelines [31] .

2.1.5. Outcome measures and statistical analysis

Our primary outcome was the number of positive bacterial dentin samples remaining in a cavity, i.e. the proportion of cavities with cultivable bacteria. Effects were estimated as odds ratio (OR), with the odds of harbouring positive bacterial samples (i.e. not being sterile) in the test compared with the control group being assessed. Superiority was de ned if a treatment yielded signi - cantly fewer positive samples than the comparator. Our secondary outcome was the reduction in the numbers of bacteria remaining in the cavity, with colony forming units as effect measure. For bacterial reduction, log CFU reported by some studies were back transformed into original values, and bacterial reduction calculat- ed as standardized mean differences (SMD: before and after lining) as effect estimate. Superiority was de ned if a treatment induced a signi cantly greater bacterial reduction than the comparator. The unit of analysis was the sample, i.e. the evaluated lesion or lesion site; no adjustment for the possible effects of clustering was undertaken as not all studies reported the number of treated patients for each group. Treatments were classi ed into the following categories being used: (1) No active liner/material, i.e. direct restoration using composite/amalgam/compomer, or (wax/gutta-percha) placebo being used as liner, (2) calcium hydroxide (both setting and not setting), (3) glass ionomer cement (standard or resin-modi ed GIC used as liner; in case GIC and other lining materials like antibiotics were mixed, we considered this treatment not as GIC, but the other lining material), (4) mineral trioxide aggregate (MTA), (5) calcium phosphates (hydroxyl apatite [HA] or tricalcium phosphate [TCP]), (6) antibiotic [ATB] or disinfectant (triclosan, thymol, chlorhexidine, stannous uoride), (7) black copper cement (BCC), (8) zinc oxide eugenol [ZOE] liner. Note that for comparison reasons, we summarized certain groups (e.g. calcium phosphates) and combined restorative materials and liners where applicable (e.g. for GIC). If the same study compared

the same group of treatment, but different products, only one product was chosen at random for statistical evaluation.

d j ; b ; k

N ð d Ak d Ab ; s 2 Þ

where b is the baseline treatment for the jth trial, d Ak is the estimated difference in treatment effect between A and k , and d Ab

is the estimated difference between A and b , The variance of d jbk is s 2 . For the bacterial reduction, a normal likelihood was used to model the data instead [35,36]. To t the model, we used a non-informative priors for the basic parameters from a normal distribution: N(0,10 4 ), and a at prior U(0,2) for the random-effects standard deviation. The convergence was assessed based on the Brooks Gelman Rubin criteria [37] and inspection of trace plots. The rst

50000

iterations were discarded as burn-in and then a further

50000

iterations were undertaken for 2 chains at a thinning

intervals of 5. We reported posterior median OR or SMD and their 95% credible intervals (95% CrI). Credible intervals are the range of estimated parameters after exclusion of extreme values [38] . Different treatments were ranked according to their probability of having the lowest versus the highest odds or differences [39] , and the surface under the cumulative ranking (SUCRA) line plotted and calculated. Heterogeneity within pairwise comparisons was assessed quantitatively using I 2 -statistics [40] . Loop inconsistency, i.e. the difference between direct and indirect estimates for three treatments within a loop, was evaluated by the inconsistency factor (IF) for the loop [41,42] . Within each loop, the IF value was de ned as IF = E direct E indirect ( E : estimate). We rejected the null hypothesis that the evidence is consistent (H 0 :IF = 0), when IF was signi cantly greater or smaller than 0. Funnel plot analysis was performed to assess small study effects or publication bias of pairwise estimates. Trim-and- ll was used to evaluate the effects

of such bias [43] . Given the limited available data and the lack of detailed reporting (see below), no sensitivity or subgroup analyses were performed to assess the potential impact of effect modi ers on our ndings [44] .

3. Results

3.1. Results of the search and risk of bias

Using electronic databases, 113 studies were found to be possibly eligible. 43 studies were analysed full-text, with 14 studies

F. Schwendicke et al. / Journal of Dentistry 43 (2015) 1298 1307

1301

Table 1 Included studies. A total of 14 studies were included, two studies [61,62] used the same control group and were thus combined for this analysis. The number of participants, type of teeth and lesions varied between studies, as d id the excavation and restoration procedures. Different groups of liner treatments were compared: no or placebo liner, calcium hydroxide (CH), mineral tr ioxide aggregate (MTA), glass ionomer cement (GIC), black copper cement (BCC), mineral liners like tri-calcium phosphate (TCP) or hydroxyl apatite (HA), antibiotic or disinfectant liners, zinc oxide eugenol (ZOE) liners. If th e same study compared the same group of treatment, but different products, only one group was chosen at random for statistical evaluation. Groups included in the quantitative analysis are highlighted in bold. Microbiologic evaluat ions used different methods and were performed after varying time periods.

Study

RCT

Participants (age) (n teeth)

Excavation,

Evaluation

Treatment(s) A (product), samples ( n )

Results treatment(s) A, samples ( n )

Treatment(s) B (product), samples ( n )

Results treatment(s) B, samples ( n )

Drop-out in

yes/

restoration and

method

different groups

no

sampling

 

(mean follow-up)

Bressani

yes

30 children (4 8) with vital and asymptomatic primary molars with occlusal or occluso-proximal active deep lesions (30) 44 patients (11 35) with one or more deep (>2/3 dentin) caries lesions on occlusal or- occluso-proximal surfaces in permanent vital and asymptomatic teeth (60) 20 children (4 8) with

SE, resin restoration (Scotchbond, Z250), re-entry and sampling after 3 mos

Microscopic

No liner (wax placebo) (15)

8/15 samples remained positive

CH (Dycal) (15) 7/15 samples remained positive

0/0 (3)

et al.,2013

analysis

 

[19]

 

Corallo and

yes

SE, ZOE or GIC restoration, re-entry

Microscopic

CH (Dycal) (20)

19/19 samples remained positive

No liner (wax placebo) (20)

19/19 samples remained positive

1/1/1 (3 4)

Maltz,

analysis

2013

[11]

and sampling after 3

 
 

4

mos

Duque et al.,

yes

SE, ZOE restoration,

Microaerophilic cultures on BHI, MRS, TSB agar

GIC (Vitrebond [10] or Fuji Lining [10])

8/19 MS and 5/19 LB samples remained positive, median difference MS: 3388441, range: 138 120226443 ( n = 19) and 55, range 1 11749 ( n = 19); median difference LB: 100, range 0 15488166 ( n = 19) and 10, range 0 4677351 ( n = 19)

CH (Dycal) (10) 1/8 MS and 2/8 LB samples remained positive; median difference MS:

1/0/2 (3)

2009

[17]

painless vital primary teeth with caries into inner half of dentin (30)

re-entry and collection after 3 mos

     

1949848445, range: 44668 120226443 ( n = 19); median difference LB: 4, range:

 

0 7079457 ( n = 8)

Fairbourn et al., 1980

yes

Patients with deep caries

half of each lesion

Anaerobically in blood, MSB and MRS medium

CH (Dycal) (20)

TBC: 11/20 samples remained positive; TBC: SMD 8397, 95% CI: 6557-10238 ( n = 20); LB: SMD 2090, 95% CI: 183 1632 2548 ( n = 20)

No liner, ZOE restoration

TBC: 15/20 samples remained positive; TBC: SMD 15361, 95% CI:

0/0 (5)

lesions in proximity to pulp in excavated, ZOE

[62]

vital asymptomatic permanent teeth (40)

restoration, re-entry and sampling after

5

mos

(20)

 

11995-18726 ( n = 20); LB: SMD 1626, 95% CI: 1270 1982 ( n = 20) TBC: SMD: 21, 95%CI: 27 15 (n=15); LB:SMD 60, 95%CI: 45 75 ( n = 15)

Foley et al.,

yes

45 patients (4 9) with more than one vital and asymptomatic primary molar tooth with cavitated dentinal lesions of different depths

(45)

SE, GIC restoration, extraction and sampling after 1 mo (30) or 6 mos (15)

Anaerobically on MSB, MRS, FA agar

BCC liner (15) TBC: SMD 114, 95%CI: 85 142( n = 15); LB:SMD 114, 95%CI: 45 142 ( n = 15)

No liner, GIC restoration

0/0/0 (1 or 6)

[63]

 

(15)

 

Hoshino et al., [64]

no

13 children (7 15) with vital and asymptomatic teeth and proximal or occlusal dentinal lesions (13)

SE, GIC or resin restoration AR (Photo Clear l A), re-entry and sampling after

Anaerobically on BHI, YE, blood agar

Antibiotic TCP

0/9 samples remained positive

TCP (4)

4/4 samples remained positive 0/0 (5)

(9)

 

1

12 mos (7), 24 mos (2) SE, amalgam restoration, re-entry and sampling after

d (8), 1 mo (9),

 

King et al.,

no

26 children (4 10) with deep caries lesions in vital and asymptomatic teeth (57)

Aerobically and anaerobically on TSB agar or in TG

No liner (8)

8/8 samples remained positive

CH (Pulpdent) (21), ZOE liner (USP) (22)

8/21 CH samples remained positive, 4/22 ZOE samples remained positive

Total 7 (5)

[14]

 
 

25

206 days

broth Anaerobically in blood, MSB and MRS medium

 

Leung et al.,

yes

Patients with deep caries lesions in proximity to pulp in vital asymptomatic permanent teeth (40)

half of each lesion

CH (Dycal) (20)

TBC: 8/20 samples remained positive; TBC: SMD 22278 95% CI: 17396 27160

No liner (wax placebo) (20)

TBC: 20/20 samples remained positive; TBC: SMD 75894, 95% CI2 92524 59263 (20); LB: SMD 1198, 95% CI: 1461 935 ( n = 20)

0/0 (1)

[15]

excavated, ZOE

restoration, re-entry and sampling after

(20);

 

LB: SMD 540, 95% CI: 422 658 ( n = 20)

 

1

mos

1302

F. Schwendicke et al. / Journal of Dentistry 43 (2015) 1298 1307

Table 1 ( Continued )

Study

RCT

Participants (age) (n teeth)

Excavation,

Evaluation

Treatment(s) A (product), samples ( n )

Results treatment(s) A, samples ( n )

Treatment(s) B (product), samples ( n )

Results treatment(s) B, samples ( n )

Drop-out in

yes/

restoration and

method

different groups

no

sampling

 

(mean follow-up)

Neelakantan et al., [16]

yes

50 patients(13 30) with lower permanent vital and asymptomatic molars with deep occlusal cavities (50)

CE, amalgam restoration, re-entry and sampling after

Anaerobically on BHI, tomato juice agar, FA

MTA (ProRoot) MTA: 0/7, AB: 5/8, HA: 0/9 samples

CH (Dycal) (10), no liner (10)

No liner: 4/8, CH: 4/8 samples remained positive

3/2/1/2/2 (3)

(10),

remained positive

 

polyantibiotic

3

mos

agar

paste (10), HA (LimeLite) (10) No liner, ZOE restoration

 

Nordstrom et al., [65] , NRCT

yes

Patients (4 22) with

SE, ZOE restoration,

Anaerobically in MRS and TG medium

14/32 (MRS) and 14/32 (TG) samples remained positive

CH (32)

13/32 (MRS) and 19/32 (TG) samples remained positive

3 in total (3)

25

deciduous and

re-entry and

39

permanent vital and

sampling after 3 mos

(32)

 

asymptomatic teeth with large carious lesions close to

pulp (67) Children (4 10) with vital asymptomatic primary molars with active dentin lesions (40) 20 children (4 7) with primary vital and asymptomatic molars with deep occlusal/occluso- proximal/occluso-lingual lesions with margins in enamel (42) 22 adults (18 67) with primary dentinal non- cavitated occlusal/occlusal- proximal lesions in permanent vital and asymptomatic molars or premolars (30)

 

Pinheiro et al., [66]

yes

SE, resin restoration (Single Bond, Z100), re-entry and sampling after 24 h SE, resin restoration (Z250), re-entry and sampling after 4

Anaerobically on blood agar

Antibiotic GIC

TBC: 16/20 samples remained positive, TBC: SMD 1.4, 95% CI: 0.7 2.1( n = 20)

GIC (Vidrion F)

TBC. 20/20 samples remained positive, TBC: SMD 1.2, 95%CI 0.6 1.8 ( n = 0) TBC: SMD 5223, 95% CI 3604

0/0 (0)

(20)

(20)

Pinto et al.,

no

Anaerobically on MSB, MRS BHI agar

CH (Hydro C)

TBC: SMD 2479, 95%CI: 1711 3247; LB: SMD 141 95% CI: 97 184 ( n = 19)

No liner

2/1 (4 7)

[67]

(21)

(gutta-percha

6841,

 

placebo) (21)

LB: SMD 2, 95% CI: 1 3 ( n = 20)

7

mos

Wicht et al., 2004, 2005 [68,69] a

yes

SE, resin restoration (Prime and Bond NT, Dyract AP), re-entry and sampling after

Anaerobically on blood, MSB, MRS agar

Disinfectant

TBC: 10/10 and 10/10 samples remained positive; LB: 5/10 and 3/10 samples remained positive; TBC: SMD 2204 95% CI 1521 2287 ( n = 10) and SMD 8152 95% CI: 5626 10679 ( n = 10); LB: SMD 134, 95% CI: 93 176 ( n = 10) and SMD 71, 95% CI: 49 93 ( n = 10)

No liner

TBC 10/10 remained positive; LB: 10/10 samples remained positive; TBC: SMD 17979 95% CI 12406 23551 ( n = 10); LB: SMD 17 96% CI: 12 22 ( n = 10)

0/0 (1.5)

liner (Cervitec)

(gutta-percha

(10), antibiotic

placebo) (10)

   

liner

6

weeks

(Ledermix)

 

(10)

Abbreviations: BHI brain heart infusion, CE complete excavation, CFU colony forming units, EDJ enamel-dentin junction, FA fastidious anaerobe, LB lactobacilli, MRS deMan Rogosa Sharpe, MS mutans streptococci, MSB mitis salivarius bacictracin, (N)RCT (non)-randomised controlled trials, RD rubber dam, SE selective excavation, SMD standardised mean difference, TB C total bacterial counts, TG thioglycollate, TSB trypticase soy broth, YE yeast extract. a These studies were published separately, but reported data on the same control group. Thus, only data from the rst study (Wicht et al., 2004) was used.

F. Schwendicke et al. / Journal of Dentistry 43 (2015) 1298 1307

1303

being included ( Fig. 1 ). Reasons for study exclusion at the full- text-stage can be found within the appendix (Table S1). Included studies represented minimum 270 patients (not all studies reported this) with 500 treated lesions. Most patients were children or adolescents; 6 and 4 studies reported solely on primary and permanent teeth, respectively. One study did not explicitly state the dentition. The age of the patients ranged from 4 to 67 years. Median follow-up (time between lining and re-entry) was 3 months (range 1 day to 24 months). Eleven studies were RCTs and three studies were non-RCTs. None of the studies reported to use a split-mouth design, whilst clustering of treated lesions in one mouth was common ( Table 1 ). Risk of bias was found high or unclear in all studies. This was mainly due to lack of blinding the operators, as commonly found in operative studies. Moreover, only two studies reported on proper allocation concealment, and only half of the studies reported how random treatment allocation was performed (Table S2) Seven studies had a placebo group with no liner application (total number of teeth was 104). CH was used as a liner in nine

studies (169 teeth), ZOE in three studies (74 teeth), ATB in four studies (59 teeth) and GIC in three studies (55 teeth). Both BCC and MTA each were used in one study (15 and 10 teeth, respectively). Various materials had been employed to temporarily seal the cavity between the two visits ( rst treatment and re-entry):

Composite and zinc oxide eugenol had been used in four studies, amalgam in two and GIC study. Two studies used various materials. Given the limited data and the great number of potential confounders, we did not statistically assess their impact on our

ndings.

3.2. Number of positive samples (sterility of the cavity oor)

The number of positive bacterial dentin samples after a certain lining or sealing period was reported by 11 studies, with 403 bacterial samples (cavities or cavity sites) having been sampled. Pairwise comparisons ( Fig. 2 a) found no signi cant difference between any of the groups in achieving sterility of the cavity oor, with only three comparisons including more than one study: Cavities without liners had 1.5-times the odds of yielding positive samples than cavities lined with calcium hydroxide (OR [95% CI]: 1.50 [0.90; 2.51]). This probability did not signi cantly differ in cavities lined with calcium phosphates (TCP/HA) compared with antibiotic or disinfecting liners (OR: 1.23 [0.01, 257]), cavities lined with antibiotic liners compared with no active liner (OR: 0.92 [0.34, 2.48]), or cavities lined with ZOE compared with calcium hydroxide (OR: 1.20 [0.61, 2.33]). Only for one comparison, heterogeneity was high. Funnel plots analysis ( Fig. 2 b) did not indicate risk of publication bias. Based on these data, a network of interventions was con- structed ( Fig. 2 c), connecting seven groups via 14 direct compar- isons. The network formed a nearly fully connected polygone, indicating that uncertainty exists as to which comparators to choose, i.e. a gold standard being absent. When submitted to network meta-analysis, the chances of having no positive bacterial samples was highest when using MTA liner (73%), followed by antibiotic/disinfectant liners (8%) and ZOE (7%); the last ranked strategy was lining using glass ionomer cement ( Fig. 2 d). This was con rmed calculating and plotting SUCRA lines ( Fig. 2 e), with SUCRA values being highest for MTA (89.4), followed by antibiotic/disinfectants (59.6) and CH (55.0), and lowest for GIC (25.0) and no liner/placebo (22.3). Concurrent with traditional meta-analysis, estimates for network meta-analysis did not yield signi cant differences (Fig. S1). From the 9 loops formed, one loop (involving treatments CH, TCP/HA, and ATB/disinfectant liners) showed evidence of statistical inconsistency (IF: 5.48, 95% CI:

0.80 10.2) (Fig. S2).

3.3. Bacterial reduction

The bacterial reduction after a certain lining or sealing period was reported by six studies with 209 cavities or cavity sites being analysed. The available data indicated that not using any lining (or only placebo lining) seems less likely to reduce the bacterial load in the cavity than lining with calcium hydroxide (SMD [95% CI]: 6.05 ( 28.89, 14.78)). The difference between GIC compared with BCC was statistically signi cant (SMD: 14.06 [ 17.81, 10.31]), as was the difference between antibiotic or disinfectant liners and GIC, or ZOE and calcium hydroxide ( Fig. 3 a). Only one pairwise comparison was based on more than one study; the heterogeneity for this comparison was high ( I 2 = 99%). Again, funnel plot analysis did not indicate publication bias ( Fig. 3 b). Based on these data, another network was constructed ( Fig. 3 c), connecting six treatments using a linear structure, the latter being the product of paucity of data rather than a potential evolution of tested treatments. Using network meta-analysis, ZOE was found to have the highest probability of achieving a bacterial reduction, whilst no lining was ranked lowest ( Fig. 3 d). This was con rmed calculating and plotting SUCRA lines ( Fig. 2 e), with values being highest for ZOE (100.0), followed by BCC (64.6) and antibiotic/ disinfectants (32.8) ( Fig. 3 e). Moreover, pairwise estimates for network meta-analysis con rmed the presence of signi cant differences between materials (Fig. S3). It should be highlighted that not all treatments compared for the odds of yielding positive bacterial samples (see above) were included into this analysis; comparisons between the two different outcomes should thus be interpreted with caution. Assessment of inconsistency was not feasible due to absence of closed loops.

4. Discussion

One main aim of using cavity liners is to reduce the load of remaining bacteria, as these or their metabolic by-products might harm the pulp and compromise pulpal health, especially when treating deep pulpo-proximal lesions. This review systematically appraised the available clinical data supporting such antibacterial effects of different liners and comparatively synthesized these data using network meta-analysis. Our primary analysis focused on the number of cavities which remained bacterially infected after a lining/sealing period or, vice versa, the numberofcavitieswhichwere renderedbacteria-free after certain time periods. MTA was found most suitable for achieving this aim, with a 73% probability of being the best material for this purpose, followed by antibiotic or disinfectant and zinc oxide eugenol. The commonly used calcium hydroxide was found less potent, and not performing any or only placebo lining had only a small chance of achieving this goal. However, the underlying data for these ndings are sparse, the ranking should thus be interpreted with caution, as indicated by the absence of statistically signicant differences in both pairwise and network meta-analyses estimates. Our secondary analysis, which was based on a smaller number of studies and did not comprise all lining groups (e.g. MTA was not included), assessed the magnitude of bacterial reduction during the lining/sealing period via comparing the number of colony-forming units before and after this time. In this analysis, calcium hydroxide was found most probable of achieving a bacterial reduction, as were antibiotic liners. Again,not performing any oronly placebolining had only limited antibacterial effects. Mineral trioxide is not commonly used as cavity liner yet, but has been found to induce an alkaline pH shift whilst leaching calcium [45] . The results regarding its antibacterial properties, however, are ambiguous so far [45] , which is why the ndings of this review are of relevance. If assessed for further properties, MTA also enhances the viability of pulpal cells, induces angiogenesis [46] , and stimulates

1304

F. Schwendicke et al. / Journal of Dentistry 43 (2015) 1298 1307

et al. / Journal of Dentistry 43 (2015) 1298 – 1307 Fig. 2. The number of

Fig. 2. The number of positive bacterial samples after different lining treatments. Overall, seven treatments were compared: no or placebo lining, calcium hydroxide (CH), mineral trioxide aggregate (MTA), glass ionomer cement (GIC), mineral liners (tri-calcium phosphate [TCP] or hydroxyl apatite [HA]), antibiotic (ATB) or disinfectant liners, or zinc oxide eugenol (ZOE) liner. (a) Pairwise meta-analysis was used to compare the probability of positive dentin samples after different liner treatments. Odds Ratios and 95% Con dence Intervals were calculated; heterogeneity was assessed using I 2 -statistics. (b) Funnel plot analysis to assess potential publication bias. (c) Network of the comparisons for the Bayesian network meta-analysis. For each node (lining treatment), the number of treated cavities or cavity sites is indicated by the width of each circle. Direct comparisons are indicated by lines between nodes, with the width of lines being proportional to the number of trials comparing connected treatments. Indirect comparisons are not shown. (d) Strategy ranking according to network meta-analyses. Strategies were ranked for having the highest chance of no positive bacterial samples, i.e. sterility of the cavity. The probability of being ranked 1st, 2nd etc. of each strategy (different colours) is given. (e) Cumulative ranking probability plots. On the horizontal axis is the possible rank of each treatment (from rst best rank to worse). On the vertical axis is the cumulative probability for each treatment to be the best, second best etc. option (11). (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

odontoblastic differentiation and mineralization [47] . MTA has been introduced to resin composite materials, possibly allowing to omit its separate use as a liner, which could improve clinical results, since the material itself has only low compressive strength and could compromise the restoration placed above [45] . The inferior effects of calcium hydroxide compared with MTA have been found in other regards as well, with lower success chances after direct capping using calcium hydroxide compared with MTA [4850]. A direct comparison

of the clinical effects of both materials used as liners has been performed in a recent randomized trial and conrmed the benets of MTA [51] . In general, the suitability of calcium hydroxide for cavity lining is increasingly questioned, and our results might partially support these growing doubts. However, we found calcium hydroxide to exert stronger antibacterial effects than other materials also commonly used for lining (especially GIC), and to reduce bacterial numbers much more effectively than only sealing the cavity.

F. Schwendicke et al. / Journal of Dentistry 43 (2015) 1298 1307

1305

et al. / Journal of Dentistry 43 (2015) 1298 – 1307 1305 Figure 3. Bacterial reduction

Figure 3. Bacterial reduction by lining and/or sealing. Six treatments were compared: no or placebo lining, calcium hydroxide (CH), glass ionomer cement (GIC), antibiotic (ATB) or disinfectant liners, zinc oxide eugenol (ZOE), or black copper cement (BCC) liner. (a) Pairwise meta-analysis compared the standardized mean difference (SMD) reduction of bacterial numbers at the cavity oor in different treatment groups. A negative SMD indicates a lower reduction in the test compared with the control group, whilst positive SMDs indicate a higher reduction. SMD, 95% con dence intervals and I 2 -values are shown. (b) Funnel plot analysis to assess potential publication bias. (c) Network of the comparisons for the Bayesian network meta-analysis. (d) Strategy ranking according to network meta-analyses. Strategies (different colours) were ranked for having the highest chance of bacterial reduction, i.e. those with the lowest reduction were ranked lowest. (e) Cumulative ranking probability plots. On the horizontal axis is the possible rank of each treatment (from rst best rank to worse). On the vertical axis is the cumulative probability for each treatment to be the best, second best etc. option (11). (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

In this review, zinc oxide eugenol was highly ranked for achieving sterility of the cavity, which con rms the high antibacterial potential of ZOE compared with CH, GIC or MTA [52,53] . Other studies found ZOE to reduce bio lm adhesion [54] , which supports the notion of it being suitable for exerting antibacterial effects within the oral environment. Similarly, our secondary analysis found ZOE to have only limitedly antibacterial effects if measured via reduction in bacterial units remaining in the cavity. This nding was based on a single study, and caution is required when interpreting it. GIC was ranked low with regards to its probability of achieving total bacteria elimination and for reducing bacterial numbers. However, GIC has been shown to have antibacterial properties [17,55,56] and to be suitable for suppressing the cariogenic activity of remaining bacteria by releasing uoride ions [57] . It was further reported to allow remineralisation of the remaining carious dentin and to actively bond to dental hard tissues, which could have mechanical advantages [10,58] . Moreover, clinical studies found teeth lined with GIC to have high chances of success compared with other lining materials [11,51] . In consequence, there seems to be a con ict between clinical studies, which found GIC or providing no lining highly successful, and the studies included in this review, which found MTA or antibacterial liners

more suited for exerting antibacterial effects. It can be deducted that reducing the number of remaining bacteria might not be as relevant as commonly thought, especially when considering the demonstrated antibacterial effects of a tight seal [18] . Here, again, our results are in obvious con ict with existing studies, which found sealing itself to drastically reduce the number of bacteria within the cavity [18] , which was not con rmed by our review. There are several reasons for these discrepancies. First, we did not mainly assess the bacterial reduction itself, but the relative differences of this reduction between different groups: It might be that sealing is ef cacious for reducing bacterial numbers, but the use of liners might further increase this effect. Second, all but one included study did not only seal the cavity without any liner (direct adhesion), but used placebo liners (wax, gutta-percha) instead. It could be hypothesized that the remaining space beneath the restoration is detrimental, since it might allow bio lm re- formation. One effect of sealing, which has not been extensively assessed so far, might thus not only be the starvation of bacteria by carbohydrate deprivation, but also the inhibition of bio lm formation due to lack of space. The remaining bacterial units within dentin tubules or the hybrid layer might not achieve a critical mass to survive or to be harmful. However, it is also possible

1306

F. Schwendicke et al. / Journal of Dentistry 43 (2015) 1298 1307

that studies focussing on clinical outcomes underestimate the harm from remaining bacteria, as pulp histology is unknown and pulpal health most likely over-estimated [59] . This review has several limitations, which impact on our conclusions. First, our search had only low sensitivity, with a relatively large number of studies being identi ed by hand searches or cross-referencing. However, only few of these studies were eventually included. Second, some included studies were non-randomized, which introduces bias by selection. However, we assessed studies for the comparability of lesions and conditions across treatment groups, and only included trials with similar baseline conditions in both groups. Third, the performed categorisation of treatments might distort the results of the network meta-analysis. This categorization, however, was re- quired, and was justi ed by the similarity of material classes (e.g. TCP and HA). Fourth, this review did not assess the comparative antibacterial effects of restorations, but compared the effects of different liners or liners versus restorations. Antibacterial compo- sites, for example, might well be more effective for reducing bacterial numbers than many liners [60] , as might antibacterial adhesives [61] . Last, the analysis was also affected by the variety in outcome measurement methods: oftentimes, culture conditions (agar/broth, milieu) differed between studies, as did baseline bacterial loads and follow-up (sealing) periods. Whilst within each study, this was not an issue as most studies were RCTs, such difference across studies impacts on the transitivity of the network analysis [42] . Transitivity could have been further affected by the variety of effect modi ers with potential impact on our results. For example, the dentition, the lesion type (number of surfaces, lesion extension and depth), the performed excavation technique ( complete or selective/incomplete excavation), or the used restoration material might have affected our ndings. We did not assess their impact due to the relatively small number of available studies, but no pattern of potential confounding occurred when descriptively evaluating studies. Similarly, the small number of studies and the resulting network did allow only limited assessment of consistency, especially for our secondary analysis. In general, the number of studies and their low quality should be born in mind when interpreting the results of this review. Given these limitations and the potentially limited correlation between the remaining number of bacteria in a cavity and clinical outcomes, this review cannot recommend any speci c approach toward cavity lining. Based on our data, certain liners might be more suitable to achieve sterile cavities or reduce bacterial numbers within these cavities. In clinical practice, the performed excavation of carious dentin and the quality of the subsequently placed restoration might be more decisive than the decision for or against a speci c liner. Clinical trials should record both clinical and microbiologic outcomes to explore their potential association. The effects of remaining bacteria on pulp histology and pulpal health should be evaluated, as this might be more relevant than antibacterial effects, especially when considering the effects of sealing on remaining bacteria. When assessing antibacterial effects of liners, standardized approaches for microbiologic analysis should be sought. Current evidence is insuf cient for supporting speci c cavity treatments prior restoration.

Acknowledgments

The authors have no any con icts of interest.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jdent.2015.07.001.

References

[12]

[9]

[18]

[21]

F. Schwendicke et al. / Journal of Dentistry 43 (2015) 1298 1307

1307

97.

[50]

[55]

Centres d'intérêt liés