Before moving on to the commentary for this systematic review I think it would be beneficial to reboot our scientific thinking systems. To do this I recommend reading or listening to Richard Feynman's 1974 Caltech Commencement Address.1

As the number of dental implants in the population increases, so has the prevalence of inflammatory lesions around these implants. The definition of peri-implant disease in its most basic sense is ‘a collective term for inflammatory reactions in the tissues surrounding an implant. Peri-implant mucositis is used to describe the presence of inflammation in the mucosa at an implant with no signs of loss of supporting bone. Peri-implantitis in addition to inflammation in the mucosa is characterised by loss of supporting bone’.2 In a recent systematic review the prevalence of peri-implantitis mucositis and peri-implantitis were 43% and 22% respectively,3 but as the authors point out, due to high heterogeneity this figure needs to be interpreted with caution. The authors here wished to investigate the impact of peri-implant maintenance therapy (PIMT) on biologic complications and implant survival, and a comprehensive literature search was undertaken which generated 13 studies which fulfilled their inclusion criteria.

The first point to note is that the authors state that the study followed the PRISMA (Preferred Reporting for Systematic Reviews and Meta-analysis) protocol,4 there was also a mention of following AMSTAR (A Measurement Tool to Assess Systematic Reviews).5 As part of both these tools the authors undertook a ‘risk of bias assessment’ of the included studies using the Newcastle-Ottawa Scale for cohort/case control studies.6The idea behind this is that only the better primary research available goes into the final analysis, so out of the 13 papers the authors whittled it down to ten. One paper scored eight out of a possible nine points (low risk of bias) and the mode being four/five (unclear to high risk of bias).

Now if you have done your homework and read Feynman you would know that it is important in science to be able to reproduce the original experiment before starting a new investigation. The best paper in this review is by Cho-Yan7 which scored eight/nine. It appears that the data set used was not independently validated and came from another study8 which was a retrospective paper where there was also no mention of independent validation either. The sample data are therefore two degrees removed from the original patient population and susceptible to selection bias, and should be downgraded from eight/nine to seven/nine. Similarly, in the two papers by Aguirre-Zorzano,9,10 the 2015 paper may possibly include the patients from the 2013 paper, and this could lead to the problem of ‘double-counting’. In the paper from Costa11 it was unclear how patients were allocated to maintenance/no maintenance groups, creating the risk of further selection bias, as the less motivated patients may be more likely to end up in the no maintenance group.

Unfortunately, it is not easy for most clinicians reading these reviews to access the full text of the primary literature or contact the authors for clarification and we have to appraise the information as presented. This is quite a common problem in systematic review12 and needs more attention when creating the research protocols. The process of conducting long-term studies is time consuming and expensive and almost impossible to reproduce in the dental environment so the ‘devil remains in the detail’. In this case, independent validation of the original data set and randomisation of the study groups. Further problems occur due to varying durations of study not mentioned in Table 4, with one paper running a study for one year post loading and others for five to ten years. Since peri-implantitis takes time to develop, one study was too short in real terms to have any external validity.

If the authors were in fact following the PRISMA protocol it would have been interesting to have followed Section 16 (additional analysis and how to make use of the risk of bias assessments) of the protocol and redo the meta-analysis without the unclear to high risk-of-bias papers. The problem with excluding two thirds of the implants is that the study becomes dramatically underpowered to detect such a small effect size. Keeping the numbers of participants high is known as ‘vote counting’ as described in a paper by Ioannidis.13

The second major problem with this review focuses on the statistical analysis and presentation of the results. Results were provided for the effects of peri-implant maintenance therapy in relation to peri-implant mucositis, peri-implantitis, periodontitis and the secondary outcome of implant survival, however they are difficult to interpret or place into context, and a table may have helped in comparing effect sizes. Charts were provided which did help in interpreting the results but they did not appear to link with the results provided in the main text. Beyond showing that the effect size between the groups was quite small, this may have been one of the reasons to include all the available data in the analysis regardless of quality. In relation to the results of the secondary outcome of implant survival rate the authors state that ‘implants under PIMT have 0.958 the incident event than those with no PIMT’ which might mean twice the survival rate or a 4% increased or decreased risk of failure. Clearer presentation of the results may have helped answer this question.

The authors did comment on the problems of bias and how previous studies could not draw a conclusion of treatment effect14,15,16 regarding maintaining health around dental implants. Unfortunately, the authors then sited two studies that are at high risk of bias to defend their argument regarding the impact of maintenance therapy.

Possibly the most interesting information that could be extracted from this study appears to have been missed by the authors. This is the difference between peri-implantitis at the implant level and patient level as shown in Figure 2 of the original review. Extrapolating from the chart itself the absolute risk reduction at the implant level of PIMT is (-)5% but at the patient level it is (-)25%. For this difference to occur there are several questions that come to mind. When the groups are allocated, do the higher risk patients end up in the no PIMT groups since they are less concerned about rigorous maintenance (selection bias) and do these higher risk patients have more implants due to increased tooth loss as a function of previous periodontal disease or recurrent periodontal (RP) disease. This factor was mentioned in Cho-Yan's paper7 where they concluded ‘that it is the presence of recurrent or RP, rather than a simple history or periodontitis, which is the risk factor for peri-implantitis’.

At the implant level it is interesting to note that even with a (-)33% relative risk reduction in infection, roughly 10% of the total implants have peri-implantitis despite the maintenance therapy which concurs with the authors' conclusion that other patient-, clinical- and implant-related factors need to be taken into consideration.

In conclusion, this review shows there is a small improvement in peri-implant tissue health with regular periodontal/PIMT but patients can still develop soft and hard tissue complications despite good maintenance. Though not mentioned directly, patients with a history of recurrent periodontitis prior to implant placement may be at the highest risk of peri-implantitis despite good oral hygiene, clinical technique or implant type. The progress with solving the problem of peri-implantitis, especially at the patient level lies with well-crafted prospective studies. These need to follow recognised reporting guidelines such as CONSORT (Consolidated Standards of Reporting Trials)17 for randomised trials and STROBE (Strengthening The Reporting of OBservational studies in Epidemiology)18 for observational studies. Dental research is always going to suffer from limitations such as study size, duration, patient retention etc, so we owe it to our patients to get the best out of every study undertaken.