Effects of immediate and delayed loading protocols on marginal bone loss around implants in unsplinted mandibular implant-retained overdentures: a systematic review and meta-analysis

Background Immediate loading has recently been introduced into unsplinted mandibular implant-retained overdentures for the management of edentulous patients due to their increasing demand on immediate aesthetics and function. However, there is still a scarcity of meta-analytical evidence on the efficacy of immediate loading compared to delayed loading in unsplinted mandibular implant-retained overdentures. The purpose of this study was to compare the marginal bone loss (MBL) around implants between immediate and delayed loading of unsplinted mandibular implant-retained overdentures. Methods Randomized controlled trials (RCTs), controlled clinical trials (CCTs), and cohort studies quantitatively comparing the MBL around implants between immediate loading protocol (ILP) and delayed loading protocol (DLP) of unsplinted mandibular overdentures were included. A systematic search was carried out in PubMed, EMBASE, and CENTRAL databases on December 02, 2020. “Grey” literature was also searched. A meta-analysis was conducted to compare the pooled MBL of two different loading protocols of unsplinted mandibular overdentures through weighted mean differences (WMDs) with 95% confidence intervals (95% CIs). The subgroup analysis was performed between different attachment types (i.e. Locator attachment vs. ball anchor). The risk of bias within and across studies were assessed using the Cochrane Collaboration’s tool, the Newcastle–Ottawa scale, and Egger’s test. Results Of 328 records, five RCTs and two cohort studies were included and evaluated, which totally contained 191 participants with 400 implants. The MBL of ILP group showed no significant difference with that of DLP group (WMD 0.04, CI − 0.13 to 0.21, P > .05). The subgroup analysis revealed similar results with Locator attachments or ball anchors (P > .05). Apart from one RCT (20%) with a high risk of bias, four RCTs (80%) showed a moderate risk of bias. Two prospective cohort studies were proved with acceptable quality. Seven included studies have reported 5.03% implant failure rate (10 of 199 implants) in ILP group and 1.00% failure rate (2 of 201 implants) in DLP group in total. Conclusions For unsplinted mandibular implant-retained overdentures, the MBL around implants after ILP seems comparable to that of implants after DLP. Immediate loading may be a promising alternative to delayed loading for the management of unsplinted mandibular implant-retained overdentures. PROSPERO registration number: CRD42020159124. Supplementary Information The online version contains supplementary material available at 10.1186/s12903-021-01486-3.

Consequently, we quantitatively compared the effects of immediate and delayed loading protocols on MBL around implants for unsplinted mandibular implant-retained overdentures in this systematic review and meta-analysis, to inform dental practitioners about the selection of appropriate loading protocol for the long-term clinical success of restorations.

Methods
The present meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.
The protocol was registered priori in the International Prospective Register of Systematic Reviews (PROSPERO) (www.crd.york.ac.uk/PROSPERO/) (registration number: CRD42020159124)

PICOS question
According to the recommendations of the Centre for Evidence-Based Medicine (University of Oxford, Oxford, UK), the PICOS (participants, interventions, comparisons, outcomes, and study designs) question was as follows: How is the effect on MBL around implants after ILP in the unsplinted mandibular implant-retained overdentures when compared to DLP?
Intervention: ILP for unsplinted mandibular implant-retained overdentures. In addition, if a study met any of the following exclusion criteria, it was excluded from the study: Case reports, review papers.

Overdentures retained by a single implant only
Diameter of implants narrower than 3 mm (mini-implant) Duplicate studies based on the same patient cohorts Studies with sample size less than ten Information sources and literature search The literature search was conducted independently by two independent assessors (W.L. and H.C.). Any disagreement was be resolved by discussion of the two assessors. Three online electronic databases, including PubMed, EMBASE, and CENTRAL (Cochrane Library), were searched for relevant scienti c reports published in the English language on April 28, 2020. No time lter was applied. The online search was conducted with the search strategy combining both the MeSH and free text words with high sensitivity and adaptation for the databases (Table 1). For the "grey" literature (e.g. unpublished and ongoing studies, conference abstracts, dissertation and thesis), the ClinicalTrials.gov, System for Information on Grey Literature in Europe (OpenGrey), National Technical Information Service (NTIS), and ProQuest Dissertation Abstracts, and Thesis databases were also searched. Furthermore, hand search was performed to identify the eligible reports based on the reference lists of related trials and reviews as a complement.

Study selection
After pooling the full search results from all database, literatures with repetitive contents were excluded. Two independent assessors (W.L. and H.C.) independently screened the titles and abstracts of studies, and irrelevant reports were discarded. Then the full-text evaluation of articles was carried out by two independent assessors (W.L. and H.C.) to select reports that met all inclusion criteria as well as to exclude reports according to any of the exclusion criteria.
Any disagreement about whether a study should be included was resolved by discussion or arbitrated by a third assessor (L.S.). In addition, the kappa statistic was used to measure agreement between the independent assessors.

Data collection and data items
The data were extracted from included reports and cross-checked by two independent assessors (W.L. and H.C.) independently. If there was a discrepancy on the data extraction during this process, the third assessor (L.S.) was consulted and an agreement was nally reached through a consensus discussion. A data collection form was developed a priori to record the extracted information. The following data were included: study, study design, total number of patients, age, edentulous region, number of implants (per patient), implant system, implant diameter, implant length, torque of implants, attachment type, comparison, number of patients in ILP/DLP, loading time of ILP/DLP, radiographic method, marginal bone loss, dropout (patient).
During the data extraction, it was found that one of the studies reported the mean and the standard deviation of vertical bone loss at four different sites (i.e. distal, labial, mesial, and lingual), while the other studies measured MBL at distal and mesial sites around the inserted implants. Thus, only the average values of MBL at distal and mesial sites in these studies were included in the following meta-analysis.

Risk of bias in individual trials
The risk of bias in the included RCTs were evaluated by the Cochrane Collaboration's tool (RevMan version 5.3) [38]. According to the bias indices, three different levels including low, moderate, and high were used to classify the risk of bias within RCTs [39]. The Newcastle-Ottawa scale (NOS), as an ordinal starrating scale, was employed for the assessment of methodological quality of non-RCTs. In NOS, a higher score represented a higher report quality of cohort study [40]. The assessments were independently carried out by two independent assessors (W.L. and J.Z.). Any disagreements were discussed and resolved until consensus was reached.

Summary measures and synthesis of results
Statistical analyses were performed via the RevMan (RevMan v5.3, Cochrane Collaboration) and Stata (Stata MP v14, StataCorp LP) softwares. To compare MBL of ILP group with DLP group, the weighted mean differences (WMDs) with 95% con dence intervals (95% CIs) for these continuous outcomes was calculated. The results were provided with a xed-effect or random-effects model [41]. Statistical heterogeneity was measured by the Chi 2 statistic and I 2 statistic [39].

Risk of bias across studies
When there are at least 10 studies included in the present meta-analysis, tests for funnel plot asymmetry was drawn. Otherwise, Egger's test was employed to assess the publication bias [39,42].

Additional analyses
The subgroup analysis was carried out among different attachment types of unsplinted mandibular implant-retained overdentures including Locator attachments, ball anchors and magnetic attachments, in order to reveal whether the overall estimate effect would be in uenced by different attachment types. Sensitivity analyses were performed by removing individual trials from the meta-analysis, to see whether the overall effect would be affected and thus to reveal the robustness of the results.

Study selection
Five hundred and eighty-one and six records were identi ed through database search and hand search respectively. Through removal of repetitive records and initial screening by titles and abstracts, 34 records remained. After full-text review, 27 studies were excluded as they did not meet the eligibility criteria and a total of seven articles were included for this review and meta-analysis (Fig. 1).
The kappa value for the inter-investigator agreement of initial screening by titles and abstracts was 0.96 and that of full-text evaluation was 0.91, both presented an "almost perfect" inter-agreement [43]. Table 2 showed the characteristics of the included seven studies. All studies were published from 2007 to 2020, and 191 recruited participants were included all together. Among these seven studies, ve (71%) were RCTs and two (29%) were prospective cohort studies. In six studies [44][45][46][47][48][49], overdentures in ILP groups were attached to the implants immediately after implantation, while in the other one study, the restorations in the ILP group were performed seven days after the surgery [50]. In DLP groups of all included studies [44][45][46][47][48][49][50], the healing time after implant surgery was no less than three months. Four studies [44][45][46][47] disclosed the speci c ranges of inserting torque, and the other three studies [48][49][50] did not report the insertion torque. Moreover, Locator attachments were employed in three studies [44,46,47] and ball anchors were utilized in the other four literatures [45,[48][49][50]. In regard to dropout of implants, four studies [46][47][48][49] reported that only one or two participants happened implant failure in ILP group, while one study [49] revealed implant failure of one participant in DLP group.

Study characteristics
Risk of bias within studies Figure 2 illustrated the quality assessments and risk of bias of the included RCTs. Four RCTs [44,45,47,48] were of unclear risk of bias while one study [46] included a high risk of bias. Two RCTs had a low risk of selection bias and three RCTs did not provide detailed information in terms of random sequence generation or allocation concealment. Only one RCT [46] (20%) reported a high risk of performance bias as only one experienced operator performed all the surgeries in ILP group and DLP group. Three studies [45][46][47], which did not provide enough information about the binding of outcome assessment, had an unclear risk of detection bias. All studies reported low risk of attrition bias with no missing data [44,45,48] or reasonable explanations for only one missing data [46,47]. The reporting bias was classi ed at an unclear level in 80% of the trials [44,45,47,48], for insu cient information was available to judge the risk level. These studies appeared to be free of other sources of bias. Table 3 presented the quality assessment results of two prospective cohort studies [49,50].
The overall risk of bias was deemed to be low (seven stars) in the two cohort studies, and these studies were proved to be of an acceptable quality [40,51,52]. As the two studies did not report the derivation of the cohorts, it was unclear to judge the representativeness of their exposed cohorts. Though two subjects of one study [49] lost to follow up, the description of those lost was provided in detail and then the adequacy of follow up was considered with low risk of bias.

Results of individual studies and synthesis of data
In this review, 191 patients with a follow-up of no less than 12 months were pooled for the synthesis of data and the meta-analysis result of seven studies was illustrated with the forest plot (Fig. 3). As a result of the comparison of MBL between ILP and DLP groups, a substantial heterogeneity [38] (P = .04, I 2 = 55.06% > 50%) was found. Therefore, instead of the xed-effect model, DerSimonian-Laird model as a random-effect model was applied according to the STATA technical bulletin and Cochrane handbook [39,53]. No statistically signi cant difference of MBL was detected (WMD 0.04, CI -0.13 to 0.21, P = .68) between ILP and DLP group.

Risk of bias across studies
Because the limited number of included studies (n < 10), the publication bias assessment was conducted with Egger's test in the current meta-analysis. The result of Egger's test indicated that no signi cant bias could be found among seven included studies (P > .05). Nevertheless, this evaluation of risk of publication bias should be considered as a reference only owing to the limited quantity of articles.

Additional analysis
According to the different attachments applied in these studies, two subgroups (i.e. Locator attachment vs. ball anchor) were set. Figure 4 showed the comparison of average MBL in the subgroups respectively. In either the Locator attachments (three trials, WMD -0.08, CI -0.50 to 0.34, P > .05) or the ball anchors (four trials, WMD 0.05, CI -0.15 to 0.25, P > .05) subgroup, no signi cant difference was detected between ILP and DLP. In the sensitivity analyses, one single study was deleted from the overall pooled analysis each time and it was found that the exclusion of individual studies did not show any in uence on the overall MBL (Fig.5). Moreover, with the omission of two prospective studies, there was still no signi cant difference between ILP and DLP groups (WMD -0.02, CI -0.23 to 0.19, P > .05).

Discussion
The present systematic review was based on the results from ve RCTs and two non-RCTs with a total of 191 patients. The meta-analysis was conducted to compare the MBL around inserted implants of unsplinted mandibular overdentures between ILP and DLP groups, and the result did not reveal any signi cant difference (P > .05). Further, the subgroup analysis showed similar results about the MBL between the two different loading protocols in either the Locator attachments or the ball anchors subgroup (P > .05), suggesting that different attachment types employed in the included studies did not result in any difference on the result of meta-analysis, and might not be the source of cross-study heterogeneity. Additionally, in both ILP and DLP groups of these included studies, the average MBL surrounding implants of unsplinted implant-retained overdentures were lower than 2 mm, indicating that the results were all clinically acceptable as far as MBL was concerned [31].
Several studies reported that the initial healing of implants with immediately loaded mandibular overdentures might be impaired by the resultant restoration movement, immediate abutment connection, and early contact with oral microbial plaque [54][55][56]. However, the present meta-analysis reported no detectable difference in the MBL around immediately loaded implants compared to the DLP group, which is also in line with the evidence from some previous systematic reviews [33,57]. This may be attributed to the early mechanical strain and a lack of the second-stage surgery in ILP [58][59][60]. Mechanical loading from the overdentures in ILP might act as a stimulator to alveolar bone formation and then lead to high bone fractions [60]. The early mechanical strain in the bone-toimplant contact surface was found to have a positive effect on the initial phase of bone healing [58,59]. Moreover, second-stage surgical operations could bring additional trauma and damage tissues, leading to marginal bone loss [61,62]. DLP with a second-stage surgery might be accompanied by the loss of underlying bone around implants compared with ILP [63]. Noticeably, Sanz-Sanchez et al. [24] and Kern et al. [64] synthesised both the xed implant-supported and removal implant-retained restorations and reported that the ILP group presented an even less MBL compared to DLP group. This result may due to the combined discussion of xed implant-supported and removal implant-retained restorations. the early mechanical strain from xed dentures is usually much higher than that from rmovable restorations, thus the positive effect of immediate loading on MBL [60] is supposed to be more obvious.
Besides different loading protocols, marginal bone level could also be in uenced by some factors such as implant primary stability and alveolar bone condition. In seven included studies, the effect of these factors on the MBL has been carefully controlled. The insertion torque value and implant stability quotient (ISQ) in these studies revealed clinically acceptable micromotion of implants [17,65], and all implants sites of included cases were the alveolar bone of anterior mandible which was considered with the highest bone density compared to the alveolar bone of other oral regions [66]. Therefore, the primary stability, as an essential prerequisite for implant success, met the implant loading requirement for both ILP and DLP.
In this review, the early loading protocol, in which implants were put in function between one week and two months after placement [22], was not discussed, for this loading protocol could not reduce healing time signi cantly compared with ILP, whereas would increase the period with low masticate e ciency, and is seldom applied on the implant-retained overdentures in the clinic.
MBL was employed as the only criterion to evaluate the therapeutic effects of immediate and delayed loading protocols in this study. However, this parameter indeed has its drawback. The raw MBL data at 1-year post-loading is a static index and it could only present the status of peri-implant tissue at one single time point, thus the reliability of MBL is controversial. Accordingly, the rate of MBL as a new index was proposed by Galindo-Moreno et al. recently [67]. As remodelling of marginal bone is a dynamic process, the rate of MBL is calculated in millimetre/month (mm/m) and could change over time [67,68]. Galindo-Moreno et al. [67] found that the progression of MBL tends to be higher and the risk of implant failure could be signi cantly increased when the rates of MBL was higher than 0.44 mm at six months post-loading. The new index may better help dentists predict future bone changes in the early stage and then establish a strict maintenance recall for patients. Meanwhile, a more de nite evaluation of clinical outcomes between these two different loading protocols could be carried out in a short observation time rather than a minimum follow-up of one year. Thus, the rate of MBL might be a more suitable criterion for implant success in the clinic.
This was the rst study to systemically evaluate the MBL of immediate loading compared to delayed loading in unsplinted mandibular implant-retained overdentures. The protocol of this study was registered in PROSPERO in advance and performed strictly in accordance with PRISMA guidelines. Apart from online search, manual search was performed based on the references of selected studies and related reviews, in order to discover quali ed trials which might not be included in the databases. Therefore, compared with individual studies, it is a more convincing clinical suggestion of loading protocol selection of unspinted mandibular implant-retained overdentures to practitioners. However, there are some limitations of this study. The implant number and implant system in the included studies had not achieved complete consistency, which might cause the considerable heterogeneity in the meta-analysis. Additionally, only seven clinical trials were included and two of them were prospective cohort studies. Due to the limited number of the trials, the results of this study might lack su cient evidence. Moreover, although the result of meta-analysis after removing two included prospective studies [49,50] remained the same, the validity of analysis might also be slightly compromised since no randomization of participant allocation could be employed in cohort studies. Thus, further high-quality, well-designed RCTs with large sample size are required to appraise the e cacy of different loading protocols on MBL around implants in unsplinted mandibular implant-retained overdentures.

Conclusions
Based on the results of this systematic review and meta-analysis, the MBL around implants restored with ILP showed no signi cant difference with that of implants restored with DLP for unsplinted mandibular implant-retained overdentures. The subgroup analysis suggested that either the Locator attachments or the ball anchors employed in the included studies would not result in any difference on the result of meta-analysis. However, considerable heterogeneity was observed across the included studies. Also, the limited number of trails and no randomization of participant allocation in the included cohort studies might compromise the validity of analysis. Further high-quality RCTs with robust study design and large sample size are needed to strengthen the evidence base and identify the effect of immediate and delayed loading protocols on MBL around implants in unsplinted mandibular implant-retained overdentures. We declared that materials described in the manuscript, including all relevant raw data, will be freely available to any scientist wishing to use them for noncommercial purposes, without breaching participant con dentiality.

Competing interests
The authors declare that they have no competing interests. Funding Nr, number; ILP, immediate loading protocol; DLP, delayed loading protocol; RCT, randomized controlled trial; NR, not reported; CBCT, cone beam computed tomography Table 3. Quality assessment and risk of bias of the included non-randomized studies.

Study
Coding Manual for Cohort Studies Newcastle-Ottawa Scale Turkyilmaz et al., 2012 Selection 1) Representativeness of the exposed cohort d 2) Selection of the non-exposed cohort a 3) Ascertainment of exposure a 4) Demonstration that outcome of interest was not present at start of study a Selection: 1) d: no description of the derivation of the cohort; 2) a: drawn from the same community as the exposed cohort ; 3) a: secure record (e.g., surgical records) ; 4) a: yes . Comparability: 1) a: study controls for ____ (select the most important factor) . Outcome: 1) a: independent blind assessment ; 2) a: yes (select an adequate follow up period for outcome of interest) ; 3) a: complete follow up -all subjects accounted for ; b) subjects lost to follow up unlikely to introduce bias -small number lost -> ____ % (select an adequate %) follow up, or description provided of those lost) .