Systematic Review and Meta-Analysis on Incidence of Altered Sensation of Mandibular Implant Surgery

Altered sensation (including paresthesia, dysesthesia and hypoesthesia) after mandibular implant surgery may indicate transient or permanent injury of the inferior alveolar nerve and the mental branch, and considerably lower patients’ satisfaction about the therapy. Previous studies have shown a great degree of variability on the incidence of altered sensation. We here reported the incidence of altered sensation after mandibular implant surgery based on a meta-analysis of 26 articles published between 1990.1.1 and 2016.1.1. Study quality and risk of bias was assessed and the studies with a lower score were excluded in the meta-analysis. Data synthesis was performed using the logistic-normal random-effect model. The meta-analyses revealed that the short-term (10 days after implant placement) and long-term (1 year after implant placement) incidence was 13% (95% CI, 6%-25%) and 3% (95% CI, 1%-7%), respectively. (2) For the patients who initially reported altered sensation, 80% (95% CI, 52%-94%) of them would return to normal sensation within 6 months after surgery, and 91% (95% CI, 78%-96%) of them would return to normal sensation one year after surgery. We concluded that dentist-patient communication about the risk of altered sensation is critical to treatment planning, since the short-term incidence of altered sensation is substantial (13%). When a patient reports altered sensation, regular assessment for 6 months would help tracing the changes of symptoms. In terms of long-term follow-up (1 year after surgery), the incidence is much lower (3%) and most patients (91%) would return to normal sensation.


Introduction
One of the most disturbing complications associated with implant surgery is post-operative altered sensation, including paresthesia, dysesthesia and hypoesthesia [1,2]. For mandibular implant surgery, altered sensation may be associated with an injury of the inferior alveolar nerve (IAN) and the mental branch. Persistent altered sensation may indicate neurosensory dysfunction or IAN neuropathy [3,4]. Even transient altered sensation, such as numbness or tingling, may considerably lower patients' satisfaction about implant surgery and, occasionally, result in medico-legal problems [2,5]. Therefore, the occurrence of post-operative altered sensation, such as paresthesia, has been widely considered as one of the success criteria of implant surgery [6,7], and recently, the diagnosis and prevention of altered sensation have been extensively discussed in the literature [1,8,9]. The occurrence of altered sensation has been investigated in previous reviews, as one of the general complications of implant surgery [10][11][12]. However, an estimation of the incidence of altered sensation after mandibular implant surgery, based on systematic review and meta-analysis, has not been reported. Based on clinical literature, the incidence of altered sensation varied considerably across studies, from 0.13% [13] to 37% [14]. Therefore, researchers can only estimate the incidence with a great degree of variability, such as 0-15% [8] or 0-40% [15]. In addition, the recovery rate of altered sensation, i.e. the proportion of the patients with altered sensation returning to normal sensation, has still remained unclear. More precise estimates of the incidence and the recovery rate of altered sensation, based on meta-analysis, would contribute to the risk assessment of surgery and patientdentist communication, during treatment planning. Finally, elderly patients usually have a severely atrophic ridge, with a short distance between the ridge and the IAN [16,17]. The influences of patients' age and the degree of ridge atrophy on the incidence of altered sensation have not been fully assessed.
The current systematic review and meta-analysis aimed to estimate the incidence of altered sensation after mandibular implant surgery. Specifically, we focused on the following aims: 1. First, we estimated the short-term (assessed postoperatively or within 10 days after implant placement) and long-term (assessed 1 year after implant placement) incidence of altered sensation after mandibular implant surgery.
2. Secondly, we estimated the recovery rate of initial altered sensation, i.e., the proportion of the patients with altered sensation returning to normal sensation.

Materials and Methods
The review and meta-analysis were performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Protocols (PRISMA-P, see S1 Table for the checklist) [18]. The review protocol has been registered based on PROSPERO (http://www.crd.york. ac.uk/PROSPERO) (CRD42015016250) and the review was performed accordingly.

Eligibility Criteria
We selected the original research articles based on the following criteria: • Participants: The participants included in the review were the adults who were eligible for implant surgery, according to the criteria established in the individual studies. All participants received mandibular implant surgery.
• Outcomes: The individual studies should report the number of patients who showed altered sensation. Our primary outcome was the incidence of altered sensation, defined as the proportion (i.e., the event rate) of the number of patients who showed altered sensation to the number of total patients being assessed after the surgery. 'Altered sensation' was identified if one of the following conditions was reported: dysesthesia, paresthesia (including changes in sensation, such numbness or tingling), hypoesthesia, anesthesia, or sensory changes as assessed by sensory testing [9]. The outcomes were further categorized according to the time point of clinical assessment. We considered the altered sensation recorded postoperatively or within 10 days after implant placement to be short-term, and the altered sensation recorded 1 year after implant placement to be long-term. Additionally, we considered the altered sensation recorded 3-6 months after implant placement to be intermediate-term, based on the model of nerve injury [19]. The articles that reported altered sensation during the period other than the previous time points would still be included. The model predicts that spontaneous healing would occur within 4 months, if nerve sheet is not damaged by trauma (i.e., neuropraxia and axonotmesis) [19,20].

Search Strategy
We performed a computerized search of original research articles on the electronic database MEDLINE and The Cochrane Library, with the keywords related to implant surgery and altered sensation. It should be noted that, because the term 'altered sensation' may refer to different experiences in studies, we included a variety of keywords related to altered sensation in our search, including allodynia, hyperalgesia, dysesthesia, paresthesia, sensory disturbance, numbness etc. (see Table 1 for the complete search strategy). Additionally, a manual search was performed by screening the reference list from the previous reviews on implant-related altered sensation [12,21,22]. The findings from the computerized and the manual search were pooled for subsequent selection. Search was limited to the journal articles published in English, during the period from 1990 Jan.1 to 2016 Jan. 1. Case reports were excluded.

Study Selection
We manually screened the initially found articles by excluding those who focused on the following topics: transmandibular implant surgery, zygomatic or orbital implant surgery, nerve transposition, nerve lateralization, distraction osteogenesis, osteotomy for orthodontic purposes, or the bone-harvesting surgery without implant placement. The articles that no full text can be retrieved, according to PubMed, were also excluded (Fig 1). Review articles were excluded. Two authors (CSL and SYW) independently conducted literature search and the assessment of eligibility and study selection, primarily according to the title and the abstract of the selected articles. Decisions were made by a consensus between the authors. The selected articles were sorted using Endnote (Thomson Reuters, New York, NY, USA).

Data Collection and Extraction of Data Items
Printed or electronic full text of all the included studies was retrieved. The following data items were manually extracted: (A) Demographic and clinical characteristics (5 items, see Table 2), including (1) the design of the study (e.g., a randomized control trial of a retrospective/prospective cohort study), (2) the number of patients (total and by each sex), (3) the mean or the range of patients' age, (4) patients' ridge condition, focusing on alveolar bone height, the ridge-to-IAN canal distance (RCD), the Cawood's class of mandible atrophy [23], and the Lekholm and Zarb's class of jaw bone quality [24], (5) the surgical methods for ridge augmentation.
(B) Methodological characteristics (4 items, see Table 3), including (1) the length and diameter of implant (2) pre-operative imaging examination, focusing on the use of panoramic radiograph or tomograph/computed tomography, (3) the methods for assessing altered sensation, and (4) the specific experience of altered sensation (e.g., paresthesia or hypoesthesia) being assessed.
(C) Outcomes of altered sensation (4 time points and 3 items for each time point, see Table 4), including (1) the time point when the investigation was performed, (2) the number of the patients initially showing altered sensation, and (3) the number of total patients being investigated, and. The data were extracted separately for short-term, intermediate-term, longterm outcomes, and the outcomes reported at other time point. Therefore 12 items in total were recorded (Table 4).
It is noteworthy that some of the studies have reported the occurrence of altered sensation from more than one time point. For example, 6 studies have reported both short-term and long-term outcomes from the same cohort of patients [14,21,[25][26][27][28]. Statistically, this overlap would increase the dependency between the results of different time point (see Discussion). However, because it is common for the clinical studies to examine patients' symptoms for more than one time point, we here separately treated the outcomes from the same studies for different analyses. For example, the results from the study by Boven et al. ([25]) would be analysed for both short-term and long-term altered sensation.

Assessment of Study Quality and Risk of Bias
Assessment of study quality and risk of bias was performed according to a 7-item customized assessment sheet, which was modified from the Newcastle-Ottawa Quality Scale (NOQS) [29], a scale specific for non-randomized cohort study. The customized scale assessed the representativeness of cases, the research methods and the outcomes, of an individual study (see Table 5 for details the items).

Data Synthesis
We calculated the pooled event rate using Metaprop [30], a Stata-based program specifically designed for binominal data. The confidence intervals (CIs) were calculated using an exact binominal approach. Conventionally, a proportion, such as the event rate of altered sensation, can be approximated by the normal distribution. However, when the proportions are close to the boundary (i.e., 0 and 1), the normal distribution may not be a good approximation of proportions [30]. We did not adopt the approximation approach, because in our meta-analysis, the event rates of altered sensation were 0 or 100% in several studies. Specifically, the analysis was performed based on Stata version 13.1, using the program metaprop_one [30]. We used the exact method (i.e., the option cimethod(exact)) to calculate the confidence intervals of each study. A random-effect analysis was performed, (i.e., with the option logit) so that a binomial distribution is used to model the within-study variability and the parameters were estimated using a maximum likelihood procedure. Two analyses were performed to estimate the incidence and the recovery rate of altered sensation: 1. Analysis 1: We investigated the incidence of altered sensation by calculating the pooled event rate of altered sensation, using a random-effect model. The analysis was performed respectively for (1) short-term (within 10 days after surgery) and (2)   The number in the brackets denotes the number of responders. 2 Degree of bone atrophy was assessed by Cawood's classification; degree of bone quality was assessed by Lekholm and Zarb's classification; RDC and height is measured in mm. 3 The study adopted a within-subject split-mouth design. 4 Mean age was calculated based on Table 1 of the study. 5 Bone quality was assessed respectively for two study groups. 6 Most implants were placed with bone quality between 1 and 3, based on Table 5 of the study. 7 The bone quality of 57% of the mandibles was of grade 3 or 4. surgery) can be too ambiguous, and therefore would result in great between-study heterogeneity. Between-study heterogeneity was assessed using τ 2 statistics. Here the value τ is variance in incidence between studies (here treated as a random effect) [31,32]. Statistical significance of τ 2 was assessed using a random-effect model vs. fixed-effect model likelihood ratio (LR) test. We adopted a random-effect model effect because of the statistical heterogeneity between the studies (see Results) and the clinical heterogeneity due to experimental design and sampling [31].
2. Analysis 2: We investigated the recovery rate of initial altered sensation. Recovery rate was defined as the ratio of the patients whose altered sensation restored to normal sensation to those who had reported altered sensation postoperatively. Recovery duration was defined as the time when the clinical assessment was performed and normal sensation was recorded. The pooled recovery rate was calculated, respectively, for recovery duration ≦ 6 months and recovery duration = 1 year, using a random-effect model. These outcomes from were extracted based on the descriptions of the results or discussion in the original articles. For example, in one study (Study ID = 4), the authors have explicitly stated that 'the maximum duration of these sensory disturbances was 6 weeks' [33]. Therefore, we considered that all the patients who initially reported altered sensation (5 patients) returned to normal sensation within 6 months.

Assessment of Publication Bias
The degree of publication bias was visualized with funnel plots, and the degree of asymmetry of the funnel plots were estimated based on the Egger's regression test [34]. The analyses were performed using statistical software Comprehensive Meta-analysis 2 (Biostat, Engelwood, NJ).

Results of Search and Study Selection
Our search yielded 196 articles, including 188 articles from computerized search and 8 articles from manual search (Fig 1). After screening, 170 articles were excluded for a variety of reasons (see S2 Table for the reasons of excluding the articles). 26 articles were finally included. Among these, 2 articles (Study ID = 5, 6) reported the findings about two different surgical procedures [35,36]. These articles were each considered as two studies, and thus 28 studies were included (Fig 1).

Demographic, Clinical and Methodological Characteristics of the Included Studies
As shown in Tables 2 and 3, the included studies demonstrated a considerable degree of heterogeneity in the demographic, clinical and methodological aspects. The studies were predominantly an observational cohort study. Regarding the demographic and clinical aspects (Table 1), there were more female patients recruited in the studies, and the age was relatively elder, mostly 40-60 y/o. Patients' mandibular ridges were moderate-to-severely atrophic, with inadequate bone height and/or poor bone quality. Reduction in the ridge-to-IAN canal distance was noted in several studies [35][36][37]. Augmentation procedures were performed in 5 studies [25,33,35,36,38]. Regarding the methodological aspects (Table 3), in most studies, an implant with a regular size (i.e., >10mm in length and 3.75-5mm in diameter) [39] was used. Panoramic radiography was frequently taken preoperatively, either alone (n = 9) or together with tomograph/computed tomography (n = 4). Regarding the assessment of altered sensation, questionnaires were used in 6 studies, and sensory testing was performed in 8 studies, among which a light touch test was the most frequently performed (n = 6). Paresthesia was the most recognized experience of altered sensation (n = 10), followed by pain (n = 6), anesthesia (n = 5), hypoesthesia (n = 4), and dysesthesia (n = 2).

Study Quality and Risk of Bias
Based on the results of study quality and risk of bias assessment, we found 3 studies with a score between 0 and 2 (Study ID = 2, 7, 12), 4 studies with a score 3(Study ID = 13, 16, 20, 26), and 21 studies with a score between 4 and 7 (for details see S3 Table). The average score of the assessment was 4.1±1.4 (±standard deviation), based on our customized scale (Table 5). In general, the study quality and risk of bias was moderate, with a substantial degree of variability across the studies. Regarding the representativeness of sampling, 50% of the studies have provided clear information about both the inclusive and exclusive criteria of patient selection. 64% of the studies have clearly stated the source of patient sampling (e.g., from multiple centers or a single hospital). 79% of the studies have clearly stated the surgical methods for implant placement. Regarding outcome assessment, only 36% of the studies have clearly defined the experience of altered sensation before assessment, and 54% of the studies have clearly stated the methods for assessment. Finally, 68% and 64% of the studies have clearly stated the duration of follow-up and success/survival rate of surgery, respectively (for detailed results see S3 Table).

Analysis 1: Estimating the Short-term and Long-term Incidence of Altered Sensation
The meta-analyses included 16 studies for estimating the short-term incidence (  [33]. The outcomes from these studies were categorized as 'Other time point of assessment' in Table 4. These outcomes would be discussed descriptively and not be included in the meta-analysis. It should be noted that the results of study quality and risk of bias showed a higher level of variance, which may influence the pooled estimation of incidence. Therefore, we conducted data synthesis after excluding the studies with a lower score (0-2) of the study quality and risk of bias assessment. After the adjustment, 3 studies (Study ID = 2, 7, 12) were excluded for the analysis of short-term incidence and 1 study (Study ID = 2) was excluded for the analysis of long-term incidence. The short-term incidence was 13% (95% CI, 6%-25%, Fig 2A) and the long-term incidence was 3% (95% CI, 1%-7%, Fig 2B). A significant degree of between-study heterogeneity was found for both the short-term analysis (LR test: χ 2 (1) = 348.2, p<0.01, τ 2 = 2.13) and the long-term analysis (LR test: χ 2 (1) = 23.6, p<0.01, τ 2 = 1.86). Without the adjustment for study quality and risk of bias, the short-term incidence was 11% (95% CI, 5%-21%; LR test: χ 2 (1) = 350.9, p<0.01, τ 2 = 2.07) and the long-term incidence was 3% (95% CI, 1%-7%; LR test: χ 2 (1) = 22.3, p<0.01, τ 2 = 1.60).

Analysis 2: Estimating the Recovery Rate of Altered Sensation
Fifteen studies that provided the information about recovery rate were included in the analysis (for detailed results see Table 6). Note that the studies included in Analysis 2 ( Table 6) may not be identical to the studies included in Analysis 1 (Table 4). For example, one study (Study ID = 16) reported the patients with postoperative altered sensation, but it did not explicitly state the duration that the patients recovered from altered sensation [40]. These studies were excluded in Analysis 2. For another example, one study (Study ID = 4) reported the duration of recovery, but it did not explicitly state the time point of assessment [33]. These studies (Study ID = 4, 9, 18) would be included in Analysis 2 but not in Analysis 1. As shown in Fig 3A, the studies have shown a declined incidence of altered sensation during the follow-up sessions. After the excluding the studies with a lower score (0-2) from the study quality and risk of bias assessment (Study ID = 2, 12), 9 studies (Study ID: 4, 5a, 5b, 8,9,14,15,20,21) reported the recovery rate within 6 months after surgery with the pooled recovery rate 80% (95% CI, 52%-94%, Fig 3B). A significant degree of between-study heterogeneity (LR test: χ 2 (1) = 7.9, p<0.01, τ 2 = 1.30) was found. Five studies (Study ID: 14,20,21,22,25) reported the recovery rate 1 year after surgery (Table 6), with the pooled recovery rate 91% (95% CI, 78%-96%, Fig 3C). The between-study heterogeneity was not statistically significant (LR test: χ 2 (1) = 2.26, p = 0.07, τ 2 = 0.64). Without the adjustment for study quality and risk of bias, the  recovery rate within 6 months was 77% (95% CI, 53%-91%; LR test: χ 2 (1) = 7.0, p<0.01, τ 2 = 1.05). The recovery rate for 1 year did not differ before and after adjustment, because no study was excluded in this analysis.

Publication Bias
Regarding the analysis of short-term and long-term incidence, the Egger's regression test showed significant asymmetry (P = 0.03 and P = 0.03, respectively) of the funnel plot. Regarding the analysis of intermediate-term, Egger's regression test did not show significant asymmetry (P = 0.13) of the funnel plot (see S1 Fig for the funnel plots).

Incidence of Altered Sensation and the Recovery Rate after Mandibular Implant Placement
Estimates of the incidence of altered sensation varied dramatically across previous studies. A frequently cited number is 8.5% according to Bartling and co-workers, based on the results of subjective assessment and sensory testing of 94 patients [41]. An earlier retrospective research by Ellies reported an incidence 37%, based on the questionnaires from 226 multi-center patients [14]. Another earlier systematic review by Berglundh and co-workers estimated that 41% of the 159 reviewed studies have reported an occurrence of persistent (lasting more than 1 year) sensory disturbance [10]. In contrast, according to Vazquez and co-workers, based on a larger cohort of patients (n = 1527), the incidence was very low (0.13%) [13]. Therefore, most researchers would estimate the incidence with a great degree of variability, such as 0-15% [8] or 0-40% [15]. The huge variation may be associated with the heterogeneity in research methods. While some studies only focused on persistent changes (e.g., disturbances lasting for more than 1 year), others focused on transient ones (e.g., disturbances occurring immediately after surgery). We here estimated the short-term and long-term incidence of altered sensation. As expected, the incidence decreased through the three time points. Our meta-analysis results are consistent with the neurophysiological model of nerve injury, which classified nerve injury into three major classes: neuropraxia, axonotmesis and neurotmesis [19,20]. A transient and reversible injury without damaging the nerve proper and nerve sheath i.e., neuropraxia or axonotmesis, would spontaneously return to normal sensation within 2 to 4 months, as shown in most of the cases of altered sensation after mandibular implant surgery [42,43].  Furthermore, we found the recovery rate was high (80%) for the first 6 months after implant placement. One year after implant placement, an even higher proportion of patients (91%) may return to normal sensation (Fig 3B). The findings suggested that, even though the shortterm assessment showed a considerably high incidence (13%) of altered sensation, most of the cases would be resolved within 1 year.

Effects of Biological Risk Factors
The elderly patients usually have a severely atrophic ridge, with a short distance between the ridge and the IAN [16,17]. The influences of patients' age and the degree of ridge atrophy on the incidence of altered sensation have not been fully assessed. The results from the review suggested that short ridge-to-IAN canal distance may not influence the short-term incidence of altered sensation. It should be noted that in 4 of the short-RCD studies (Study ID = 5a, 5b, 6a, 6b), either augmentation procedures were performed or a short implant (5mm or 7mm) was used, in order to compensate the inadequate bone height [35,36]. The finding suggested that ridge atrophy (RCD < 10mm) may not significantly increase the risk of altered sensation, given that augmentation procedures are performed or a shorter implant is used. The finding highlighted the importance of pre-surgery treatment planning. The anatomical relation between the implant and the IAN, especially the mental nerve branch, plays a critical role in the occurrence of altered sensation [42,43]. Imaging techniques, such as the computed tomography would be valuable tools for precisely evaluating the spatial relation between bone and nerve structures [44]. The problems are particularly critical to the elderly patients of implant surgery, because both the degree of ridge atrophy and bone quality is related with aging [45]. Our analysis did not reveal a significant association between age and the incidence of altered sensation. This negative finding implies that age per se may not have a direct effect on altered sensation. Instead, its effect may be associated with multiple factors, such as ridge morphology and bone quality/quantity. Interaction between these factors would require further investigation.

Limitations and Further Considerations
Due to the heterogeneity in study design, patient sampling and methods of assessment across the studies, conclusions from the current review and meta-analysis should be carefully interpreted with the following considerations: 1. For the biological risk factors, we here only evaluated the effect of age and ridge-to-IAN canal distance, since these factors can be precisely quantified. However, we did not investigate the overall physical and behavioral conditions, such as diabetes and smoking, which are known associated with implant outcomes [46,47]. For example, diabetes is associated with altered pain sensitivity in the elderly patients [48]. As shown in the assessment of risk of bias, few studies have systematically evaluated these risk factors during patient sampling. The effect of these factors would require further investigation.
2. Regarding the methods of assessment, we here did not differentiate the specific experiences of altered sensation. Clinically, dysesthesia or pain may indicate the development of neuropathy or a significant influence on patients' quality of life [49,50]. However, because most studies did not clearly state how these experiences were defined to the patients, we here pooled all the symptoms under an umbrella term 'altered sensation'. In addition, very few studies have investigated the quantity of altered sensation, for example, using a psychophysical scale (e.g., the numerical rating scale). Therefore, we cannot evaluate the change about the intensity of altered sensation through a period of time.
3. Regarding the method of data synthesis, it should be noted that the estimates of short-term and long-term altered sensation were from some overlapped studies-6 studies have reported both short-term and long-term results [14,21,[25][26][27][28]. Although this is commonly seen in the clinical studies about post-surgical results, in which the authors would usually examine patients' symptoms for more than one time point. The lack of independence may lead to an underestimation of standard errors and hence confidence intervals.
4. It should be noted that, in terms of surgical procedure, both drilling and implant placement may induce IAN injury. However, the included studies did not distinguish between the two conditions. It has remained unknown whether drilling or placement is the dominant factor to IAN injury. Apart from the physical conditions, altered sensation is both a sensory and an emotional experience [51]. Very few studies have reported the emotional aspects about altered sensation, and the psychosocial profiles about patients with altered sensation were mostly under-researched (except for [52,53], for example). Importantly, when patients perceive altered sensation after surgery, the increased dissatisfaction and negative emotion may results in liability claims. The dissatisfaction would be greater if the risk of altered sensation is not notified before surgery. The association between the affective-motivational factors and the medico-legal claims have not been fully investigated.

Clinical Implications
Based on our findings, we suggested the following aspects about improving the quality of dental implant therapy.
1. On dentist-patient communication about the risk of altered sensation: our findings showed that, for the implant surgery with conventional procedures, the risk to have altered sensation immediately after surgery is substantial (13%). Nevertheless, most of these initial symptoms would spontaneously return to normal. Our findings highlighted the importance of early assessment [11]. Importantly, when a patient reports altered sensation, regular assessment for 6 months would be critical to trace the changes of altered sensation.
2. On the timing of assessment: our findings suggested that an assessment during 3 to 6 months postoperatively would be valuable for subsequent management of altered sensation. Our finding echoed the proposal that if the symptom does not reduce by the time of 12 weeks, further intervention (e.g., neurosurgical referral) may be considered [42,43]. We suggested that follow-up assessment should be scheduled no shorter than 6 months, since most patients with initial altered sensation may return to normal at that stage.