Influence of Prognostic Risk Indicators on Osseointegrated Dental Implant Failure: A Matched Case-Control Analysis

The predictability of osseointegrated root form implants with well-documented and long-term functional and esthetic outcomes has led to their acceptance as a practical treatment modality in modern dentistry for rehabilitation purposes in fully edentulous and partially dentate patients, or for single-tooth replacement.

The success of implant treatment and maintenance of its highly desirable outcomes are dependent upon several prerequisites to accomplish a dynamic association of functional and esthetic results. These factors may include adequacy of bone quantity and quality, biocompatibility of implant biomaterial and characteristics of implant surface, absence of surgical complications, absence of peri-implantitis, and prevention of excessive loading. Moreover, direct and indirect systemic factors that influence the host response appear to be of great relevance to the prediction of risk groups for implant loss.1 

Despite a few contraindications and a documented long-term success rate, it is widely recognized that dental implants sometimes can be susceptible to healing complications or disease conditions that eventually may result in implant loss.2 Thus, the focus of implant research has shifted from demonstrations of clinical success to more specific attempts at identification of factors that may contribute to implant failure.3 

The aim of this study was to identify risk indicators associated with patients with at least one implant failure using a case-control study design.

Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines4 were followed in reporting this case-control study.

All patients who had implants placed and restored at the University Dental Hospital of Manchester, Manchester, UK, between February 2000 and December 2006 were eligible for inclusion in the study, regardless of medical health status, age, gender, or race. Only one inclusion criterion was required for the study group, namely, that participants had had at least 1 failed implant that was removed or lost/exfoliated, although for the control group, no implant losses should have occurred during the study period. Exclusion criteria included inadequate or unavailable patient records.

Control subjects were selected so that they resembled the cases with regard to the following characteristics:

• Cases and controls had the same number of implants, which were placed in the same year.

• Cases and controls were the same age and gender.

• All cases that had been treated with implants in combination with bone augmentation were matched to controls that had undergone the same augmentation procedures.

The following were recorded: smoking habits, alcohol consumption or other illicit drug use, and medical conditions that may compromise wound healing including immunosuppression, diabetes mellitus, systemic steroid treatment, osteoporosis, previous chemotherapy, and radiotherapy treatment.

These included opposing dentition in relation to the implant (natural teeth, removable denture, fixed prosthesis, or implant-supported prosthesis) and implant location in the maxilla and mandible (anterior and posterior sectors: anterior implants were considered when placed in the region of incisors and canines, posterior implants were regarded as those positioned in the region of premolars and molars).

These included the type, dose, and frequency of chemotherapeutic agents, such as antibiotics and analgesics, taken postoperatively. A minimum of a 5-day treatment course recorded on the patient's prescription was required for investigators to consider that the patient had received antibiotics. Chlorhexidine use was a variable recorded only if both preoperative and postoperative use was clearly noted by clinicians.

The number of placed implants, the implant system used, and the dimensions of the implant and abutment (length and diameter) were recorded.

These were grouped into two main categories: removable (overdentures) and fixed (bridge or single-crown restorations).

Any abnormal incidents that had occurred during implant placement or the healing period were recorded (cortical plate perforation, exposed implant surface, lack of primary stability, and bone graft–related complications). Whether problems (infection, pain, paresthesia, dehiscence, and cover screw loosening) were encountered before exposure of the implants was also recorded.

Peri-implantitis was recorded when clinical manifestations of peri-implantitis such as suppuration, bleeding on probing, reddening of the peri-implant mucosa, pain on percussion, increased depth of the peri-implant pocket, mobility, and radiographic evidence of advanced bone loss around the implant were noted in the patient's clinical records.

The primary outcome in this study was implant failure. Failure was defined as removal of the implant for any reason.5 Failure was grouped into five categories according to timing (failed before exposure, failed at exposure, failed between exposure and restoration, failed during the first year of loading, failed after the first year of loading).

A 2-group continuity corrected χ² test with a 2-sided significance level of .05 would have 80% power to detect a difference in implant failure between a proportion of 0.390 and a proportion of 0.100 (odds ratio, 0.174) when sample sizes were 25 and 73, respectively (total sample size, 98). The significance of the association between implant failure and each potential risk indicator would be determined by the χ² test at a .05 level of significance. The appropriate odds ratio would be calculated for each variable.

Data obtained were processed with the Statistical Package for the Social Sciences (SPSS) for Windows, version 15 (SPSS Inc, Chicago, Ill). Basic data analysis including descriptive statistics (frequency distribution, cross tabulations, and χ² tests) was used to produce a table of frequency counts and percentages for all values associated with each variable included in this study, and to examine the association between variables. In addition, independent sample t tests were used to compare the average sizes of implants and abutments between cases and controls. Logistic regression was used in this study because the implant failure outcome variable was binary (yes or no), and predictor variables were both categorical and continuous. Associations between smoking, alcohol consumption, postoperative antibiotics, chlorhexidine use, peri-implantitis, number of implants, implant length, implant diameter, abutment height, abutment diameter, type of prosthesis, opposing dentition, and implant failure were evaluated by fitting a univariate logistic regression model for each variable. Risk factors based on univariate analyses were then entered into a multivariate logistic regression model that was adjusted for inclusion of age, sex, number of implants, and bone grafting, to estimate odds ratios and corresponding 95% confidence intervals for each variable.

Between February 2000 and December 2006, a total of 663 patients were provided with dental implants at the hospital, and 31 patients who had failed implants were identified. However, only 22 patients were suitable for inclusion in this study owing to inadequacy of required data in patient records, or the fact that it was not feasible to find matched controls according to proposed case-control matching criteria.

Twenty-two patients who had experienced a loss of at least one implant (study group) were matched to 61 patients who did not lose any implants (control group). Eighteen cases were matched with 54 controls (ratio, 1∶3). Three cases were matched to 6 controls (ratio, 1∶2), and 1 case was matched to 1 control (ratio, 1∶1) because it was not possible to find 3 matching controls.

In the study group, a total of 78 implants were inserted in 22 patients; of these, 33 (42.3%) implants were lost, whereas in the control group, 212 implants were inserted in 61 patients and none were lost. Of 83 patients included in the study, 32 (38.6%) were males and 51 (61.4%) were females, with a mean age of 51 years (range, 21–87 years; SD  =  18.7). The largest percentage of individuals corresponded to the 41–60 year age range. No statistically significant differences in implant failure between older (>60 years) and younger patients (P  =  .98), or between males and females (P  =  .80), were observed as expected owing to matching of cases and controls. Certain characteristics of case group patients and of failed implants are summarized in Table 1.

Most subjects in study group 14 (63.6%) and control group 44 (72.1%) were healthy, with no illnesses or medical conditions noted in the medical records. Although well controlled, diabetes mellitus and cardiovascular conditions were the most frequently observed conditions, and several conditions were found in the selected population, although in small prevalences. A summary of the medical history of cases and controls is provided in Table 2. No statistically significant difference in implant failure was found between patients with and without systemic conditions (P  =  .19). On the other hand, a significant difference in implant failure was noted between smokers and nonsmokers (P  =  .004). Similarly, heavy drinkers (>5 units/d) had a statistically higher percentage of implant failure compared with nondrinkers and patients who reported consumption of a lesser amount of alcohol (<5 units/d) (P  =  .002).

Jaw site (maxilla and mandible in their anterior and posterior sectors) was not significantly related to implant failure as failure was considered per patient (P  =  .21).

No significant difference was observed between groups with respect to bone grafting (P  =  .92). Autogenous bone harvested from the iliac crest was most frequently used (90%) for reconstruction of the atrophic mandible and maxilla.

Although most patients (16; 72.7%) included in the study group had implants opposed by natural dentition or an implant-supported prosthesis, no significant influence of type of opposing dentition on implant failure was found (P  =  .31).

Although a significant difference was observed between patients who had and those who did not have postoperative antibiotics with respect to implant failure (P  =  .04), chlorhexidine use did not significantly influence implant failure (P  =  .46).

The fixtures were rehabilitated with 11 single crowns and bridges and 11 dentures in the case group, and with 25 single crowns and bridges and 36 dentures in the control group. The type of prosthesis (fixed vs removable) used to restore implants did not have a significant effect on implant failure (P  =  .46).

The number of placed implants did not influence implant failure (P  =  .99). Approximately two-thirds of cases experienced 1 failure, and only 2 patients lost 4 or 5 implants, respectively.

Whereas the most frequently used implant system in the case group was Frialit-II (87.5%), the Astra system was used in most controls (76.7%). Other systems were considered but were excluded from the final model as they were rarely used; these included IMZ (7.2%) and 3i (1.2%).

The mean length of the inserted implants was 12.2 mm (minimum 10 mm, maximum 15 mm) for the study group, and 12.8 mm (minimum 9 mm, maximum 15 mm) for the control group. This means that approximately 0.6 mm longer implants were inserted in the control group compared with the study group. However, this difference between groups was not statistically significant (t test, P  =  .17). By contrast, the diameter of the implant was found to significantly influence implant failure (t test, P  =  .001); the average diameter in the case group was 3.7 mm, and in the control group, 4.1 mm. As far as the abutment size is concerned, no significant differences were found between cases and controls with respect to the height (t test, P  =  .07) or diameter (t test, P  =  .37) of abutments.

Statistical analysis of notes made at the appointment of implant placement or during the subsequent healing period revealed that the total frequency of abnormal incidents registered in the case group (36.4%) was more than twice that of the control group (14.8%). However, the difference was not statistically significant between groups (P  =  .14). Table 3 summarizes the frequencies of abnormal incidents. On the other hand, implant failure was significantly related to the presence of complications that occurred before exposure (P  =  .001). Although half of cases experienced 1 or more of the complications listed in Table 4, only 5 (8.2%) controls were associated with occurrences of such problems. Associations between the frequency of these complications and the use of postoperative antibiotics were also studied. It was observed that postoperative antibiotic administration was significantly associated with fewer complications encountered during the early healing period (P  =  .002). The frequency of complications before exposure in relation to use of postoperative antibiotics is demonstrated in Table 5.

Fifteen of 22 cases (68.2%) showed signs of infection and were diagnosed with peri-implantitis, whereas none of the controls had peri-implantitis. A highly significant difference (P < .001) for peri-implantitis was found between patients with failed implants and matched controls.

As far as the timing of implant failure is concerned, the incidence of failure among cases was 41% before loading (early failures) and 59% after loading (late failures). More than half of the late failures (54.5%) occurred after the first year of function.

Univariate logistic regression associations between study variables and implant failure indicated that variables associated with implant failure (P < .001) included current tobacco use, heavy alcohol consumption, use of postoperative antibiotics, and implant diameter. Moreover, in the multivariate regression model with age, gender, bone graft, and number of placed implants included in the analysis, all variables already found related to implant failure in the univariate model remained statistically associated with implant failure. A summary of univariate and multivariate logistic regression results can be found in Table 6.

The purpose of this study was to identify risk indicators associated with implant failure in a sample of patients treated at the University Dental Hospital of Manchester. To reduce bias when case and control groups were compared, patients in the control group were matched to the study group. This indicates that control patients were identified and then were included in such a way that both groups were as identical as feasible with regard to age, gender, number of placed implants, timing of implant placement, and reconstructive surgery (bone grafting procedures).

In the study group, 41% of cases lost their implants before loading (early failures) and 59% after loading (late failures), with more than half of late failures occurring after the first year of function. This is in accordance with results reported by Esposito et al6 in a meta-analysis of several studies, indicating a chronological distribution of 47% as early and 53% as late failures, of which 55% were detected after the first year of loading. This observation also supports several reports indicating that a higher rate of late implant failure accounted for the incidence of disease conditions such as peri-implantitis that eventually may lead to implant loss.3 It is believed that this disease may require a certain time to reach clinical significance and an even longer period to necessitate removal of the implant. Therefore, it is more likely to predict that the hazard rate for implant failure may increase as longer observation times are applied. In this study, the higher incidence of this condition for failed implants as compared with successful implants was positively clear.

Clusterization is a technical term used to describe the phenomenon in which a few individuals can concentrate risk for implant failure and subsequently experience multiple losses. In the present study, the prevalence of clustered failure was observed in 6 (27.8%) of the cases. A combination of the presence of peri-implantitis or illness (diabetes mellitus) and heavy smoking habits could explain implant failure in these subjects. These observations suggest the existence of systemic factors that may affect the survival of all implants within a given patient, leading to multiple implant failures. This finding is consistent with those of other studies.7,8 

No statistically significant difference among patients was observed for age. Lack of association between implant failure and age was previously reported.9 Although males were more prone to lose implants in previous investigations,10,11 female gender was considered by other authors to be a risk factor for implant failure.12,13 Consistent with other reports in the literature,14,15 no significant association was found between patient gender and implant failure in the selected sample of patients.

Several medical conditions that were recorded in the study group such as diabetes mellitus and Sjögren's syndrome were found by some authors to play an important role in early implant failure.16,17 However, general health conditions did not significantly influence implant failure observed in the present study. This observation is in accordance with the findings of 2 larger studies18,19 and a consensus review of the literature.1 

The adverse effect of smoking on the peri-implant tissue condition has been described in a number of studies.20,21 In the present study, a statistically significant difference was noted between smokers and nonsmokers with regard to implant failure. This may be explained by the detrimental effect of smoking on the wound healing process, particularly in the early stage of osseointegration.

As far as alcohol consumption is concerned, individuals who routinely consume alcohol may experience a delay in the healing of surgical wounds, as consumption of alcohol was found to be associated with deficiencies in the complement system and suppression of activation and proliferation of T lymphocytes, as well as with adherence, mobility, and phagocytic activity of monocytes, macrophages, and neutrophils.22 Moreover, certain substances contained in alcoholic drinks such as ethanol and nitrosamines may cause bone resorption and interfere with the stimulation of new bone formation.23 In the present study, it was observed that alcohol consumption was significantly associated with implant failure.

Characteristics of bone quality and quantity at the implantation site in relation to anatomic location are among those factors that seem to significantly influence implant failure.24 However, several authors25,26 observed no significant differences when comparing the failure rates of both jaws, or of the anterior and posterior zones. In contrast, other authors27,28 reported different results, indicating higher failure rates for implants placed in the maxilla and in posterior segments of both jaws. In this study, no significant differences in failure rate were recorded between the maxilla and the mandible.

The effectiveness of the prophylactic use of antibiotics in conjunction with implant surgery and its correlation with success or failure rates are poorly investigated in the literature; a lack of randomized controlled clinical trials is documented. It has been suggested that little or no benefit is derived by providing antibiotic coverage for implant placement.29 On the other hand, Dent et al30 reported that significantly fewer failures occurred when preoperative antibiotics were used. Although evidence-based support is still missing, a recent Cochrane review31 reported that some evidence suggests that a single oral dose of 2 g of amoxicillin given preoperatively an hour before implant placement can significantly reduce failure of dental implants. In the present study, a significant difference was found, with fewer patients experiencing implant losses when prescribed antibiotics postoperatively as compared with patients who had not received any antibiotics. This was further supported by the finding of a significant decrease in occurrence of complications during the early healing period when patients had received postoperative antibiotics.

A review of the literature6 found that the success rates of implants placed in the setting of reconstructive grafting procedures ranged from 60%–100%. Based on the inconsistency of these reported estimates of implant survival, it is reasonable to doubt whether the bone graft itself is an independent risk factor for implant failure. Similar to other studies,14,32 no negative influence for bone grafts on implant failure was observed in the present study. In contrast, in a retrospective investigation,33 a possible correlation between multiple implant failures in bone-grafted patients and relative bone mass density was observed. Furthermore, it was suggested that implant failure might be related to the complexity of the procedure and to the presence of bone graft–related complications.6 These complications were recorded in only 1 of 37 patients (44.6%) who underwent reconstructive procedures. In addition, the delayed approach used in the present study, which allowed for at least 3 months of healing time before implant placement, was believed to provide time for graft maturation and might have resulted in greater predictability of the treatment, especially when the residual ridge was insufficient for initial implant stability. Moreover, it could be hypothesized that delayed implant insertion in a 2-stage surgical approach might have permitted implant placement in more ideal positions and angulations after reconstruction of the deficient implant site.

It is generally agreed that greater implant length and diameter may increase the contact surface area at the bone-to-implant interface, subsequently resulting in better survival rates.34 Nevertheless, several authors17,35,36 failed to observe this association. In the present study, implant failure was significantly associated with implant diameter but not with implant length. Moreover, no associations were found between abutment size (height and diameter) and implant failure. The correlations between implant or abutment size and implant failure in relation to qualities of the specific implant placement site (thickness, height, density) and prosthetic needs were not studied. Also, several implant systems with different surface characteristics were included. These factors should be kept in mind when results of this study are interpreted.

Outcomes derived from well-designed retrospective studies are important for clinical knowledge that constitutes a reference point for specialists in planning, performing, and subsequently evaluating dental implant procedures. However, retrospective studies rely on the completeness and validity of data entered into patient records.37 Although data in patient records may be misfiled, misplaced, discarded, or simply missing, this lack of information was not a major restricting issue in this study. Fortunately, note entries were made by a small number of staff members at the implant clinic, where a standardized and efficient approach is put in place to ensure appropriate registration of information included in the patient notes. Nevertheless, the possibility of incomplete recording of information persists but is thought not to be statistically significant owing to the availability of key variables of interest in this study of risk factors related to implant failure.

With regard to statistical data analyses in this study, an appropriate multivariate regression analysis was necessary for better control of confounding factors with a great emphasis on a sound study design and appropriate case-control matching criteria. Despite robust statistical analyses of dental implant failure data, conclusions that can be drawn from retrospective studies may be limited in terms of generalization to a larger population. A limitation of this study is the descriptive definition of implant failure used in this study, that is, implant removal for any reason. Although definite, this did not address situations such as the nonfunctional implant, the nonrestorable implant, the failing implant, the implant associated with progressive subtle bone loss without clinically detectable mobility, or patient satisfaction. Unfortunately, in the setting of this study and given its retrospective design, it was not feasible to consistently ascertain implant failure in terms other than implant removal. The retrospective nature of this report should be taken into consideration when its results are interpreted; prospective case-control studies may be more appropriate and are needed to assess the risk of implant failure in conjunction with identified predictors.

Improving knowledge about the risk factors associated with implant failure is of clinical value in improving implant treatment outcomes. Knowledge of patient-related factors may help the clinician in patient selection, treatment planning, and obtaining appropriate consent of a patient for treatment. Equally important, careful preoperative assessment of implant-based risk could be useful in confidently selecting the type, size, location, and number of implants and in precisely designing the prosthesis according to individual prosthetic requirements. In summary, this study identified smoking and alcohol consumption as risk factors for implant failure. It was also found that the use of postoperative antibiotics or wide-diameter implants may reduce implant failure.

STROBE

Strengthening the Reporting of Observational Studies in Epidemiology