Long term clinical performance of 10 871 dental implants with up to 22 years of follow‐up: A cohort study in 4247 patients

Abstract Background The present retrospective study was aimed to assess the long‐term clinical performance of dental implants in a cohort study of 4247 patients. Methods A longitudinal observational cohort study was done on all implants performed by a single periodontist from July 1995 to April 2019. The main outcome variables of this study were implant failure and marginal bone level around implants. Results The study participants received a total of 10 871 implants with a mean of 2.56 implants per patient. The cohort was followed‐up to 22.2 years (mean = 4.5 ± 4.2). Among the 4247 patients of the current study, 140 patients (3.3%) experienced a combined total of 178 implant failures. According to life table analysis, at the implant level the cumulative survival rate at 3, 5, 10, and 15 years was 98.9%, 98.5%, 96.8%, and 94.0%, respectively while at patient level was 97.4%, 96.7%, 92.5%, and 86% at 3, 5, 10, and 15 years. Patients with multiple units were at a greater risk for having an implant failure. Baseline bone level was 0.09 ± 0.28 mm while at 8–10 years the mean bone level was 0.49 ± 0.74 mm. The incidence of peri‐implant mucositis at the implant level was 9.4% at 2–3 years, 9.3% at 4–5 years, 12.1% at 6–7 years, and 11.9% at 8–10 years. The incidence of peri‐implantitis was 2%, 2.6%, 3.2%, and 7.1% at 2–3, 4–5, 6–7, and 8–10 years, respectively. Cigarette smoking and diabetes mellitus were positively correlated with implant failure. Conclusions Though the results are promising and encouraging in terms of survival and bone level over time, it is important to emphasize the potential risk factors and consider them prior to dental implant placement.


| INTRODUCTION
Edentulism is a serious health problem involving functional, esthetic, phonetic, and psychological problems. 1,2 Despite great achievements in global oral health, edentulism remains a major and irreversible problem affecting the quality of life.
Globally, it was reported in 2017 that there were 3.5 billion cases of oral conditions, of which 2.3 billion had untreated caries in eral Dentists and Specialists in the Calgary, Alberta region. All measurements were taken by the same examiner who placed the implants (DF). The inclusion criterion was partially or fully edentulous sites, and the only exclusion criterion was patients with ASA class 3 or higher. Data regarding medical, and dental status before surgery were available for analysis. These parameters (e.g., smoking, diabetes mellitus) were considered as baseline factors. The investigated variables were grouped into: implant characteristics (e.g., length, diameter), surgical site (e.g., location, etc.), procedures (e.g., insertion torque, augmentation, etc.) and prosthetic variables. Dates of the following clinical events were recorded: implant placement, stage 2 (3 months after implant placement and prior to prosthetic connection) and all follow-up visits including the last (most recent) date the patient was seen as well as implant removal where applicable. The main outcome variable of this study were implant failure and marginal bone level around implants. Failure at implant level was defined as the removal of an implant for any reason. Failure at patient level was defined as a patient that experienced at least one implant failure during the followup period. Early failures were defined as failures occurring before implant loading, while late failures occurred after loading. Survival time was defined as the time from implant insertion to implant removal or to last follow-up for surviving implants.
Bone level measurements around implants had been performed at stage 2 surgery, and in years 1, 2-3, 4-5, 6-7 and 8-10 after installation as previously described. [12][13][14] Peri-implant soft tissue was evaluated by probing with a light vertical probe force of 17 g using a calibrated force automated probe or manual probe calibrated to about 17 g; each with a probe tip width of 0.45 mm at six locations around the implant. The soft tissue condition based on probing was determined using the implant mucosal index (IMI) whereby 0 = no bleeding, 1 = minimal single-point bleeding, 2 = moderate multipoint bleeding, 3 = profuse multipoint bleeding, and 4 = suppuration. 14,15 Scores were applied to each implant as the worst point during entire implant follow-up period. Peri-implant mucositis was defined as IMI ≥ 2 not accompanied with bone loss What is known: • Dental implants are known to have high survival and success rates.
• Long-term, large-scale, "real life" follow-ups and documentation are needed to better understand the behavior of dental implants over time as well as the factors influencing the survival and success of dental implants.
What this study adds: • This long-term, large-scale, "real life" retrospective analysis provides a statistical analysis of factors related to dental implants' survival and success.
• Though the results are promising and encouraging in terms of survival and bone level over time, it is important to emphasize the potential risk factors and consider them prior to dental implant placement. It is of utmost importance to highlight the role of proper preparation and maintenance for the long-term outcomes.
whereas peri-implantitis was defined as bone loss ≥1.0 mm in conjunction with IMI ≥ 2 at any follow-up after the stage 2 baseline. Hazard ratios (HR) were calculated in order to estimate the association between explanatory variables and failure time. Hazard ratio for categorical variable is defined as the ratio between hazards for implant failure among one group compared to another group. A ratio equal to one means that hazards are equal across groups while HR < 1 and HR > 1 means protective and risk effect, respectively. Hazard ratios were obtained by constructing the Cox proportional hazard (PH) regression model. All explanatory variables of the current study were evaluated one by one in a univariate analysis. Variables that were significantly related to failure in a univariate analysis were incorporated into a multivariate model in order to account for confounding effect between certain variables. In our model we accounted for possible Intra cluster correlation (as a result of multiple implants within certain patients) by calculating sandwich type robust standard errors.

| Statistical methods
Lastly, to use the Cox model, it was essential to check the underlying PH assumption, which states that HR is constant throughout the time under investigation. In the current analysis, the PH assumption was tested by using the Grambsch-Therneau test. In case of violation, we

| RESULTS
Overall, the study cohort included 4247 patients (56.4% females) with a mean age at surgery equal to 53.8 ± 13.5 years. The study  Modeling of time until failure as an outcome variable revealed several significant associations. Table 2 presents the results of a univariate models for the study exploratory variables. Significant variables were incorporated into a multivariate model (Table 3). Hazard ratio (HR) for time until failure, when comparing implants in patients with multiple implants vs implants in patients who only have a single implant, was 5.85 (p < 0.00001; Figure 4A).
As could be seen from Figure 4B, 6 mm implants were at greater risk for failure than longer implants (HR = 3.53, p < 0.001). Immediate implantation (n = 1254) was a significant indicator for early implant failure, but the effect disappears after 10 years postsurgery ( Figure 4C). Finally, implants combined with a GBR procedure were at greater risk for failure (HR = 1.85, p < 0.001) both in terms of early and late failures ( Figure 4D). At patient level, heavy smokers and Diabetic patient are at a greater risk for experiencing a failure during implant service (Table 3).
At stage two (n = 10 429) the mean bone level was 0.09 ± 0.28 mm while at 8-10 years (n = 1965) the mean bone level was 0.49 ± 0.74 mm (Figure 4). Bone loss following the first year was F I G U R E 2 Conditional frequency distribution of implant diameter by location  Throughout the study period, bone loss between two successive time points was prominent between stage 2 and year one. It slowed down until the fourth to fifth year. After that, bone loss was negligible and remained near to zero until 8-10 years of follow-up, which again had greater amounts of bone loss ( Figure 5).

| DISCUSSION
Implant therapy is regarded as a safe and reliable method of treating patients with complete or partial edentulism. The use of dental implants as a replacement for missing teeth has been increasing steadily, probably owing to the high predictability and survival rates, as reported in numerous studies. 4,16 Given the increasing popularity of dental implants, it is highly important to have a long-term "real-life" evaluation and analysis of this treatment performance.
The current study could serve as an example for a long-term methodological data collection and analysis that can be performed in niably critical for patient care; however, we must be able to apply it and therefore, there is a need for more implementation science in dentistry.
Overall, the survival rates of the implants in the long-term evalua- Immediate implantation was found to be a significant indicator for implant failure, but the effect disappears after 10 years postsurgery. This might imply that even though immediate implant placement should be considered as a risk factor for early implant failure, once the implant is stable and functioning for as long as 10 years following placement, it is no longer considered as a risk factor.
Finally, implants combined with a GBR procedure were at greater risk for failure (HR = 1.85, p < 0.001). It is highly crucial to realize that, as shown from the results of this work, implants in augmented bone are not as successful as implants placed in native bone and this risk presented not only in short term where infection may play a role but also longer term whereby the implant may be more at risk for periimplantitis or load related failures. This should be considered as part of the overall treatment plan and should be also shared with the patient.
Peri-implant mucositis and peri-implantitis are important entities that were observed in this current report. Long-term evaluation and follow-up for every implant patient are particularly important in order to identify early signs of these conditions. Early detection and proper intervention are crucial for favorable treatment outcomes. 21 It is noteworthy, that the present report is based on data from a periodontal practice that puts strict emphasis on prevention, proper periodontal stabilization prior to implant placement as well as longterm maintenance program and follow-up. This might be one of the reasons for the rather predictable results shown. It is of utmost importance to highlight the role of proper preparation and maintenance for the long-term outcomes. 19 Retrospective studies, as their nature might present some risk of bias, which is a limitation of this study, however, studies like that are still important to assess risk factors over a long-term follow-up of a large number of patients and implants. Some of these limitations are related to confounding factors that cannot always be identified in retrospective analysis of cases. Multivariate analysis is an attempt to control some of the confounders but bias can still be present as part of the retrospective nature of this study.
Long-term data from other practice-based groups will enable comparison of the results and further analysis of confounding factors.

| CONCLUSIONS
This study reported on long term follow-up and analysis of success and survival of dental implants in a large cohort of patients treated in a periodontal clinic. Though the results are promising and encouraging in terms of survival and bone loss, it is important to emphasize the potential risk factors and consider them prior to dental implant placement.

ACKNOWLEDGMENTS
The authors declare there are no competing interest for the above manuscript. This study was carried out independent of financial support with the exception of financial assistance from Institut Straumann AG, Basel, Switzerland for statistical analysis performed at the department of Statistics and Operations Research, Tel-Aviv University, Israel. All authors have made substantial contributions to conception and design of the study. DF and RO has been involved in data collection and data analysis. All authors have been involved in data interpretation, drafting the manuscript and revising it critically and have given final approval of the version to be published.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.