Since their introduction, the use of orthodontic

ORIGINAL ARTICLE Survival analysis of orthodontic mini-implants Shin-Jae Lee, a Sug-Joon Ahn, a Jae Won Lee, b Seong-Hun Kim, c and Tae-Woo Kim d Seoul and Gyeonggi-Do, Korea Introduction: Survival analysis is useful in clinical research because it focuses on comparing the survival distributions and the identification of risk factors. Our aim in this study was to investigate the survival characteristics and risk factors of orthodontic mini-implants with survival analyses. Methods: One hundred forty-one orthodontic patients (treated from October 1, 2000, to November 29, 2007) were included in this survival study. A total of 260 orthodontic mini-implants that had sandblasted (large grit) and acid-etched screw parts were placed between the maxillary second premolar and the first molar. Failures of the implants were recorded as event data, whereas implants that were removed because treatment ended and those that were not removed during the study period were recorded as censored data. A nonparametric life table method was used to visualize the hazard function, and Kaplan-Meier survival curves were generated to identify the variables associated with implant failure. Prognostic variables associated with implant failure were identified with the Cox proportional hazard model. Results: Of the 260 implants, 22 failed. The hazard function for implant failure showed that the risk is highest immediately after placement. The survival function showed that the median survival time of orthodontic mini-implants is sufficient for relatively long orthodontic treatments. The Cox proportional hazard model identified that increasing age is a decisive factor for implant survival. Conclusions: The decreasing pattern of the hazard function suggested gradual osseointegration of orthodontic mini-implants. When implants are placed in a young patient, special caution is needed to lessen the increased probability of failure, especially immediately after placement. (Am J Orthod Dentofacial Orthop 2010;137:194-9) Since their introduction, the use of orthodontic mini-implants or screws has become widespread and has gained in popularity. 1 The abundance of recent studies is evidence of this popularity, 2 and, more favorably, patients seem to have high acceptance and satisfaction with this relatively new treatment modality. 3 When researching orthodontic implants, clinicians are frequently concerned about the ultimate success rate and the risk factors of failure. Survival analysis is the most suitable method to analyze data that have as a principal end point the time until an event occurs; it is a popular method in biomedical research. Survival time can be defined broadly as the time until an event. 4,5 The event in this study was implant failure. Although some prosthetic dental implant a Associate professor, Department of Orthodontics, School of Dentistry and Dental Research Institute, Seoul National University, Seoul, Korea. b Professor, Department of Statistics, Korea University, Seoul, Korea. c Assistant professor, Department of Orthodontics, The Catholic University of Korea, Uijongbu St. Mary s Hospital, Gyeonggi-Do, Korea. d Professor, Department of Orthodontics, School of Dentistry and Dental Research Institute, Seoul National University, Seoul, Korea. One author, Seong-Hun Kim, uses the implant mentioned in this article. However, all authors claim to have no financial interest in any company or any products mentioned in this article. Reprint requests to: Sug-Joon Ahn, Dental Research Institute and Department of Orthodontics, School of Dentistry, Seoul National University, 28-22 Yeongeon- Dong, Jongro-Gu, Seoul 110-768, Korea (ROK); e-mail, titoo@snu.ac.kr. Submitted, January 2008; revised and accepted, March 2008. 0889-5406/$36.00 Copyright Ó 2010 by the American Association of Orthodontists. doi:10.1016/j.ajodo.2008.03.031 studies used survival analysis, 6-8 studies of orthodontic mini-implants simply reported success or failure rates. This is less meaningful than a survival analysis. For an orthodontic-implant success-rate study, survival analysis seems to have several advantages and even be superior to other methods. First, the most important feature of survival analysis is that some subjects in the study have not experienced implant failure at the end of the study or the time of analysis. Because of time limitations, a researcher often cannot wait for all events in all subjects. These situations are called censored observations. Second, some subjects are lost to follow-up, drop out, or experience failure because of an unrelated cause. Third, we usually cannot determine the exact time of implant failure but can notice that the failure happened only at discrete times ie, at the appointment or the patient s visit. Survival analysis solves these problems irrespective of the original distribution of data, produces valuable information, including hazard characteristics and survival rates, and identifies risk factors. Thus, survival analysis modeling seemed to be a valuable method for orthodontic mini-implant research, and our aim was to investigate orthodontic implant failure characteristics and risk factors with survival analyses. MATERIAL AND METHODS One hundred forty-one orthodontic patients (treated from October 1, 2000, to November 29, 2007) were 194

American Journal of Orthodontics and Dentofacial Orthopedics Lee et al 195 Volume 137, Number 2 included. Their mean age was 27 years (range, 12-51 years). All patients were treated with fixed appliance and had their 4 first premolars extracted. A total of 260 orthodontic mini-implants (C-implant, Dentium, Seoul, Korea) were placed between the maxillary second premolars and the first molars. The mini-implants were used to retract the anterior segment with a controlled tipping force. To exclude both material and design factors of the implants, only 1 type of implant was used. This implant had a sandblasted (large grit) and an acid-etched screw part. The screw part was 1.8 mm in diameter and 8.5 mm long. The head part (upper 2 mm) was designed for orthodontic use. Detailed specifications can be found elsewhere. 14 Five clinical variables (sex, age, implantation side [right or left], oral hygiene, and clinician) were investigated, and survival time was coded by weeks after implant placement. Placement of the mini-implants was performed by 2 right-handed clinicians; one, who had been in clinical orthodontic practice over 30 years, placed 174 mini-implants, and the other (S.H.K.), who had 10 years of experience, placed 86 mini-implants. The healing period after placement was 4 weeks. Oral hygiene status was determined by reviewing orthodontic records and intraoral photos subjectively. Failures of implants were coded as events data. Implants that were removed because the treatment ended and those that were not removed during the fixed period of this investigation were coded as censored data. Discrete failure time was recorded in weeks. Statistical analysis The Fisher exact test significance and odds ratio statistics were calculated. A nonparametric life table method was used to easily visualize the hazard function over time. There were only 2 failure events after 70 weeks (at 113 and 117 weeks), and the hazard function was depicted at a 100-week limit. Kaplan-Meier survival curves were generated, and the Gehan generalized Wilcoxon test was used to identify the variables associated with implant failure. Prognostic variables associated with implant failure were identified with the Cox proportional hazard model by using SPSS software for Windows (version 12.0, SPSS, Chicago, Ill). The level of statistical significance was set at 5%. RESULTS The average participating time of each implant was 74 6 41 weeks and ranged from 1 to 189 weeks. Of the 260 implants, 22 failed, and 238 were censored, 220 of which completed orthodontic treatment. The average orthodontic treatment time was 88 weeks. There was no significant difference in the success rates between implantation sides, clinicians, sex, and oral hygiene. Only the age variable had a significant association with success rate (Table I). When the success rates were evaluated for each clinician, the results were similar for the pooled data in Table I. The hazard function of implant survival time gives the conditional failure rate. Therefore, the hazard function is regarded as the instantaneous failure rate. 5 Late failure events at 70, 113, and 117 weeks were observed. When the time after implant placement was truncated at 100 weeks, the hazard function of implant failure showed that the risk was highest immediately after placement and then decreased to zero. The linear fit of the hazard function showed high quality of fit (R 2 5 0.81) with a negative slope over time, as shown in Figure 1. The survival function is also called the cumulative survival rate. The mean time of permanence for the implant placed was 156.5 weeks, and the median time of permanence far exceeded the mean orthodontic treatment time. The Kaplan-Meier survival curve (Fig 2) showed an extremely high success rate. The Gehan generalized Wilcoxon test showed that, for patients younger than 20 years, the survival rate was significantly lower than for older subgroups (P 5 0.029). The Cox proportional hazard model also showed that increasing age is a decisive factor for implant survival (Table II). The estimated probability of implant failure decreased 0.925 times with each year of the patient s age (P 5 0.016). DISCUSSION After searching published articles and the electronic database concerning orthodontic implants, we found that our survival analysis method was the second study on this topic; another study compared the survival rates of various implant types and many clinical variables for 140 implants. 15 In the first study, the sample size seemed relatively small for the 7 test variables with multi-levels. In addition, the results in parts are not easy to understand. The authors pointed out the location and emergence site as risk factors. 15 However, their confidence intervals included 1.0, and their conclusions were not exactly based on the results. In addition, because the simplistic log-rank test does not determine effectively when the failure is concentrated in an early phase over time, another method should have been considered to detect differences in survival rates.

196 Lee et al American Journal of Orthodontics and Dentofacial Orthopedics February 2010 Table I. Success rate, exact test significance, and odds ratio statistics for orthodontic mini-implants by clinical variables Variable Success n (%) Failure n (%) Total Exact test significance Odds ratio (95% CI) Implantation side 1.000 1.07 (0.45, 2.56) Right 123 (91.8) 11 (8.2) 134 Left 115 (91.3) 11 (8.7) 126 Clinician 1.000 0.94 (0.37, 2.40) 1 159 (91.4) 15 (8.6) 174 2 (S.H.K.) 79 (91.9) 7 (8.1) 86 Sex 0.598 0.76 (0.28, 2.05) Male 53 (89.8) 6 (10.2) 59 Female 185 (92.0) 16 (8.0) 201 Oral hygiene 0.508 1.40 (0.58, 3.36) Good 128 (92.8) 10 (7.2) 138 Bad 110 (90.2) 12 (9.8) 122 Age subgroups (y)* 0.025 0.31 (0.11, 0.85) \20 42 (82.4) 9 (17.6) 51 20-30 121 (93.8) 8 (6.2) 129.30 75 (93.8) 5 (6.4) 80 *Among the 3 subgroups, the odds ratio was calculated between the \20 and the 20-30 age subgroups; 2-sided exact test significance at P \0.05. Fig 1. When the time after implant placement was limited to 100 weeks, the hazard function of implant failure showed that the risk was highest immediately after placement and then decreased to zero. The linear fit of the hazard function showed high quality of fit (R 2 5 0.81). Previous articles that described the success rates and identified risk factors seemed to have relatively few subjects with many research variables. 9,11,12 A small sample size might inevitably be accompanied by an increase in type II errors and decreased statistical power. Furthermore, if there are too many variables, then higher-order interactions would be produced, and interpreting the results would be complicated. 16 A recent study showed no significant correlations between the success rate and the variables of age, sex, mandibular plane angle, anteroposterior jaw-base relationship, control of periodontitis, temporomandibular joint disorder, loading, diameter, and screw length. 9 However, because the sample size was 79, it is questionable whether the test had enough statistical power for those 9 variables. Was there a possibility of type II error? Another article

American Journal of Orthodontics and Dentofacial Orthopedics Lee et al 197 Volume 137, Number 2 Fig 2. The Kaplan-Meier survival function curve by age subgroup showed an extremely high success rate overall. For patients younger than 20 years, the survival rate was significantly lower than the older subgroups. Table II. Results of the Cox proportional hazard modeling and Wald test significance Beta SE Wald Significance Exp (B) 95.0% CI for Exp (B) Age 0.08 0.03 5.825 0.016* 0.925 (0.868, 0.985) Right vs left 0.05 0.43 0.013 0.909 0.952 (0.411, 2.204) Clinician 0.13 0.47 0.075 0.784 0.879 (0.348, 2.218) Sex 0.02 0.49 0.001 0.975 1.015 (0.386, 2.670) Oral hygiene 0.13 0.45 0.087 0.769 0.877 (0.365, 2.106) Wald, Wald statistic; Exp (B), exponential of Beta. *Significant parameter estimates at P \0.05; of the variables tested, age was the only significant one. investigated 227 implants with more than 20 variables. 11 Two hundred twenty-seven implants are not sufficient to find statistically significant interaction effects among over 20 variables. Because the success rate of orthodontic mini-implants is high, and the between-group difference is not large, a greater sample size is required to prove the hypothesis of this study. The resulting success rate was calculated by conventional ratio statistics (Table I) and showed that our rate was higher than 9,10 or similar 11 to previous reports. The proximity of an implant to the root was reported as a major risk factor for implant failure. 10,17 A smaller-diameter screw, peri-implant inflammation, and high mandibular plane angle were suspected as possible factors for failure. 12 Although Park et al 11 reported greater success on the left side because of the dominance of the righthanded patients, who usually have better oral hygiene on the left side of the dental arch, our subjects showed no significant differences between the left and right sides for implant placement. Kuroda et al 9 and Park et al 11 stated that oral hygiene and implant failure have a close relationship, but this was not shown to be true in our study. Our study showed that adolescents younger than 20 years were a high-risk group. Although a long-term follow-up of prosthetic implants in adolescents (ages, 13-17 years) reported favorable results, orthodontic implants have various features, such as smaller sizes and transverse directions of force application. 18 The hazard function linearly decreased over time; this negative slope hazard function by time might be considered evidence of osseointegration. Osseointegration was often measured with the concept of removal torque value. 19,20 The inverse of the hazard function graph (1-hazard function) demonstrates a noninvasive

198 Lee et al American Journal of Orthodontics and Dentofacial Orthopedics February 2010 way to visualize the pattern of osseointegration of the implants. The hazard function pattern in Figure 1 compared favorably with that of previous research on partial osseointegration. 21-23 Although orthodontic miniscrews achieve stationary anchorage primarily through mechanical retention, they are known to be osseointegrated after 3 weeks, increasing the difficulty of removal. The decreasing hazard function justified a clinical recommendation of loading orthodontic force after 3 months with a minimum 3-week healing period. 13 The meta-analysis of Cornelis et al, 24 a systematic review of 8 articles, concluded that healing times ranged from 0 to 12 weeks. Our data also showed late failures after 12 weeks, even at 113 and 117 weeks. The reason for the late failures might be explained by tooth movement over time that resulted in root proximity. The Kaplan-Meier method was selected for generating the survival function because (1) it applies to small, moderate, and large samples; (2) it easily visualizes and completely covers the information for censored subjects, and we had much censored data; and (3) it is popular in biomedical science. 5 The survival function showed that the survival rate of orthodontic mini-implants is sufficient for relatively long orthodontic treatments. It might be related to the decreasing pattern of the hazard function. The median and other percentiles (25th and 75th) of survival time compare the survival distributions of 2 or more groups. However, as shown in Figure 2, when the success rate is excessively high or low, the median survival is impossible to calculate. Therefore, various other approaches for comparing survival functions are available. Among them, the Gehan generalized Wilcoxon test was selected because it gives more weight to early failures than later failures and is more likely to detect early differences in the survival distributions; this was thought to be the most suitable for our data. 5 The test statistics (P 5 0.029) implied the importance of special care for younger patients when placing orthodontic mini-implants. Although age was the only variable associated with failure, further investigations with larger samples might be helpful for determining the influences of malocclusion, tooth movement, and anchorage units for those movements. The success rate of our study was higher than those in previous studies. But this does not mean that the implant system we used was of better quality. This implant system needs a 2-stage surgery procedure and an 4-week healing period (much like a prosthetic implant). However, many orthodontic mini-implants involve a drill-free procedure without the need for flap surgery. Although the success rate might be low, it can actually be more useful. Survival analysis might be a better method of investigating implant failure, instead of reporting only the ratio of successes and failures. CONCLUSIONS As a result of survival analysis, the decreasing pattern of the hazard function suggests gradual osseointegration of orthodontic mini-implants. When implants are planned for a young patient, special caution is needed to prevent greater failure probability, especially immediately after placement. We thank Dr. Kyu-Rhim Chung, DMD, MSD, PhD, for his generous assistance with data collection. We also thank the copyeditor, AJO-DO, and Dr. Richard E. Donatelli, DMD, at the Department of Orthodontics, University of Florida College of Dentistry, for their invaluable assistance to improve this manuscript. REFERENCES 1. Roberts WE, Smith RK, Zilberman Y, Mozsary PG, Smith RS. Osseous adaptation to continuous loading of rigid endosseous implants. Am J Orthod 1984;86:95-111. 2. Huang LH, Shotwell JL, Wang HL. Dental implants for orthodontic anchorage. Am J Orthod Dentofacial Orthop 2005;127:713-22. 3. Gunduz E, Schneider-Del Savio TT, Kucher G, Schneider B, Bantleon HP. Acceptance rate of palatal implants: a questionnaire study. Am J Orthod Dentofacial Orthop 2004;126:623-6. 4. Kalbfleisch JD, Prentice RL. The statistical analysis of failure time data. New York: John Wiley & Sons; 1980. 5. Lee ET, Wang JW. Statistical methods for survival data analysis. 3rd ed. Hoboken, NJ: Wiley; 2003. 6. DeLuca S, Habsha E, Zarb GA. The effect of smoking on osseointegrated dental implants. Part I: implant survival. Int J Prosthodont 2006;19:491-8. 7. Scurria MS, Morgan ZVT, Guckes AD, Li S, Koch G. Prognostic variables associated with implant failure: a retrospective effectiveness study. Int J Oral Maxillofac Implants 1998;13:400-6. 8. Becktor JP, Isaksson S, Sennerby L. Survival analysis of endosseous implants in grafted and nongrafted edentulous maxillae. Int J Oral Maxillofac Implants 2004;19:107-15. 9. Kuroda S, Sugawara Y, Deguchi T, Kyung HM, Takano- Yamamoto T. Clinical use of miniscrew implants as orthodontic anchorage: success rates and postoperative discomfort. Am J Orthod Dentofacial Orthop 2007;131:9-15. 10. Kuroda S, Yamada K, Deguchi T, Hashimoto T, Kyung HM, Takano-Yamamoto T. Root proximity is a major factor for screw failure in orthodontic anchorage. Am J Orthod Dentofacial Orthop 2007;131(Suppl):S68-73. 11. Park HS, Jeong SH, Kwon OW. Factors affecting the clinical success of screw implants used as orthodontic anchorage. Am J Orthod Dentofacial Orthop 2006;130:18-25. 12. Miyawaki S, Koyama I, Inoue M, Mishima K, Sugahara T, Takano-Yamamoto T. Factors associated with the stability of titanium screws placed in the posterior region for orthodontic anchorage. Am J Orthod Dentofacial Orthop 2003;124:373-8. 13. Deguchi T, Takano-Yamamoto T, Kanomi R, Hartsfield JK Jr, Roberts WE, Garetto LP. The use of small titanium screws for orthodontic anchorage. J Dent Res 2003;82:377-81.

American Journal of Orthodontics and Dentofacial Orthopedics Lee et al 199 Volume 137, Number 2 14. Chung K, Kim SH, Kook Y. C-orthodontic microimplant for distalization of mandibular dentition in Class III correction. Angle Orthod 2005;75:119-28. 15. Cheng SJ, Tseng IY, Lee JJ, Kok SH. A prospective study of the risk factors associated with failure of mini-implants used for orthodontic anchorage. Int J Oral Maxillofac Implants 2004;19:100-6. 16. Norman GR, Streiner DL. Biostatistics. The base essentials. St Louis: Mosby; 1996. 17. Liou EJ, Pai BC, Lin JC. Do miniscrews remain stationary under orthodontic forces? Am J Orthod Dentofacial Orthop 2004;126:42-7. 18. Thilander B, Odman J, Lekholm U. Orthodontic aspects of the use of oral implants in adolescents: a 10-year follow-up study. Eur J Orthod 2001;23:715-31. 19. FaveroLG, Pisoni A, Paganelli C. Removal torque of osseointegrated mini-implants: an in vivo evaluation. Eur J Orthod 2007;29:443-8. 20. Song YY, Cha JY, Hwang CJ. Mechanical characteristics of various orthodontic mini-screws in relation to artificial cortical bone thickness. Angle Orthod 2007;77:979-85. 21. Kravitz ND, Kusnoto B. Risks and complications of orthodontic miniscrews. Am J Orthod Dentofacial Orthop 2007;131(Suppl): S43-51. 22. Melsen B, Verna C. Miniscrew implants: the Aarhus anchorage system. Semin Orthod 2005;11:24-31. 23. Vande Vannet B, Sabzevar MM, Wehrbein H, Asscherickx K. Osseointegration of miniscrews: a histomorphometric evaluation. Eur J Orthod 2007;29:437-42. 24. Cornelis MA, Scheffler NR, De Clerck HJ, Tulloch JF, Behets CN. Systematic review of the experimental use of temporary skeletal anchorage devices in orthodontics. Am J Orthod Dentofacial Orthop 2007;131(Suppl):S52-8.