Bone dimensional variations at implants placed in fresh extraction sockets: a multilevel multivariate analysis

Transcription

1 Cristiano Tomasi Mariano Sanz Denis Cecchinato Bjarni Pjetursson Jorge Ferrus Niklaus P. Lang Jan Lindhe Bone dimensional variations at implants placed in fresh extraction sockets: a multilevel multivariate analysis Authors affiliations: Cristiano Tomasi, Jan Lindhe, Departmentof Periodontology, University of Gothenburg, Gothenburg, Sweden Mariano Sanz, Jorge Ferrus, Department of Periodontology, University Complutense of Madrid, Madrid, Spain Denis Cecchinato, Institute Franci, Padova, Italy Bjarni Pjetursson, Department of Periodontology, University of Iceland, Iceland Niklaus P. Lang, Department of Periodontology, Prince Philip Dental Hospital, University of Hong Kong, Hong Kong, SAR PR China Correspondence to: Cristiano Tomasi Department of Periodontology Institute of Odontology The Sahlgrenska Academy at University of Gothenburg Box 450 SE Göteborg Sweden Tel.: þ Fax: þ cristiano.tomasi@odontologi.gu.se Key words: clinical trial, extraction socket, immediate implants, multilevel model analysis, ridge dimension, statistics Abstract Aim: To use multilevel, multivariate models to analyze factors that may affect bone alterations during healing after an implant immediately placed into an extraction socket. Material and methods: Data included in the current analysis were obtained from a clinical trial in which a series of measurements were performed to characterize the extraction site immediately after implant installation and at re-entry 4 months later. A regression multilevel, multivariate model was built to analyze factors affecting the following variables: (i) the distance between the implant surface and the outer bony crest (S-OC), (ii) the horizontal residual gap (S-IC), (iii) the vertical residual gap (R-D) and (iv) the vertical position of the bone crest opposite the implant (R-C). Results: It was demonstrated that (i) the S-OC change was significantly affected by the thickness of the bone crest; (ii) the size of the residual gap was dependent of the size of the initial gap and the thickness of the bone crest; and (iii) the reduction of the buccal vertical gap was dependent on the age of the subject. Moreover, the position of the implant opposite the alveolar crest of the buccal ridge and its bucco-lingual implant position influenced the amount of buccal crest resorption. Conclusions: Clinicians must consider the thickness of the buccal bony wall in the extraction site and the vertical as well as the horizontal positioning of the implant in the socket, because these factors will influence hard tissue changes during healing. Date: Accepted 24 August 2009 To cite this article: Tomasi C, Sanz M, Cecchinato D, Pjetursson B, Ferrus J, Lang NP, Lindhe J. Bone dimensional variations at implants placed in fresh extraction sockets: a multilevel multivariate analysis. Clin. Oral Impl. Res. 21, 2010; doi: /j x The hard tissue dimensional changes occurring following tooth extraction and implant placement immediately into the extraction sockets (type I according to Hammerle et al. 2004) were recently described by Sanz et al. (2009) and Ferrus et al. (2009). The bone dimensional variations that occurred during healing following implant installation were studied by linear measurements performed from predetermined landmarks at the time of surgery and at re-entry operations after 4 months. The analysis of the data documented that the use of a tapered implant in comparison with a cylindrical implant did not provide benefits with respect to bone preservation. It was also observed, however, that the thickness of the buccal and palatal bony walls, as well as the dimensions of the void (defect) present between the implant and the bony walls of the socket significantly influenced bone fill in the horizontal and vertical components of these defects, as well as the resorption that occurred at the socket walls. The linear measurements used obviously depicted two-dimensional changes, which were translated to represent 30 c 2009 John Wiley & Sons A/S

2 three-dimensional alterations. Even though a correlation between each linear measurement representing different aspects of the same defect may be anticipated and an ordinary multivariate regression analysis could reveal factors of importance for bone alterations at implants immediately inserted into extraction sockets, this would not allow the modeling of the correlations between each outcome variable. The application of multilevel techniques to the multivariate structure allows to draw conclusions on the correlations between the dependent variables and to test the effect of explanatory variables on specific dependent variables and the specific effect for the single dependent variable with a smaller standard (Snijders & Bosker 1999). The aim of this analysis was to use multivariate multilevel models to further analyze the factors that may affect tissue alterations occurring at the buccal and palatal aspects of the bony crest during healing after the immediate placement of an implant into an extraction socket. Material and methods The data included in the current analysis were obtained from a randomizedcontrolled clinical trial (RCT designed as a parallel group multicenter study) documenting the tissue response to the immediate placement of two implants with different geometry (Fixture MicroThreadt OsseoSpeedt, Astra Tech AB, Mölndal, Sweden) into the extraction sites in the anterior and premolar regions of the maxilla (Ferrus et al. 2009; Sanz et al. 2009). Four centers, the Department of Periodontology and Fixed Prostodontics, University of Berne, Switzerland, the Department of Periodontology, Universidad de Complutense, Madrid, Spain, the Department of Periodontology, Institute of Odontology at Sahlgrenska Academy, University of Gothenburg, Sweden and the Institute Franci, Padova, Italy were involved. Details regarding the clinical protocol and overall outcomes were reported previously (Sanz et al. 2009). Study sample Briefly, 93 healthy adult subjects (18 years of age) in need of single implants replacing teeth to be removed in the maxilla within the region of were included. Subjects were recruited after having completed treatment for periodontal disease and caries and following signing of informed consent. Study procedures An independent randomization schedule was generated for each center in blocks and designed to ensure balanced distribution of treatments (cylindrical vs. tapered implant). In each center, a randomized treatment code was available in closed envelopes. After tooth extraction and immediate implant placement, a gap always occurred between the implant and the hard tissue walls of the extraction socket. This gap or defect could be present at all aspects of the implant site. In order to describe the size of the defect around the implant, the following landmarks were defined at the buccal and palatal aspects of the extraction site (Fig. 1): Surface of implant (S). Rim, the shoulder of the implant (R). Fig. 1. Landmarks used to describe the dimensions of the crestal ridge as well as the size of the gap between the implant and the socket walls. S, surface of implant; R, rim (shoulder) of implant; C, top of the bony crest; OC, outer border of the bony crest, 1mm apical of C; IC, inner border of the bony crest, 1mm apical of C; D, base of the (void) defect. Top of the bony crest (C). Outer border of the bony crest (OC), 1 mm apical of C. Inner border of the bony crest (IC), 1 mm apical of C. Apical base of the defect (D). Immediately after implant installation, the following linear dimensions were assessed at both the buccal and palatal aspects of the experimental site: S to IC: the width of the gap between the implant and the inner bone crest. S to OC: the distance between the implant and the outer surface of the bone crest. R to D: the depth of the vertical defect. R to C: the vertical distance between R and the top of the bone crest (C). This value could be assessed with a positive or negative sign depending on whether or not R was located apically (positive) or coronally (negative) of C. T: The thickness of the bony walls measured 1 mm apical of the top of the alveolar crest. Following the installation procedure, the soft tissues were replaced and sutured to obtain a tight seal. Seven days after implant placement, the sutures were removed. No implant-supported temporary restorations were used during the first 4 months. At 16 weeks after immediate implant placement, the patients returned for the reentry procedure. The healing abutment was removed and full thickness flaps were elevated. The measurements performed at the time of implant installation were repeated. Statistical analysis A regression multivariate, multilevel model was built to analyze alterations that had occurred between baseline and 16 weeks on the buccal and on the palatal aspect of the implants. A statistical package specifically designed for multilevel modeling (MLwiN 2.11 r, Center for Multilevel ling, University of Bristol, Bristol, UK) was applied. Out of 93, one single implant was lost in one patient. This patient was excluded from the analysis and therefore the multivariate model included 92 patients. c 2009 John Wiley & Sons A/S 31 Clin. Oral Impl. Res. 21, 2010 / 30 36

3 In the model applied, the subject represented the higher level and the outcomes the lower level of analysis. The outcome variables that were included in the model were: the horizontal residual gap (S-IC), the vertical residual gap (R-D), the S-OC distance at the time of reentry, the vertical position of the bone crest opposite the implant surface (R-C) at the time of re-entry. Moreover, in the build-up of the model, the baseline values of each outcome variable were first included, and subsequently all the predictors (bone wall thickness, age, gender, implant shape, position within the tooth arch, smoking, buccal palatal position of the implant in the alveolar socket (S-OC palatal S-OC buccal) and reason for tooth extraction were tested (Table 1). Finally, a parsimonious model including the predictors that had a statistically significant impact (Po0.05) on one or more of the outcomes was named as the Final. The coefficients were estimated using iterative generalized least squares and the significance of each covariate was tested using a Wald test. Nested models were tested for significant improvements in model fit by comparing the reduction in 2 log likelihood with a w 2 distribution (Snijders & Bosker 1999). Results Table 1. Predictors tested in the multivariate multilevel models Independent variables Buccal Palatal Initial horizontal gap (S-IC) Initial horizontal position (S-OC) Initial vertical gap (R-D) Initial vertical position (R-C) 2to5 5to5 Crest thickness 651 mm/2741mm 551 mm/3741 mm Age Gender 48 male/44 female Implant shape 45 test (cylindrical)/47 control (tapered) Tooth arch position 38 anterior/54 premolar Smoking during healing 30 yes/62 no Position B-L (to crest center) (palatal) 4to þ 2 (buccal) Reason for extraction 16 periodontal/76 other Table 2. Multivariate multilevel model on buccal bone alterations Buccal Empty Buccal aspect In Table 2 the multivariate models (Empty and Final ) for alterations on the buccal aspect of the implants are depicted. First the Empty without covariates was estimated. This provided predicted mean values for S-IC (residual horizontal gap) of 0.63 mm, for S-OC of 1.95 mm (residual crest dimension), for R- D of 2.14 mm (residual vertical gap) and for R-C of 0.58 mm (crestal height at time of re-entry). As a confirmation of the model fitting to the data, it may be noted that predicted mean values corresponded to the mean values calculated on a patient level analysis (Sanz et al. 2009). The Final documented that the value of each outcome variable representing 16 weeks of healing was significantly correlated with the corresponding baseline values. Thus, the greater the baseline values the greater were the values at the time of re-entry (Table 2). For example, every 1 mm of increment in the baseline measurement describing the horizontal gap, would result in an increase of 0.22 mm of the value of the predicted residual gap at the time of re-entry. In the model the buccal bone wall was divided into two categories related to the dimension 1 mm apical of the alveolar crest: (a) width 41 mm (thick) and (b) Final Fixed part S-IC Intercept Initial S-IC Po0 Crestal thickness 41mm Po0.05 Smoking Po0.05 Fixed part S-OC Intercept Initial S-OC Po0 Crestal thickness 41 mm Po0.05 Fixed part R-D Intercept Initial R-D Po0.05 Age Po0.05 Fixed part R-C Intercept Initial R-C Po0.05 Crestal thickness 41 mm Po0.05 Periodontal Po0.05 Implant B-L position Po0.05 Random part Variance S-IC Covariance S-IC/S-OC Variance S-OC Covariance R-D/S-IC Covariance R-D/S-OC Variance R-D Covariance R-C/S-IC Covariance R-C/S-OC Covariance R-C/R-D Variance R-C log likelihood mm (thin). At sites with a thick wall compared with sites with a thin wall: (i) there was a smaller horizontal gap (difference 0.25 mm, Po0.05) at the time of reentry, (ii) there was a smaller decrease of the S-OC dimension (difference 0.45 mm; Po0.05), (iii) a smaller reduction of the height of the bony crest (difference 0.59 mm; Po0.05). Furthermore, the model documented that (1) the residual mean horizontal gap remained larger (0.25 mm; Po0.05) if the patient was a smoker, (2) an older patient presented a deeper residual vertical gap 32 Clin. Oral Impl. Res. 21, 2010 / c 2009 John Wiley & Sons A/S

4 Table 3. Correlation between the final values of covariates (buccal) S-IC S-OC R-D S-OC 0.43 n R-D 0.61 n 0.25 n R-C n n Significant correlation Po0.05. (0.03/year; Po0.05), (3) the amount of vertical resorption of the bony crest opposite the rim of the implant was influenced by the reason for tooth extraction (periodontal sites showing less change; 0.78 mm; Po0.05), (4) the amount of vertical crest resorption was influenced by the horizontal positioning of the implant. Thus, every increase of 1 mm in buccal direction from the center of the alveolus, resulted in an apical displacement of the buccal bony crest of 0.22 mm (Po0.05). The model including all the covariates reduced the unexplained variance between subjects for the S-IC (by 21%) at the time of re-entry, for the S-OC (by 39%) at the time of re-entry, for R-D (by 6%) and for R-C (by 19%). The analysis of the covariances revealed a significant positive correlation between final S-IC and S-OC and R-D at the time of re-entry (Table 3). In other words the greater the horizontal gap at the time of reentry, the larger was the residual S-OC distance and the residual depth of the vertical defect. A negative correlation was detected between residual R-D and the residual R-C. Palatal aspect In Table 4 the multivariate models (Empty and Final ) describing alterations on the palatal aspects of the implant sites are depicted. First the Empty without covariates was estimated, giving a predicted mean value for S-IC of 0.45 mm, for S- OC of 1.79 mm, for R-D of 1.14 mm and for R-C of 0.38 mm. The Final demonstrated that the value of each outcome variable representing 16 weeks were significantly correlated with the corresponding baseline value. The presence of a thick palatal wall (T) in comparison to a thin wall had a significant impact on the horizontal gap at the time of re-entry, with a reduction of 0.2 mm (Po0.05). Table 4. Multivariate multilevel model on palatal bone alterations Palatal Empty The age of the patient had a significant impact on the reduction of the palatal bone wall; with increasing age the value S-OC at the time of re-entry was lowered by 0.02/year (Po0.05). The model including all the covariates reduced the unexplained variance between subjects for the S-IC at the time of re-entry (by 17%), for S-OC at the time of re-entry (by 37%), for R-D (by 12%) and for R-C (by 16%). The analysis of the covariances revealed (Table 5) a significant positive correlation between the S-IC at the time of re-entry and the S-OC at the time of re-entry and R-D. predictions (buccal and/or palatal aspects) The models were used to draw predictions in a graphical way for a better understanding of the effects of the covariates on the measured outcomes. In Fig. 2, the predicted values of the residual horizontal gap are related to the initial value of S-IC for the buccal and the palatal sides and for thick and thin bony crests. The linear relationship between initial and residual gap was similar for both aspects of the implant. In Fig. 3, the linear relationship between the initial and the S-OC at the Final time of re-entry is demonstrated; the dotted line represents an initially thin bone crest. The predicted vertical residual gap for both sides is presented in Fig. 4. It was demonstrated that an initial vertical gap of 7 mm resulted in a residual gap of around 1.5 mm at the palatal and 2 mm at the buccal aspects of the implants. Finally, a graph representing the predicted resorption of the bony crest opposite the rim of the implant is presented in Fig. 5. It was demonstrated that the closer to the buccal aspects of the extraction socket the implant was placed, the more advanced was the non-bone covered portion of the buccal portion of the implant at the time of re-entry. Discussion Fixed part S-IC Intercept Initial S-IC Po0.01 Crestal thickness 41mm Po0.05 Fixed part S-OC Intercept Initial S-OC Po0 Age Po0.01 Fixed part R-D Intercept Initial R-D Po0 Fixed part R-C Intercept Initial R-C Po0 Random part Variance S-IC Covariance S-IC/S-OC Variance S-OC Covariance R-D/S-IC Covariance R-D/S-OC Variance R-D Covariance R-C/S-IC Covariance R-C/S-OC Covariance R-C/R-D Variance R-C log likelihood Table 5. Correlation between the final values of covariates (palatal) S-IC S-OC R-D S-OC 0.26 n R-D 0.59 n 0.34 n R-C n Significant correlation Po0.05. The present investigation demonstrated that a multivariate multilevel analysis c 2009 John Wiley & Sons A/S 33 Clin. Oral Impl. Res. 21, 2010 / 30 36

5 The current analysis provided data documenting that Fig. 2. Predicted horizontal gap dimension at the time of re-entry in relation to initial gap for buccal (blue) and palatal (red) sites according to initial bone thickness. The outcomes were consistently dependent on the baseline characteristics. As an example, the residual horizontal and vertical depth of the void at the time of re-entry presented linear relations with the baseline values. The amount of hard tissue fill was dependent on the age of the patient, whether she/he was a smoker and on the thickness of the bony walls and the degree of crestal resorption. In turn, the degree of crestal resorption was found to be dependent on the thickness of the buccal/palatal bony walls as well as the reason for tooth extraction. Fig. 3. Predicted buccal horizontal crest position at the time of re-entry in relation to surface of the implant. Fig. 4. Predicted residual vertical gap at the time of re-entry. could provide information regarding various aspects of hard tissue healing following immediate implant installation into extraction sockets. Moreover, the findings of the analysis may be translated and used by the clinician in the decision-making process regarding the immediate positioning of implants into the socket. Position of the implant in the socket In a previous clinical study involving implants immediately placed into extraction sites (Botticelli et al. 2004, 2008), no information was provided regarding the positioning of the implant into the socket. The authors reported that, during 4 months of healing following implant installation, marked alterations of the edentulous sites occurred. Thus, 450% of the buccal dimensions and close to 30% of the palatal/ lingual dimensions were lost at the implant sites. In animal studies, it was observed that implants immediately placed into the center of extraction sockets could not counteract such dimensional changes and hence, the buccal aspect of the implant during tissue modeling became exposed to the mucosa (Araujo & Lindhe 2005). Similar observations were reported in a clinical study (Evans & Chen 2008). When the implant shoulder was located in a buccal position in the fresh extraction socket, three times as much tissue recession occurred as when the implant had been placed in a lingual position. In another animal experiment (Araujo et al. 2006), it was demonstrated that by placing the implants in a lingual position into the extraction socket, the large buccal gap that occurred between the titanium device and the bone wall during healing became filled with bone. The implant thus became fully integrated into bone. Similar results were recently demonstrated in another study in which the immediate positioning of the 34 Clin. Oral Impl. Res. 21, 2010 / c 2009 John Wiley & Sons A/S

6 Fig. 5. Predicted buccal bone dehiscence at the time of re-entry in relation to initial vertical and horizontal positioning. implant into extraction sockets approximately 1 mm deeper than the level of the buccal alveolar crest and in a lingual position in relation to the center of the alveolus was able to reduce or eliminate the exposure of the implant above the alveolar crest (Caneva et al. 2010). In essence, the position of the implant placed immediately into the extraction socket appears to be of outmost importance for treatment outcomes. This conclusion is clearly validated by the clinical data of the present study. Thus, the multivariate multilevel analysis showed that the further to the palatal aspect of the alveolus the implant was placed, the less implant exposure had occurred at the buccal aspect after 4 months of healing. Furthermore, the current data also documented that more apically positioned implants suffered less implant exposure at buccal aspects than did implants with a shoulder (rim) more closely positioned to the alveolar crest. It must be emphasized, that this conclusion is valid irrespective of other potential influencing factors included in the model (e.g. thickness of the remaining bone walls, age of the patient, smoking and reason for tooth extraction). Thickness of the buccal/palatal bone crest The present multilevel analysis showed that the thickness of the buccal/palatal bony crest markedly influenced the bone fill that occurred in the void (defect) between the implant surface and the socket walls. Thus, at sites with thick bony walls (41 mm), there was more bone fill than at sites with a thin alveolar crest ( 1 mm). Moreover, it was observed that the amount of bone fill on the buccal as well as on the palatal aspects was similar and dependent on the thickness of the alveolar crest. In this context, it is important to realize that the term bone fill was based on the S-IC and R-D measurements. In other words, two-dimensional parameters were meant to illustrate a three-dimensional hard tissue change. Furthermore, it must be understood that one of the tissue landmarks (IC; according to the study protocol; Sanz et al. 2009) was dependent on the location of the alveolar crest (C). Because the alveolar crest during healing was displaced in an apical direction, the IC also became displaced. Hence, the location of S-IC at baseline was not identical to S-IC at 4 months. The statistical model, however, took into account such deficiencies in the measurements by modeling the covariances between the variables and illustrated the importance of the thickness of the alveolar crest for the amount of bone fill. Patient-related factors The multilevel model identified age and smoking as patient-related factors of importance. Thus, the vertical gap fill (RD) was less pronounced in older than in younger subjects and S-IC change was smaller in smokers than non-smokers. The observed effect of aging on bone healing is in agreement with findings on bone fracture healing (Doll et al. 2002). Studies in mice have demonstrated that age affects vascularization during fracture repair and is associated with a decreased rate of chondrogenesis, decreased bone formation, reduced callus vascularization, delayed remodeling, and altered expression of genes involved in repair and remodeling (Quarto et al. 1995; Lu et al. 2005, 2008; Naik et al. 2009). The observation that new formation of bone in the void (defects) was less pronounced in smokers than in non-smokers is in agreement with findings in the periodontal literature. As an example, GTR procedures consistently provided less clinical attachment gain and less bone fill in angular bony defects in smokers than in non-smokers (Trombelli et al. 1997; Cortellini & Tonetti 2004; Stavropoulos et al. 2004). Furthermore, it has been demonstrated that smoking may affect the bone remodeling process after tooth extraction, leading to a significant dimensional reduction of the residual alveolar ridge and a postponed post-extraction socket healing compared with non-smokers (Saldanha et al. 2006). In vivo and in vitro studies have shown that nicotine may inhibit revascularization (Daftari et al. 1994), have a negative impact on bone healing (Hollinger et al. 1999) and inhibit expression of a wide range of cytokines including those associated with neo-vascularization and osteoblast differentiation (Theiss et al. 2000). In conclusion, when considering the immediate installation of implants into the extraction socket, clinicians should consider the thickness of the buccal bony walls in the extraction sites and the vertical as well as the horizontal positioning of the implants into the sockets, as these factors will influence hard tissue changes during healing. In addition, age and smoking habit of the patient should also be considered. Acknowledgements: The study has been supported by research grant from AstraTech AS, Mölndal, Sweden. The authors want to acknowledge the diligent support regarding study monitoring and data management provided by Ann-Sofie Andresson and Frederick Cedar at the AstraTech. c 2009 John Wiley & Sons A/S 35 Clin. Oral Impl. Res. 21, 2010 / 30 36

7 References Araujo, M.G. & Lindhe, J. (2005) Dimensional ridge alterations following tooth extraction. An experimental study in the dog. Journal of Clinical Periodontology 32: Araujo, M.G., Wennstrom, J.L. & Lindhe, J. (2006) ing of the buccal and lingual bone walls of fresh extraction sites following implant installation. Clinical Oral Implants Research 17: Botticelli, D., Berglundh, T. & Lindhe, J. (2004) Hard-tissue alterations following immediate implant placement in extraction sites. Journal of Clinical Periodontology 31: Botticelli, D., Renzi, A., Lindhe, J. & Berglundh, T. (2008) Implants in fresh extraction sockets: a prospective 5-year follow-up clinical study. Clinical Oral Implants Research 19: Caneva, M., Salata, L.A., Scombatti de Souza, S., Baffone, G., Lang, N.P. & Botticelli, D. (2010) Influence of implant positioning in extraction sockets on osseointegration. Histomorphometric analyses in dogs. Clinical Oral Implants Research 21: Cortellini, P. & Tonetti, M.S. (2004) Long-term tooth survival following regenerative treatment of intrabony defects. Journal of Periodontology 75: Daftari, T.K., Whitesides, T.E. Jr., Heller, J.G., Goodrich, A.C., McCarey, B.E. & Hutton, W.C. (1994) Nicotine on the revascularization of bone graft. An experimental study in rabbits. Spine 19: Doll, B.A., Tegtmeier, F., Koch, H., Acarturk, O. & Hollinger, J.O. (2002) Evidence for a cellular andmolecular decline in bone healing with age. Operative Techniques in Orthopaedics 12: Evans, C.D. & Chen, S.T. (2008) Esthetic outcomes of immediate implant placements. Clinical Oral Implants Research 19: Ferrus, J., Cecchinato, D., Pjetursson, B., Lang, N., Sanz, M. & Lindhe, J. (2009) Factors influencing ridge alterations following implant placement in the fresh extraction socket. Clinical Oral Implants Research, doi: /j x [epub ahead of print]. Hammerle, C.H., Chen, S.T. & Wilson, T.G. Jr. (2004) Consensus statements and recommended clinical procedures regarding the placement of implants in extraction sockets. The International Journal of Oral & Maxillofacial Implants (Suppl.): Hollinger, J.O., Schmitt, J.M., Hwang, K., Soleymani, P. & Buck, D. (1999) Impact of nicotine on bone healing. Journal of Biomedical Materials Research 45: Lu, C., Hansen, E., Sapozhnikova, A., Hu, D., Miclau, T. & Marcucio, R.S. (2008) Effect of age on vascularization during fracture repair. Journal of Orthopaedic Research 26: Lu, C., Miclau, T., Hu, D., Hansen, E., Tsui, K., Puttlitz, C. & Marcucio, R.S. (2005) Cellular basis for age-related changes in fracture repair. Journal of Orthopaedic Research 23: Naik, A.A., Xie, C., Zuscik, M.J., Kingsley, P., Schwarz, E.M., Awad, H., Guldberg, R., Drissi, H., Puzas, J.E., Boyce, B., Zhang, X. & O Keefe, R.J. (2009) Reduced cox-2 expression in aged mice is associated with impaired fracture healing. Journal of Bone and Mineral Research 24: Quarto, R., Thomas, D. & Liang, C.T. (1995) Bone progenitor cell deficits and the age-associated decline in bone repair capacity. Calcified Tissue International 56: Saldanha, J.B., Casati, M.Z., Neto, F.H., Sallum, E.A. & Nociti, J.F.H. (2006) Smoking may affect the alveolar process dimensions and radiographic bone density in maxillary extraction sites: a prospective study in humans. Journal of Oral and Maxillofacial Surgery 64: Sanz, M., Cecchinato, D., Ferrus, J., Pjetursson, B., Lang, N. & Lindhe, J. (2009) A prospective, randomized, controlled study to evaluate bone preservation using implants with different geometry placed in fresh extraction sockets in the maxilla. Clinical Oral Implants Research, doi: /j x [epub ahead of print]. Snijders, T.A.B. & Bosker, R.J. (1999) Multilevel Analysis: An Introduction to Basic and Advanced Multilevel ing. Thousand Oaks, CA: Sage. Stavropoulos, A., Mardas, N., Herrero, F. & Karring, T. (2004) Smoking affects the outcome of guided tissue regeneration with bioresorbable membranes: a retrospective analysis of intrabony defects. Journal of Clinical Periodontology 31: Theiss, S.M., Boden, S.D., Hair, G., Titus, L., Morone, M.A. & Ugbo, J. (2000) The effect of nicotine on gene expression during spine fusion. Spine 25: Trombelli, L., Kim, C.K., Zimmerman, G.J. & Wikesjo, U.M. (1997) Retrospective analysis of factors related to clinical outcome of guided tissue regeneration procedures in intrabony defects. Journal of Clinical Periodontology 24: Clin. Oral Impl. Res. 21, 2010 / c 2009 John Wiley & Sons A/S