Search methods for identification of studies. We obtained article citations through an electronic search of databases and bibliographic reference listings of published primary and review studies, as well as through a hand search. We contacted authors of relevant studies for additional information and accepted their responses until June 12, 2007.

Electronic searches. An information specialist at the University of Washington Health Sciences Library assisted us in developing a search strategy, which included terms related to types of treatment and outcomes. (For the search strategies used in one electronic database, see Appendix 1 in the supplemental data to the online version of this article at "http://jada.ada.org".) The search strategy did not include search terms relating to the study type because doing so could have excluded almost one-half of the pertinent publications.¹⁴ The electronic databases we used—all for the period from January 1980 through June 2006—were MEDLINE; Web of Science; and the Cochrane Library, including the Cochrane Central Register of Controlled Trials (CENTRAL), the Cochrane Database of Systematic Reviews (CDSR), the Database of Abstracts of Reviews of Effects (DARE) and Health Technology Assessment (HTA).

Electronic searching of unpublished or "gray" literature. We searched ClinicalTrials.gov ("www.clinicaltrials.gov") and the National Research Register (UK) ("www.controlled-trials.com") using the following phrase: orthodontic AND periodontal. We also searched Pro-Quest Dissertation Abstracts and Thesis database ("www.lib.umi.com./dissertations") using the truncated terms "orthodontic*" and "periodont*."

Hand searching. We selected six journals—American Journal of Orthodontics and Dentofacial Orthopedics, Angle Orthodontist, European Journal of Orthodontics, Journal of Periodontology, Journal of Dental Research and Journal of Clinical Periodontology—because they published the majority of reports on orthodontics and periodontal diseases, according to a search on Web of Science. We hand searched these dental journals for studies published between January 1980 and June 2006 that reported on the periodontal effects of orthodontic treatment.

Two reviewers (J.C.-C. and D.W.B.) coded reports identified through hand and electronic searches according to type of study, participants, presence of an orthodontic intervention, presence of a periodontal outcome, presence and type of the comparison group. In addition, they screened the references in selected papers for eligible studies.

Assessment of methodological quality. Two of the reviewers (J.C.-C. and A.-M.B.) independently assessed the quality of the identified studies and resolved any disagreements through discussion.

For RCTs, the reviewers assessed eight items: random allocation; allocation concealment; baseline similarity of groups; reporting of eligibility criteria; point estimates; measure of variability of primary outcome; blinding of outcome assessors, care provider and patient; and intent-to-treat analysis.¹⁵ For cohort studies, case-control studies and cross-sectional studies, they used the Newcastle-Ottawa Quality Assessment Scale, which consists of eight items:

– four items regarding selection—representativeness of exposed cohort, selection of nonexposed groups, ascertainment of exposure and baseline assessment of outcome (not applicable for cross-sectional studies);

– one item regarding comparability of the groups in terms of design or analysis (confounding);

– three items regarding outcome assessment—blinding, follow-up duration, and dropout or response rates.¹⁶

The reviewers included in the quality assessment one additional item not present in the scale: the reporting of number of observations, point estimates and measure of variability for the primary outcome. On the basis of these items, they classified all studies according to risk of bias, whether low (all quality items were met), moderate (one or two quality items not met) or high (three or more criteria not met).

Data extraction and synthesis. One reviewer (J.C.-C.) performed data extraction. She extracted patient mean and standard deviations for the continuous outcomes from the reports or estimated them by pooling the results presented by surfaces or sites (for details on the estimation of means and standard deviations, see Appendix 2 in the supplemental data to the online version of this article).

The reviewer synthesized the results according to study design and type of outcome. She assessed heterogeneity by means of the I-squared statistic. For continuous variables, she calculated weighted mean differences between groups using RevMan 4.2.7 (Plone Foundation, Houston) on the basis of a fixed-effects model. If the statistical heterogeneity levels were higher than 70 percent with fixed and random effects models,¹⁷ she did not present summary estimates.

We planned to perform sensitivity analyses, subgroup analyses (for example, children, adolescents and adults; surrogate and true endpoints; and fixed and removable appliances) and publication type bias analyses¹⁸ as feasible to determine bias and the significance of contributing factors (study quality, diagnostic criteria and severity of the outcome, duration of orthodontic treatment and duration of follow-up) to the overall results. However, there were too few studies categorized by outcome to allow us to assess publication bias and to perform sensitivity and subgroup analyses.

	RESULTS

TOP ABSTRACT MATERIAL AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
Search results. The electronic searches retrieved 3,552 unique citations. After evaluating titles and abstracts (when available) of all citations and the full text of 104 citations, as well as querying primary authors, we determined that 14 studies met the inclusion criteria (for reasons for exclusion, see Appendix 3 in the supplemental data to the online version of this article). We found no studies that met the inclusion criteria through the electronic searching of "gray" literature. Of these 14 studies, we excluded one article because it reported on a study which we already had included and two studies because of their absence of statistics (Figure 1).

View larger version (66K): [in this window] [in a new window]   Figure 1. Flow diagram of the selection of studies for the review of the effects of orthodontic therapy on periodontal health.

We considered agreements between examiners of subsamples of the hand-searched journals and the electronically searched reports to be adequate ( = 66 percent and 48 percent, respectively).

Of the 19 studies retrieved by either the electronic or hand searches, seven were duplicates. We retrieved one study through the hand search only. Thus, we included 12 studies in this review.

Description of included studies. We identified one RCT that lasted 16 weeks (12 weeks of orthodontic treatment and follow-up four weeks later)¹⁹; three prospective cohort studies that followed patients for three, five and 14 years^20–22; and eight cross-sectional studies^23–30 (Table, pages 418–419). (For more detailed descriptions of the included studies and assessment of methodological quality, see Appendixes 4 and 5 in the supplemental data to the online version of this article.)

The participants’ ages ranged from 12 to 47 years old, with the majority of subjects undergoing orthodontic treatment as adolescents or young adults. The orthodontically treated groups had varying forms of malocclusion such as anterior crowding and Class II. Three studies^23,28,29 did not report this group’s malocclusion status. One study restricted the inclusion of participants to those who had anterior crowding on both arches.¹⁹ The malocclusion status of the untreated groups were similar to that of the orthodontically treated group,^19,21,22,30 different from the treated group (untreated comparison groups with no or minor malocclusion)^20,26–28 or not reported.^23–25,29 The types of orthodontic treatments were fixed appliances,^{19,20,24,25,29} fixed appliances with premolar extractions,^26,27,30 removable appliances,²⁹ both fixed and removable appliances,^21,22 or not reported.^23,28

The selected reports identified five surrogate markers of periodontal status: alveolar bone loss or cementoenamel junction–alveolar crest height,^20,24,26 periodontal pocket depth^19,27,29,30 (measured, but not reported in one study²⁹), clinical attachment loss,^22,27,29,30 gingival recession^22,25,27,28 and gingivitis.^{19,21,23,27,30} None of the studies reported on tooth loss, tooth mobility or other adverse effects.

Assessment of methodological quality. Of seven quality items we reviewed for RCTs, the one available RCT¹⁹ satisfied two quality items; it did not report information on four items. Of 10 quality items we reviewed for cohort studies, two studies^20,21 satisfied six and the other one²² satisfied seven. One cohort study selected the nonexposed cohort from a different source (people with mild or no malocclusion) and, although the groups were matched, a conditional analysis was not reported.²⁰ The seven cross-sectional studies satisfied from one to seven quality items of a total of nine quality items we reviewed for cross-sectional studies. Several items were not reported in the cross-sectional studies. The orthodontic groups in most of the cross-sectional studies were not selected randomly or systematically, and the non-treated groups were selected from a different source (people with no or mild malocclusion, dental students or staff). None of the cross-sectional studies adjusted for confounding, and those that reported matching by age or by another characteristic did not report a conditional analysis. We classified all 12 studies as having a high risk of bias (see Appendix 5 in the supplemental data to the online version of this article).

Results of the comparison of orthodontic treatment versus no treatment. This comparison involved 1,670 participants from 12 studies: 821 participants in the intervention groups and 849 in the untreated comparison groups.

RCT. We evaluated two periodontal outcomes in the RCT.

– Pocket depth: Subjects in the orthodontically treated group had, on average, a pocket depth 0.3 millimeter deeper than that of subjects in the untreated groups. Standard deviations were not reported.

– Gingivitis: Teeth that had been orthodontically treated had a similar number of bleeding sites as did teeth that had not been orthodontically treated (20 of 96 sites versus 18 of 96 sites; no within-patient correlation coefficients reported).

Cohort and cross-sectional studies. We evaluated five periodontal outcomes in the cohort and cross-sectional studies.

– Alveolar bone loss (three studies^20,24,26): The mean alveolar bone loss was 0.13 mm (95 percent confidence interval [CI], 0.07–0.20) greater among the groups that had been orthodontically treated than among the groups that had not (Figure 2, page 420).

– Periodontal pocket depth (two studies^26,27): The orthodontically treated group had pocket depths that were, on average, 0.23 mm (95 percent CI, 0.15–0.30) deeper than those of the untreated groups (Figure 2).

– Gingival recession (three studies^22,25,27): The orthodontically treated group had gingival recession 0.03 mm (95 percent CI, 0.01–0.04) greater than did the untreated group (Figure 2).

– Attachment loss (three studies^22,27,30): We did not calculate a summary estimate because of the high level of heterogeneity among the studies (I² = 93 percent). Those who had undergone orthodontic treatment had 0.11 mm (95 percent CI, 0.07–0.15) greater attachment loss in one study²⁷ and 0.05 mm (mean: –0.05; 95 percent CI, –0.11–0.01) and 0.06 (mean: –0.06; 95 percent CI, –0.12–0.00) lower attachment loss in the other studies,^22,30 compared with those who had not received treatment.

– Gingivitis (four studies^21,23,27,30): We calculated no summary estimate of the four studies owing to high levels of heterogeneity among the studies (I² = 97 percent). The gingival index was 0.15 points (95 percent CI, 0.07–0.23) and 0.05 points (95 percent CI, –0.13–0.26) higher in the group that had received orthodontic treatment in two studies, respectively.^27,30 The mean percentages of sites with bleeding on probing were six points (95 percent CI, –10.7- -1.3) and 23.8 points lower in the orthodontically treated group than in the group that had not received orthodontic treatment in two studies, respectively.^21,23

View larger version (74K): [in this window] [in a new window]   Figure 2. Forest plot of summary estimate and individual results of cohort and cross-sectional studies reporting on the effect of orthodontic therapy on periodontal health. Periodontal outcomes (alveolar bone loss, periodontal pocket depth and gingival recession) were measured after treatment only. The period between treatment and outcome measurements are listed on the right. IV, fixed, 95% CI: Weighted mean difference from a fixed-effects inverse variance (IV) model and 95 percent confidence interval (CI). df: Degrees of freedom. I2: Heterogeneity beyond chance. NR: Not reported. z: z score.

	DISCUSSION

TOP ABSTRACT MATERIAL AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

The studies included in the review had a potential for bias. We identified one RCT; however, it was of relatively low quality, and the 16-week trial (12 weeks of treatment and a follow-up four weeks later) cannot be considered a true representation of regular comprehensive treatment. The remaining studies in the review were cohort and cross-sectional studies. While all of these studies included nontreated controls, the majority involved controls with mild or no malocclusions. In addition, the compared groups may have differed in terms of factors such as oral hygiene practices, socioeconomic status or smoking. The investigators had not evaluated such factors at baseline or at follow-up, and in no studies had they performed statistical adjustments to evaluate the effect of potential imbalances in subjects’ characteristics on the study findings. The reported outcomes are posttreatment means only. The lack of information on pretreatment periodontal status precludes assessment of whether the reported differences are the result of the orthodontic treatment or of selection bias. The common lack of reporting regarding the number of participants who were lost to follow-up or for whom data were missing in the longitudinal and cross-sectional studies is a potential for bias.

The available data do not allow us to determine whether these small adverse changes are indicative of large site-specific changes (such as molar bands or extractions) or host-specific factors (such as alteration of oral hygiene habits during orthodontic therapy), an inherent adverse consequence of the orthodontic forces to move teeth or study biases (such as mal-occlusion and socioeconomic differences between people receiving and not receiving orthodontic therapy). The results of one study²² suggest a greater presence of few large periodontal defects versus generalized small changes as a result of the orthodontic treatment. In addition, many of the subjects were treated when they were adolescents or young adults, and the various studies’ relatively short periods of follow-up may not reflect the long-term orthodontic benefit or harm to the periodontium.

Two issues that may elicit debate are our choice of the inclusion/exclusion criteria (untreated comparison group and no fully-banded appliances) and the presentation of summary statistics when perhaps none should have been presented. In addition, a single reviewer extracted the data, and this may have been a potential source of bias. We established the inclusion criteria to identify the best available literature and to restrict the included orthodontic treatments to contemporary orthodontic methods. Therefore, we included only studies that had an untreated comparison group and did not employ fully banded appliances. Untreated controls are important to account for naturally occurring periodontal changes. However, patients in the untreated groups from the studies we included often had minor malocclusions, which may have biased the results.

The periodontal effects of banded appliances may differ from those of bonded appliances, with banding being associated with increased inflammation and loss of attachment when compared with bonding.³⁴ Although clinicians no longer place bands on anterior teeth, placing bands on molars still is a common procedure, and an evaluation of the findings from studies regarding the effects of fully banded treatment on molars may provide some valuable information. We excluded from this systematic review four studies that were reported in seven publications because they reported on fully banded orthodontic treatment.^35–41 Investigators in two studies observed higher levels of periodontal diseases on the distal surface of molars^35–37 and in the maxillary posterior areas^38,39 of orthodontic patients, while investigators in the other two studies reported conflicting results for gingivitis.^40,41 The findings of these four studies of fully banded orthodontic treatment, as well as from two reports published before 1980,^42,43 suggest that patients who have had orthodontic treatment have a slightly worse periodontal status than that of untreated controls when assessed over the long term, a finding similar to that of our review.

Strengths of this systematic review include the comprehensive literature search, including gray literature and studies published in languages other than English, and the assessment of the scientific quality of the included studies.

	CONCLUSION

TOP ABSTRACT MATERIAL AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
This comprehensive search indicates that there is an absence of reliable evidence on the positive effects of orthodontic therapy on periodontal status. Epidemiologic studies with adequate comparison groups and follow-up time are lacking, as are studies that focus on adult populations or specific techniques such as current orthodontic expansion methods. The evidence that does exist suggests a small mean worsening of periodontal status after orthodontic therapy. Claims that orthodontic therapy results in overall improved periodontal health cannot be supported with the existing controlled evidence.