CLINICIANS' CORNER The accuracy of video orthognathic surgery imaging in Peter M. Sinclair, DDS, MSD, a Pauli Kilpelainen, DDS, b Ceib Phillips, PhD, MPH, Raymond P. White, Jr., DDS, PhDfl Lyna Rogers, RDH, BS, e and David M. Sarver, DMD, MS f Chapel Hill, N.C. Video imaging is an important emerging technology in planning orthognathic surgery and educating patients about the esthetic effects of treatment. The presurgical cephalograms of 56 patients, 41 with mandibular advancement alone and 15 with mandibular advancement plus genioplasty, were digitized, and the computer-generated soft tissue "line drawing" predictions were compared with the actual postt(eatment cephalograms. Video images of the patients' presurgical lateral view were obtained, and two experienced clinicians compared the computer generated video image predictions with the actual posttreatment profile. Video images judged very good or excellent were considered acceptable for treatment planning; 6% to 83% met this criterion, depending on the profile area viewed. The percentage of acceptable images in the lower lip, labiomental fold, and chin area decreased with the addition of a genioplasty. The predicted and actual posttreatment soft tissue line drawings were quite similar for all areas except for the lower lip region, where statistically significant differences were noted, with the predicted lower lip more retrusive and thinner than the actual contours. For the lower lip and chin, a mm or more discrepancy was observed in % of the patients. In all cases, the actual image was judged more esthetic than the predicted image, allaying fears of unrealistically optimistic computer generated predictions. (AM J ORTHOD DENTOFAC ORTHOP 1995;17:177-85.) With over 7% of prospective orthognathic surgery patients citing esthetics as a motivator for seeking treatment, 1 there is clearly a need for a treatment prediction technique that gives the clinician the diagnostic information required to plan treatment accurately, while also providing patients with a realistic simulation of the esthetic outcomes of their treatment. Currently, there are three methods by which orthognathic treatment outcomes can be predicted by using lateral cephalometric radiographs. The first method, most commonly used for the past years, involves manually repositioning acetate tracings of skeletal segments over the original Supported in part by NIH grant DE-515 from the National Institute of Dental Research, and by the Dental Foundation of North Carolina. "Professor and Chairman, Department of Orthodontics, University of Southern California, Los Angeles. bfaculty of Dentistry, Department of Orthodontics, University of Kuopio, Kuopio, Finland. ~Associate Professor, Orthodontic Department, University of North Carolina. dprofessor, Department of Oral and Maxillofacial Surgery, University of North Carolina. Research Associate, Orthodontic Department, University of North Carolina. fadjunct Professor, University of North Carolina; private practice, Birmingham, Ala. Copyright 1995 by the American Association of Orthodontists. 889-546/95/$3. + 8/1/53318 cephalometric tracing to simulate the proposed treatment. The posttreatment soft tissue outline is then added based on accepted ratios of soft tissue to hard tissue changes. This approach suffers from two major weaknesses: (1) the variables of soft tissue thickness, tonicity, individual patient response, as well as differences in the surgeon's soft tissue manipulation introduce uncertainties that make the soft tissue prediction as much an art form as a science; and () the "line drawing" of the soft tissue, while relatively informative to the clinician, often does not provide the patient with an adequate image of the proposed result. In the second technique, appropriate landmarks from the cephalometric tracing are digitized and entered into a computer. 3-6 Surgical movements are simulated on the screen and the treatment options compared. Hard copies of each option serve as visual aids in discussions with the patient. Manipulating the image by computer, however, is no more accurate than doing a prediction by hand, since the predictions are based on the same guidelines used in manual predictions. Although the provision of hard copies of several options for discussion with the patient may facilitate communication, they are still line drawings and do not provide a lifelike esthetic representation of the predicted outcome. 177
178 Sinclair et al. American Journal of Orthodontics and Dentofacial Orthopedics February 1995 Fig. 1. Cephalometric superimposition showing actual postsurgical tracing (dotted line) and computer generated prediction (solid line). The third prediction technique involves blending the digitized image of the lateral cephalometric tracing with a video image of the patient. 7-~1 The orthodontic and surgical predictions produced from the digitized cephalometric tracing are combined with the video image, so that the prediction includes both a line drawing and a corresponding facial image. Images of treatment options can then be displayed on the computer terminal, or a hard copy can be produced. Video imaging offers two significant advantages over previous prediction techniques. First, the predicted image facilitates communication between doctor and patient by establishing visual treatment goals for orthodontics and surgery. This addresses the patient's basic "need to know" and, by involving the patient in the selection of the treatment options, should improve patient acceptance of the outcome. 7'8 By patients developing joint goals with their doctors, clinicians with experience in video imaging report that there are fewer surprises, fewer unrealistic expectations, and a greater "bonding" between doctor and patient during treatment. The second suggested advantage of video imag- ing relates to its ability to aid in treatment planning decisions. 9-11 It provides the orthodontist and surgeon with a manipulable image so a consensus decision can be made on the desired soft tissue outcome. For example, the esthetic improvement of correcting a Class III malocclusion by advancing the maxilla can be compared with setting back the mandible. Video imaging can also be particularly helpful in deciding whether or not to perform adjunctive soft tissue procedures? The advantage of having an actual facial image to review and manipulate rather than just a "line drawing" or cutting up enlarged photographs to perform "photograph surgery" is particularly important when attempting to meet the patient's principal facial esthetic concerns. The questions regarding video imaging revolve around several issues. First, does the presentation of an image to a patient create an unrealistic expectation of the final result? Does sharing the image with the patient represent an implicit warranty as to the treatment outcome? In addition, can the surgeon actually produce the suggested outcome, or is the video image simply the result created by a skilled computer operator? 7 Some of these questions regarding patient expectations of video imaging were addressed by Sarver et al. 8 Their findings showed that 89% of the patients believed that the image predictions were realistic and the desired results were achieved. In their sample, 83% of the patients believed that the imaging process was beneficial in helping them make their decision whether or not to have surgery, and 7% believed that the imaging process allowed their participation in specific treatment decisions. The fear that a patient's expectations might become too great if provided a presurgical video prediction did not appear to be supported. In comparison, Kiyak's studies show that less than 45% of patients from a nonimaged population expressed esthetic satisfactionj This would support the contention that patients who have been imaged have more realistic expectations as to treatment outcome, and therefore the chances of dissatisfaction are reduced considerably. A major concern regarding video imaging centers on the accuracy of the prediction. Is it truly reflective of the probable outcome, too optimistic or too pessimistic? Are specific areas of the image, i.e., lips and chin, accurate enough only for display to the patient as a rough guideline, or can they be used as specific templates from which to plan the orthodontic and surgical movements.
American Journal of Orthodontics and Dentofacial Orthopedics Sinclair et al 179 Volume 17, No. Table h Difference between actual and predicted line drawing soft tissue values Discrepancy X+_SD Range P value < - mm (percent) Upper lip Upper lip-snv (ram) -.1 + 1.1-3./+3..49 5 Upper lip--e-plane (ram) -.39 -+ 1.7-4.4/+4..9 Upper lip-hard tissue +. -+ 1.58-4.3/+5..99 11 (ram) Upper lip-ui labia 1 sur- -.7 -+ 1.57-4.3/+4.9.71 8 face (mm) Subnasale-stomion (ram) -.4 -+ 1.86-4.1/+5..11 1 Superior labial sulcus - SNV +.4 -+.7-1.5/+ 1.9.6 (mm) Lower lip Lower lip-snv line (mm) -.76 ±.5-5./+4.5.1 3 Lower lip-e-plane (ram) -1. -+.18-7.1/+3.7.1 3 Lower lip-hard tissue -.86 -+ 1.76-5.5/+4.4.1 4 (mm) Lower lip-li labial sur- -.84 + 1.8-6.3/+4.9.1 5 face (mm) Stomion-ILS (mm) +.39 -+ 3.38-9.3/+11.44.38 18 Chin ILS-SNV (ram) +.44 -+ 1.83-5./+4.5.7 4 Pgs-SNV (ram) +.1 -+.3-5.4/+11..99 7 ILS-Me (ram) +.57 -+.98-6.9/+6.5.15 4 Pogonion (soft) to hard +.13 -+.17-4./+6..5 15 tissue Soft to hard tissue Me +.4 -+ 1.91-4.5/+6.7.1l 11 (ram) Pogonion to soft Pg (ram) -.19 -+.19-5.8/+7..5 18 Menton to soft Me (ram) +.7 -+.19-5.8/+7..35 15 > + mm (percent) 1 6 6 7 1 6 4 - indicates prediction value smaller or more retrusive than actual value. + indicates prediction values larger or more protrusive than actual value. These questions regarding the accuracy of video imaging have yet to be scientifically investigated and therefore led to these two research questions: (l) Is the video image accurate and realistic enough to use in presentation to patients? and () Are specific areas of the image accurate enough for use in treatment planning and, if different, in what way and why? This study addresses these questions. MATERIALS AND METHODS The sample used in this study consisted of 56 patients drawn from the office of one of the authors (D.S.), who had completed treatment involving orthodontics and orthognathic surgery to advance the mandible a minimum of 5 mm. Forty-one patients underwent a mandibular advancement only, whereas the remaining 15 patients had an additional advancement genioplasty. All the patients were over 18 years of age, white, and selected on the basis of the availability of presurgical and posttreatment lateral cephalograms and lateral profile photographs. All records were taken with teeth together in centric occlusion, and the lips in repose. The presurgical and posttreatment cephalograms were digitized with the Prescription Planner/Portrait software program (Rx Data Inc, Ooltewah, Tenn.), and a cranial base superimposition was performed. The actual amounts of anteroposterior and vertical surgical change that had occurred during treatment was measured at the lower incisor tip, B point, pogonion, and menton. With these values, a line drawing profile prediction was generated from the presurgical cephalogram. The hard tissue cephalometric prediction was then superimposed on the digitized posttreatment hard tissue cephalogram. Thus, with the computerized hard tissue prediction and the actual hard tissue final result superimposed, it was possible to compare and analyze the line drawing cephalometric soft tissue outlines (i.e., the actual final soft tissue outline versus the predicted soft tissue outline) to determine the accuracy of the soft tissue line drawing prediction (Fig. 1). With a special soft tissue analysis module of the software program developed for this project, 18 linear or angular soft tissue measures (Table I) on the upper lip, lower lip, and chin were obtained on the digitized posttreatment cephalogram and on the prediction line draw-
18 Sinclair et al. American Journal of Orthodontics and Dentofacial Orthopedics February 1995 Fig.. A, Presurgical video image with computer generated postsurgical predicted cephalometric line drawing superimposed on it. B, The resultant computer generated surgical video image prediction for this patient. ing. The differences between the actual posttreatment soft tissue and the predicted posttreatment soft tissue "line drawing" were analyzed with paired t tests. The difference in the discrepancies observed in the two surgical groups was compared with unpaired t tests. Errors in the soft tissue of the predicted line drawing will be reflected in the predicted video image since the image is based on the predicted line drawing. Level of significance was set at.1 because of the number of analyses performed. Video images of the patient's presurgical and posttreatment lateral facial photographs were captured with a CCD-F33 camera (Sony Corp., Tokyo, Japan) set at a standardized distance with uniform background lighting, and displayed on a flat screen RGB monitor using the Prescription Portrait software program. The predicted posttreatment line drawing was overlayed on the presurgical video image and a predicted posttreatment video image generated (Fig. ). All the predictions were created without the computer operator employing any of the smoothing and blending functions available in the program. Although these functions can improve the prediction image, the use of these functions is subjective and may vary from patient to patient. To assess whether the video image predictions were accurate and representative enough to display to patients, the actual initial, actual final, and predicted final video images of one patient at a time were simultaneously displayed on the monitor (Fig. 3). These three images were evaluated and scored independently by an oral surgeon (R.W.) and an orthodontist (P.S.) for how well the predicted final image matched the actual final image. Both clinicians, who had been in clinical practice for at least 1 years, were full-time faculty members and team members in the Dentofacial Deformity clinic through which potential surgical-orthodontic treatment patients are evaluated. Neither clinician had been involved in any aspect of treatment for these 56 patients. The clinician's perception of the concordance between the actual posttreatment and predicted video images was evaluated by using the following 5-point scale: : Poor-little agreement between predicted and actual images 1: Fair-general form of prediction acceptable but not clinically representative : Good-predicted image clinically representative but with noticeable (by any observer) differences from actual image 3: Very good-predicted image clinically accurate with only minor (only noticeable by a trained clinician) differences from actual image 4: Excellent-predicted image indistinguishable from actual image The clinicians discussed the use of the scale before scoring began but no formal calibration session was held. Perceptual assessments were made at the upper lip, lower lip, labiomental fold, chin, and submental areas. Small differences (one scale unit) occurred frequently between the examiners, resulting in low intraclass correlation values. Rho values ranged from.3 for the upper lip to.53 for the labiomental fold. However, a difference of two scale units occurred in only 4% of the
American Journal of Orthodontics and Dentofacial Orthopedics Sinclair et al. 181 Volume 17, No. Fig. 3. Representation of actual presurgical image, computer generated postsurgical prediction, and actual postsurgical video image (left to right) as displayed on screen for evaluation. patients for the upper lip and in only 1% for the least consistent point, the labiomental fold. There were no differences of three scale units. Because of the low proportion of major discordances between the two observers bat the low exact agreement, the sum of the two observer's scores for each area was used as the outcome measure. The Mantel Haenszel row mean score test was used to compare the predictive utility of the areas of the face between the mandibular advancement with and without genioplasty groups and the Mantel Haenszel test for matched designs, the equivalent to the Cochran's Q criterion for measures with more than two responses, was used to compare the predictive utility for the areas on the face. 13 RESULTS Actual versus predicted line drawings Because the average discrepancies for the genioplasty group were not statistically different (p >.3) from those of the mandibular advancement alone group, the data were combined for analysis. Differences between actual and predicted measures on the posttreatment soft tissue profile are shown in Table I. As the table shows, the mean difference between actual and predicted was quite small for measures other than those of the lower s lip and were not statistically significant. For the lower lip, the differences between actual and predicted are statistically significant at the.1 level. The computer predicted lip was found to be significantly (p <.1) more retrusive when com- pared with both the E-line and a vertical line dropped through subnasale (SNV-line). The lower lip was also predicted to be considerably thinner (p <.1) than was actually the case, both in the area of B point and overlying the lower incisor. The predicted length of the lower lip as measured from stomion to inferior labial sulcus, however, was not sitgnificantly different from the real outcome. Since the average can mask differences in the distribution of discrepancies, the percentage of the patients with a difference of mm or more between actual and predicted line drawings also is shown in Table I. For 11 of the 18 measures, a mm or greater discrepancy was noted in at least % of the patients. Note that the percentage with differences at chin points is higher than might have been expected from the mean changes alone. In part, this is due to the genioplasty patients whose changes at the chin were not as consistent as the nongenioplasty group. For all the measures except the inferior labial sulcus depth, the percentage of patients for whom the mm discrepancy was an "under prediction" exceeded the "over prediction" percentage. Actual versus predicted video images Data for the judgment of the video images by the two clinicians are shown in Tables I! and III. For the mandibular advancement only group, the predicted images were perceived as agreeing with
18 Sinclair et al American Journal of Orthodontics and Dentofacial Orthopedics February 1995 Table II. Accuracy of video imaging predictions for mandibular advancement patients (n = 41) Scale Area evaluated 5 Clinically acceptable Poor 1 Fair Good 3 Very good 4 Excellent (3 and 4 combined) Upper lip % % 15% 51% 3% 83% Lower lip % % 7% 39% 3% 71% Labiomental fold % 7% 31% 39% 3% 6% Chin % 5% 4% 49% % 71% Submental area % 1% 8% 39% 1% 6% Percentage of patients showing: O: Poor-little agreement between predicted and actual images; 1: Fair-general form of prediction acceptable but not clinically representative; : Good-predicted image clinically representative but still not with noticeable (by any observer) differences from actual image; 3: Very' good-predicted image clinically representative with only minor (only noticeable by trained clinician) differences from actual image; 4: Excellent-predicted image indistinguishable from actual image. Table III. Accuracy of video imaging predictions for mandibular advancement and genioplasty patients (n = 15) Scale Area evaluated Poor 1 Fair Good 3 Very good 5 Clinically acceptable 4 Excellent (3 and 4 combined) Upper lip % % 1% 43% 57% 9% Lower lip 3% 1% % 37% 3% 67% Labiomental fold 7% 13% 3% 33% 17% 5% Chin % 1% 37% 5% 3% 53% Submental area % % 3% 58% % 77% the actual image most frequently in the upper lip area; the labiomental fold and submental areas showed the poorest agreement. For the genioplasty group, the percentage of acceptable images was higher for the submental area but lower for the lower lip, labiomental fold, and chin; these differences were not statistically significant, but approached significance at the chin (p >.4). The predicted images were perceived as agreeing with the actual image most frequently in the upper and lower lip regions. The scores given the upper lip were significantly higher (p <.1) than the labiomental fold, chin, and submental areas. At the chin, 71% of the advancement alone and 53% of the genioplasty patients had acceptable predictions (Fig. 4). None of the predicted images were viewed by either clinician as more favorable than the actual image. DISCUSSION As the average discrepancies between the actual soft tissue outline and the predicted line drawing were small, and the predictions can be judged accurate enough for display to patients and for most treatment planning purposes. Nevertheless, it must be kept in mind that a mm or more discrepancy occurred in more than % of the patients for all the lower lip and several of the chin measures, so only in 39 of the 56 cases were the predictions judged accurate enough in these areas for detailed treatment planning. This probably reflects the paucity of published data on the response of the lower lip to orthognathic treatment, especially when face height is changed as the mandible is advanced. Current prediction algorithms that rely on the ratios of averages do not reflect the variability of the responses and the interrelationships of horizontal, vertical, and transverse dimensions. To improve the accuracy of the computer generated line drawings, better data are needed for the soft tissue changes that accompany vertical and horizontal changes in the position of bone and teeth. The predicted line drawing of the soft tissue profile was used in this study for two reasons: First, the software program did not, in the version used, have the capability to measure directly on the
American Journal of Orthodontics and Dentofacial Orthopedics Sinclair et al 183 Volume 17, No, 1 gox [] Mandibular Advancement Only roll U) 8,~. 6.~ 4 C G) L. 1. Upper Lip Lower Lip Labiomental Chin Submental Fold Area Fig. 4. Percentage of video image predictions judged to be clinically acceptable (levels 3 and 4 from Tables II and III). predicted video image; and second, the assumption was that the video image modification was directly based on the predicted line drawing. However, even if the actual cephalogram and predicted line drawings were coincident, the video image did not always exactly reflect the line drawing on which it was based. The specific landmarks the computer used to generate the video image differed slightly, in this version of the software program, in a few areas from those used in generating the line drawing (e.g., the most anterior point on the lower lip versus the junction of thevermillion and the rest of the lip). Such deviations in computer software will only compound the prediction error if soft:hard tissue prediction ratios derived from cephalograms are applied to noncoincident measures. 14 This may require the development of specific video imaging soft tissue ratios for specific programs (Table IV). As subjectively viewed by the two clinicians, all the video image predictions were considered acceptable for patient education, even without the use of any computerized smoothing or blending functions to improve the esthetics of the image. With the addition of this step, which should take only 1 to minutes per case, the performance of this or any other program is likely to be significantly improved (Fig. 5). While these functions are in essence "prettifiers," which do not improve the scientific information content, they are useful in providing "clean" images for patient education. It was also clear to the evaluators that when simultaneously viewing the actual and predicted images on Table W. The ratios developed in this study of the soft tissue to hard tissue changes seen in the video image predictions Nose anterior.17.15 Nose base.4.8 Soft tissue point A.5.6 Upper lip anterior point.6.31 Upper lip stomion.6. Lower lip stomion.4.1 Lower lip anterior point.7.35 Soft tissue point B 1. 1. Soft tissue pogonion 1. 1. the screen, in all cases the acutal final images was superior in esthetics to the prediction. In no case did the predicted image produce an outcome better than the surgeon was able to achieve in real time. This was due to the computer image always presenting with a slightly rasterized or jagged appearance. With this fear allayed, we found considerably greater interest from the surgeons in using the imaging system as a tool for patient education. Concerns about the image providing an implied warranty can be addressed by having phrases such as Treatment Simulation Only appear on the screen overlayed on the computer image, or by the use of a video imaging informed consent form, such as that recommended by the American Academy of Facial Plastic and Reconstructive Surgery. 15 As
184 Sinclair et al American Journal of Orthodontics and Dentofacial Orthopedics February 1995 Fig. 5. Video image prediction before and after application of computerized smoothing function. Sarver has suggested, ~ with the majority of patients having seen and recognized many other computer simulations in other situations for what they are, just "simulations," the advantages of having a clinically acceptable image for patient education and discussion seems to far outweigh this potential risk. As one gains experience with video imaging, there is a learning curve for the operator that relates to the accuracy and ease of use of this technology. It is clear that taking accurate and reproducible records is a critical first step. The presurgical video images in this study were obtained from slides rather than directly from the patient. The discrepancies between the actual and predicted records observed in this study likely are larger than would be found in a prospective patient evaluation with direct imaging, since imaging from slides allows no correction of background or patient positioning. Video imaging requires excellent lighting and a white background behind the patient; otherwise the software program may experience difficulty in defining the edges of the patient's profile against the background. This can lead to a "fuzzy" soft tissue image. It is also vital that the video image be captured and the cephalogram be taken with the patient in the same head position and with the lips in a similar repose position. Ideally, the physical arrangement of the cephalometer and camera should allow for the simultaneous capturing of the two images with the patient in a reproducible position. Differences in head position, particularly rotational, as well as differences in lip posture, were probably responsible for some of the imaging errors seen in this study. Although this study primarily focused on mandibular anteroposterior movements, considerable vertical change occurred in many of the advancements. It appeared that the greater the mandibular advancement and particularly the greater the vertical change, the poorer the lower lip prediction became. In addition, the larger adjustments showed greater need for smoothing in the submental area. Future video imaging studies must therefore carefully evaluate vertical changes, including those involved with maxillary movements. It should be pointed out that this study was carried out with the technology available in the first half of 199 and considerable improvements in methods have occurred since that time. Also, the study represents the results of only one software program's approach to video imaging, and studies of other programs should be initiated to compare outcomes and broaden our knowledge and understanding of video imaging. Future studies should also include psychologic evaluations of patients who were video imaged as part of their orthognathic treatment so as to understand their responses, as has been done by Kiyak et al.' for several groups of orthognathic patients. Improved soft tissue ratios specific to imaging need to be developed and comparisons of the same patients imaged on different software systems need to be
American Journal of Orthodontics and Dentofacial Orthopedics Sinclair et al. 185 Volume 17, No. carried out so that the best approach to imaging can be identified. Currently video imaging is being applied to the mature adult face. Considerable scientific evaluation of the technique is required to determine whether it represents a scientific breakthrough, or whether its principal role is in patient education. Any consideration of its application to the growing face should await resolution of these questions. REFERENCES 1. Kiyak HA, Bell R. Psychosocial considerations in surgery and orthodontics. In: Proffit WR, White RP, eds. Surgical orthodontic treatment. St Louis: Mosby Year-Book, 1991: 71-95.. McNeil RW, Proffit WR, White RP. Cephalometric prediction for orthodontic surgery. Angle Orthod 197;4:154-64. 3. Bhatia SN, Sowray JH. A computer-aided design for orthognathic surgery. Br J Oral Maxillofac Surg 1984;:37-53. 4. Harradine NWT, Birnie DJ. Computerized prediction of the results of orthognathic surgery. J Maxillofac Surg 1985;13: 45-9. 5. Walters H, Walters DH. Computerised planning of maxillofacial osteotomies: the program and its clinical application. Br J Oral Maxillofac Surg 1986;4:178-89. 6. Lew KK. The reliability of computerized soft tissue prediction following bimaxillary anterior subapical osteotomy, Int J Adult Orthod Orthognath Surg 199;7:97-11. 7. Laney TJ, Kuhn BS. Computer imaging in orthognathic and facial cosmetic surgery. Oral Maxillofac Clinics North Am 199;:659-68. 8. Sarver DM, Johnston MW, Matukas VJ. Video imaging for planning and counseling in orthognathic surgery. J Oral Maxillofac Surg 1988;46:939-45. 9. Sarver DM, Johnston MW. Video imaging: techniques for superimposition of cephalometric radiography and profile images. Int J Adult Orthod Orthognath Smg 199;5:41-8. 1. Takahashi I, Takahashi T, Mitsuhiko H, et al. Application of video surgery to orthodontic diagnosis. Int J Adult Orthod Orthognath Surg 1989;4:19-. 11. Turpin DL. Computers coming on line for diagnosis and treatment planning. (Editorial) Angle Orthod 199;6: 163-4. 1. Sarver DM, Matukas VJ, Weissman SM. Incorporation of facial plastic surgery in the planning and treatment of orthognathic surgical cases. Int J Adult Orthod Orthognath Surg 1991;6:7-39. 13. Duritz SJ, Landis JR, Koch GG. A general overview of Mantel-Haenszel methods: applications and recent developments. Annu Rev Public Health 1988;9:13-6. 14. Jensen AC, Sinclair PM, Wolford LM. Soft tissue changes associated with double jaw surgery. AM J ORTHOD DENTO- FAC ORTHOP 199;11:66-75. 15. Hopping SB. Image thyself. Facial Plast Surg 199;7:45-56. Reprint requests to: Dr. Peter M. Sinclair University of Southern California School of Dentistry Orthodontic Department University Park Los Angeles, CA 989-641 JOURNAL BACK ISSUES ON CD-ROM Back issues of the AMERICAN JOURNAL OF ORTHODONTICS AND DENTOFACIAL OR- THOPEDICS are now available on CD-ROM. The complete and full text of 15 articles has been reproduced. Twenty-six volumes consisting of more than 13, pages, 1, black and white pictures and diagrams, 5 tables, and 1 color pictures have all been compressed onto compact disk. In addition, the disk will contain a Book of Abstracts and a Book of Landmark Articles as suggested by Editor-In-Chief, T.M. Graber. The new AJO-DO on CD-ROM will be of enormous benefit to education, research, and publication of articles. Direct all inquiries to a Gold Optimedia/ORMCO representative; 1-8-854-1741, ext. 777; international: (818) 85-91, ext. 777.