| Sign In to gain access to subscriptions and/or personal tools. |
Predictors of Obstructive Sleep Apnea-Hypopnea Treatment Outcome
1 Department of Oral and Maxillofacial Surgery, Correspondence: * corresponding author, a.hoekema{at}kchir.umcg.nl
Oral appliance therapy is an alternative to continuous positive airway pressure (CPAP) for treating the obstructive sleep apnea-hypopnea syndrome. However, the ability to pre-select suitable candidates for either treatment is limited. The aim of this study was to assess the value of relevant variables that can predict the outcome of oral appliance and CPAP therapy. Fifty-one patients treated with oral appliance therapy and 52 patients treated with CPAP were included. Relevant clinical, polysomnographic, and cephalometric variables were determined at baseline. The predictive value of variables for treatment outcome was evaluated in univariate and multivariate analyses. The outcome of oral appliance therapy was favorable, especially in less obese patients with milder sleep apnea and with certain craniofacial characteristics (mandibular retrognathism in particular). Neither univariate nor multivariate analyses yielded variables that reliably predicted the outcome of CPAP. We conclude that the variables found in this study are valuable for pre-selecting suitable candidates for oral-appliance therapy.
Key Words: sleep apnea syndromes • orthodontic appliances • positive-pressure ventilation • treatment outcome • predictors
The obstructive sleep apnea-hypopnea syndrome is a common sleep-related breathing disorder characterized by disruptive snoring and repetitive upper airway collapse (Malhotra and White, 2002). Continuous positive airway pressure (CPAP) is the preferred treatment for sleep apnea (Giles et al., 2006). Since maintaining CPAP requires that patients wear an obtrusive device, patients may abandon therapy. Oral appliance therapy is an alternative to CPAP that relieves upper-airway collapse during sleep by modifying the position of the mandible, tongue, and pharyngeal structures (Cistulli et al., 2004). Although there is evidence that oral appliance therapy is effective for sleep apnea, it is generally considered less effective than CPAP (Barnes et al., 2004; Hoekema et al., 2004; Lim et al., 2006). Nevertheless, many patients prefer an oral appliance over CPAP therapy (Hoekema et al., 2004). Therefore, predictors of treatment outcome are important for the selection of suitable candidates who may benefit from either treatment. Numerous clinical and polysomnographic variables have been reported to correlate with increased effectiveness of oral appliance therapy. For example, the outcome of treatment is generally more favorable in patients who are less obese (Pancer et al., 1999; Liu et al., 2001) and who have a lower apnea-hypopnea index (Rose et al., 2002a). In addition to the fact that specific bony- and soft-tissue features may characterize the upper airway of sleep apnea patients (Okubo et al., 2006), numerous cephalometric variables have also been implicated in the outcome of oral appliance therapy (Horiuchi et al., 2005). Predictors of treatment outcome, however, are not uniformly reported (Henke et al., 2000; Marklund et al., 2004). The majority of these studies incorporate bias, because patients with severe sleep apnea or patients who have not adhered to therapy have been excluded (Marklund et al., 2004). In addition, predictors have not been systematically validated to evaluate their accuracy in a separate population of patients (Lim et al., 2006). Therefore, clinicians’ ability to predict treatment outcome and pre-select suitable candidates for a specific treatment modality is still limited. The aim of this study was to assess the value of relevant clinical, polysomnographic, and cephalometric variables, separately and jointly, to predict the outcome of oral appliance and CPAP therapy.
Patient Selection Patients were recruited through the Department of Home Mechanical Ventilation of the University Medical Center Groningen (The Netherlands) for a randomized parallel non-inferiority trial comparing the effects of oral appliance and CPAP therapy (Hoekema et al., 2006). Patients over age 20 years, who underwent polysomnography and were diagnosed as having obstructive sleep apneahypopnea syndrome, were eligible (AASM, 1999). Patients were selected based on medical, psychological, and dental criteria. The trial was approved by the Groningen University Medical Center’s ethics committee. Written informed consent was obtained from patients before enrollment. Details of the trial are provided in the APPENDIX.
Study Design After patients had used an oral appliance or CPAP for approximately a two- to three-month period, the treatment effect was assessed with polysomnography. At the final follow-up review, treatment was considered effective when the apnea-hypopnea index either was <5 or showed "substantial reduction", defined as reduction in the index of at least 50% from the baseline value to a value of < 20 in a patient who had no symptoms while using therapy (Hoekema et al., 2004). Patients not meeting these criteria at their final review were considered "non-responsive" to treatment. Patients who discontinued treatment for any reason were considered "non-adherent" to treatment.
Clinical Predictors
Polysomnographic Predictors
Cephalometric Predictors
Statistical Analysis First, in each treatment group, variables were submitted for univariate analysis. Categorical variables (i.e., sex, sleep apnea-severity, and supine-dependence of sleep apnea) were submitted only for multivariate analysis. The univariate analysis consisted of calculation of ‘receiver-operating characteristics’ curves of each variable, with ‘treatment effectiveness’ and an ‘apnea-hypopnea index < 5 following treatment’ being the dependent variables, respectively. To obtain a summary measure for the predictive ability of each variable, we calculated the area under the curve of each ‘receiver-operating characteristics’ curve. The predictive ability of a variable was classified based on the area under the curve (excellent = 0.9 to 1, good = 0.8 to 0.9, fair = 0.7 to 0.8, poor = 0.6 to 0.7, or non-discriminative = 0.5 to 0.6) (Swets, 1988). All variables with at least fair predictive ability were admitted for logistic regression analyses. By excluding variables stepwise backward, we constructed predictive models in both the oral appliance and CPAP groups, with ‘treatment effectiveness’ and an ‘apnea-hypopnea index < 5 following treatment’ being the dependent variables, respectively. Logistic regression analyses also produced odds ratios associated with each predictor value. We used a discriminant analysis to select the predictive model that classified the highest percentage of patients correctly. Subsequently, the selected model was cross-validated in another discriminant analysis by the application of a "leave-one-out classification".
Two patients in both the oral appliance and CPAP groups did not return for the follow-up review. The median period to final review was 68 (interquartile range, 60–96) days in the oral appliance group and 63 (interquartile range, 60–88) days in the CPAP group (p > 0.05). At final follow-up review, mean advancement of the mandible with the oral appliance was 81 ± 19% of maximum advancement. Mean CPAP pressure was 8.1 ± 1.9 cm H2O at final review. Oral appliance therapy was effective for 39 patients (79.6%). Of the other 10 patients, eight were "non-responsive", and two were "non-adherent" to treatment. In the CPAP group, treatment was effective for 43 patients (86.0%). Of the other seven patients, two were "non-responsive", and five were "non-adherent" to treatment. Oral appliance therapy yielded an apnea-hypopnea index < 5 in 29 of the 49 patients (59.2%). CPAP therapy yielded an apnea-hypopnea index < 5 in 40 of the 50 patients (80.0%).
Clinical and Polysomnographic Predictors
Cephalometric Predictors In predicting the effectiveness of oral appliance therapy, a larger intermaxillary discrepancy (i.e., higher angle between the lines connecting point A with Nasion and point B with Nasion [ANB]) had good predictive ability, and a greater mandibular deficiency (i.e., smaller angle between the lines connecting Sella with Nasion and Nasion with point B [SNB]), a larger overjet and overbite, and a greater upper anterior face height had fair predictive ability (Fig.  , a; Table 2 ). In predicting an apnea-hypopnea index < 5 following oral appliance therapy, all cephalometric variables had poor predictive ability or were non-discriminative.
Multivariate Analysis The logistic regression analysis for predicting the effectiveness of oral appliance therapy yielded a model providing an 84% correct classification of responders and non-responders. Variables included in the model were the apnea-hypopnea index, SNB angle, and ANB angle (Table 3 ). When the selected model was cross-validated, 80% of the patients were classified correctly. The logistic regression analysis for predicting an apnea-hypopnea index < 5 with oral appliance therapy yielded a model providing an 80% correct classification of responders and non-responders. Variables included in the model were the apnea-hypopnea index and the maximum mandibular advancement (Table 3). When the selected model was cross-validated, 80% of patients were classified correctly.
The logistic regression analysis for predicting the effectiveness of CPAP therapy yielded a model providing a 65% correct classification of responders and non-responders. Variables included in the model were the body mass index, apnea-hypopnea index, sleep apnea-severity, and supine-dependence of sleep apnea (Table 3 ). When the selected model was cross-validated, 54% of patients were classified correctly. The logistic regression analysis for predicting an apnea-hypopnea index < 5 with CPAP therapy yielded a model providing a 65% correct classification of responders and non-responders. Variables included in the model were the body mass index and apnea-hypopnea index (Table 3). When the selected model was cross-validated, 65% of patients were classified correctly.
Univariate analysis demonstrated that a lower body mass index, more extended maximum mandibular advancement, smaller apnea-hypopnea index, higher ANB angle, smaller SNB angle, and more pronounced overjet, overbite, and upper anterior face height were the best predictors for outcome of oral appliance therapy. Our results concur with those of previous studies that demonstrated that these clinical (Marklund et al., 1998a; Pancer et al., 1999; Liu et al., 2001), polysomnnographic (Pancer et al., 1999; Henke et al., 2000; Mehta et al., 2001), and cephalometric (Yoshida, 1994; Mayer and Meier-Ewert, 1995; Liu et al., 2001) variables correlate with increased effectiveness of oral appliance therapy. Except for upper anterior face height, the cephalometric predictors found in the present study primarily relate to mandibular retrognathism. Contrary to other reports, variables including supine dependence of sleep apnea (Yoshida, 2001; Marklund et al., 2004) or pharyngeal dimensions and hyoid bone position (Yoshida, 1994; Mayer and Meier-Ewert, 1995; Liu et al., 2001; Mehta et al., 2001; Rose et al., 2002b; Skinner et al., 2002; Horiuchi et al., 2005) could not be implicated in the outcome of oral appliance therapy. These results indicate that patients who are less obese, have milder sleep apnea, and have certain craniofacial characteristics (mandibular retrognathism in particular) respond more favorably to oral appliance therapy. Multivariate analysis yielded predictive models that, following cross-validation, classified the outcome of oral appliance therapy correctly in 80% of patients. Variables included in the predictive models were maximum mandibular advancement, apnea-hypopnea index, and the SNB and ANB angles. Recent studies have suggested an important role for more sophisticated techniques to predict the outcome of oral appliance therapy. For instance, a remotely controlled mandibular positioner (Tsai et al., 2004) or upper airway imaging techniques, including sleep nasendoscopy (Battagel et al., 2005) or magnetic resonance imaging (Sanner et al., 2002), have been shown to be highly predictive of the patients’ response to oral appliance therapy. Although these techniques may be of additional value in the selection of suitable candidates, they are generally costly, laborious, or sensitive to a specific operator. The present study aimed at constructing a predictive model convenient for the clinical situation. By using variables that are relevant and that can be easily determined, we obtained two predictive models that allow for a reliable prediction of the effectiveness of, or an apnea-hypopnea index < 5 with, oral appliance therapy. Contrary to oral appliance therapy, clinical and polysomnographic variables had poor predictive value for the outcome of CPAP therapy. Moreover, multivariate analysis yielded predictive models for an effective treatment and apnea-hypopnea index < 5 with CPAP therapy that, following the cross-validation, classified only 54% and 65% of patients correctly, respectively. These results indicate that patients in whom CPAP therapy does not have a favorable outcome cannot be easily pre-selected. Unfortunately, in clinical practice, these patients are usually deemed the best candidates for oral appliance therapy (Lim et al., 2006). The limited predictive ability of the predictive models is possibly best explained by the fact that CPAP therapy was effective and yielded an apnea-hypopnea index < 5 in the majority of patients. One may question the extent to which the results found in this study can be extrapolated to other types of oral appliances. Although different aspects in the design of oral appliances may affect patient preference, clinical effects of different oral appliances that reposition the mandible are usually remarkably consistent (Hoekema et al., 2004). Moreover, the present study evaluated only variables that had been implicated in the outcome of oral appliance therapy in previous studies. We therefore believe that the results from this study also apply for predicting the outcome of most other types of oral appliances that reposition the mandible. A second aspect that requires consideration is the fact that cephalograms were obtained from patients in the upright position. The use of supine rather than upright cephalograms has been reported to account for the influence of posture on upper airway dimensions (Johal and Battagel, 1999). This may explain why pharyngeal dimensions and hyoid bone position could not be implicated in the outcome of oral appliance therapy. However, the added value of supine cephalometry should also be considered in the light of its time-consuming and operator-sensitive character. Finally, it should be recognized that we evaluated primarily predictors of the effect of therapy on the apnea-hypopnea index. Other important outcomes—like effects on neurobehavioral or cardiovascular parameters (e.g., sleepiness or hypertension) or therapeutic compliance—were not evaluated in the present study. In conclusion, the outcome of CPAP therapy could not be predicted reliably with the clinical and polysomnographic variables evaluated in this study. Conversely, predictive variables obtained from the univariate and multivariate analysis, including obesity, disease severity, and certain craniofacial characteristics (mandibular retrognathism in particular), were valuable for pre-selecting suitable candidates for oral appliance therapy.
The authors thank Dr. G.J. Pruim (Department of Orthodontics of the University Medical Center Groningen) and Mr. B.K. Uildriks for their assistance in the cephalometric analysis and graphics lay-out, respectively. We also thank dental laboratory Goedegebuure (Ede, The Netherlands) for their assistance in manufacturing the oral appliances. The present paper was written in partial fulfillment of the requirements for a PhD degree. Financial support for this MD-clinical research traineeship was granted by the Netherlands Organisation for Health Research and Development, The Hague, The Netherlands.
A supplemental appendix to this article is published electronically only at http://www.dentalresearch.org. Received for publication May 27, 2006. Revision received June 25, 2007. Accepted for publication September 12, 2007.
Journal of Dental Research, Vol. 86, No. 12,
1181-1186 (2007)
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
|
|
 |
|
Sign In to gain access to subscriptions and/or personal tools.
TOP
ABSTRACT
INTRODUCTION