The sample was selected from the out clinic patients of the Associação Fundo de Incentivo á Pesquisa, São Paulo (Brazil). Male patients between 25 and 65 years, with a BMI less/or equal to 35kg/m2 and an AHI greater than 5events/h were selected. We postulated that hormonal, anatomical differences and variations in the mechanisms of ventilatory control could influence the pharyngeal collapse process and consequently the requirements for titration of CPAP pressure, justifying the exclusive selection of men in this study. The patients voluntarily signed an informed consent form, and the study protocol was approved by the UNIFESP's research ethics committee (REC 0352/09). All procedures were in accordance with the 1964 Helsinki declaration.

Patients underwent full-night baseline polysomnography (Embla Systems, Inc., Broomfield, CO, USA), lateral cephalography and otolaryngological clinical evaluation. All patients also underwent a second full-night polysomnography for manual titration of CPAP pressure following Kushida's guidelines,2 using conventional nasal masks.

The image capture was done with patients keeping their teeth in normal occlusion, not swallowing and in natural head posture.18 A number of different parameters were evaluated by lateral cephalometry. The used cephalometric variables, as well as their definitions are described in Table 1.

The clinical upper airway assessment was performed by a trained otorhinolaryngologist according to the method of Zonato et al. (2003).19 Two otorhinolaryngologists with certification in Sleep Medicine were responsible for this evaluation.

We considered nasal conditions unfavorable, when one of these parameters were present: septum deviation Grade II or III; turbinate hypertrophy associated with rhinitis or complaint of obstruction of nose; septum deviation Grade I associated with rhinitis or complaint of obstruction of nose.

Oropharynx was considered unfavorable when one of the following three parameters existed: web palate; more posterior or thick palate; long or thick uvula; medial pillars; tonsils Grade III or IV.

Mallampati was considered unfavorable when modified Mallampati Class III or IV were present.

Full-night polysomnography was carried out for the baseline assessment, and the full-night polysomnography for manual titration of CPAP was performed according to the guidelines described by Kushida et al. (2008).2 The CPAP device used was REMstar Plus, with a nasal mask and the polygraph Alice 6.

For the elaboration of the equation, the following statistical methodology was adopted: the correlation coefficient was calculated using Pearson's correlation for all independent variables and for the ideal CPAP pressure. The variables that presented a statistically significant correlation or presented theorical relathionship with CPAP pressure were considered as potential predictors. A multivariate linear regression analysis was performed, using dummies variable that included categorical variable as predictor, after obtaining the predictive formula. The multiple regression procedures estimated the linear equation. To confirm this findings, the t-test was performed to compare between the pressure measured by the proposed formula and by manual titration in polysomnography.

SPSS version 18.0 (SPSS Inc., Chicago, IL) software was used, and statistical significance was set at p≤0.05.

Hoffestein and Miljeteig (1993) were one of the first groups to focus on the importance of developing a predictor equation for CPAP, correlating BMI, AHI, and neck circumference for a Caucasian population.7 Their formula was later validated by other authors.24,26–28 Oliver et al. (2000) suggested using the pressure predicted by this equation exclusively to begin the manual titration in laboratory.29 Masa et al. (2004) conducted a multicenter study in Spain with 360 patients who required CPAP therapy.23 The patients were randomized into 3 groups: manual titration during full-night polysomnography, CPAP auto-titration, and titration using Hoffestein's equation. The authors concluded that titration using this formula is as effective as the manual titration in patients with severe OSA, lowering costs and significantly decreasing waiting lists.23 It is important to emphasize that all previous studies have included patients with severe OSA, which differs from our sample, which included only mild to moderate OSA.

According to Basoglu et al. (2011),21 race and ethnicity are important factors for determining patients’ physical characteristics, the severity of OSA, and the CPAP pressure level that is necessary to resolve their apnea and hypopnea. These authors conducted a retrospective study in which 250 patients underwent polysomnography with manual titration. Various combinations of anthropometric and polysomnographic variables were tested. The obtained data a provided the development of a predictor equation for the Turkish population that included only the neck circumference and the oxy-hemoglobin desaturation index.21 In contrast, our study did not include patients with severe sleep apnea. Similarly to the Turkish population in Masa's study, the Brazilian population is highly mixed and might require the validation of a specific equation for this population.

Schiza et al. (2011)24 described a predictor equation for the Greek population that included BMI, AHI, gender, and smoking history. This prospective study compared the manual titration to Hoffestein's equation with an equation developed by the group.24 Although both equations predicted the ideal CPAP pressure, Hoffestein's equation was less effective, as determined in other populations.15,26–28 Lin et al. (2003) formulated a predictor equation based on the BMI and AHI of Taiwanese individuals and determined that neck circumference was not relevant for the Asian population,8 a fact that was also observed in our study. Likewise, in a retrospective study, Choi included 202 Koreans, 182 of whom were males with mild, moderate, or severe OSAS; a predictor equation was generated that similarly included only the two variables, BMI and AHI.11

Many studies have demonstrated that OSA patients exhibit important craniofacial alterations, which could explain the presence of OSA even in non-obese patients, especially in the Japanese population, where previous studies on cephalometric variables such as lower cranial base flexion17 and the position of the hyoid bone in relation with the chin16 were correlated with CPAP pressure.

Sforza et al. (1995) reported that the length of the soft palate affected the efficacy of CPAP in 22 French patients with OSA.30 However, because the soft palate is a component of the soft tissue, it is possibly influenced by obesity. This finding contrasts with our data in which the equation was not influenced by BMI, but was influenced by Mallampati score. In our study, the required CPAP pressure presented a positive correlation to the Mallampati III and IV classification, suggesting that the smaller volume of the upper airway requires higher CPAP pressure.

It has been suggested that retrognathia and obesity are associated with an inferior position of the hyoid bone.31 This fact might explain the negative relationship between the optimum CPAP pressure and mandibular length.

In the present study, we observed that the atlas–jaw distance and mandibular length showed a negative correlation with the optimal CPAP pressure. Even from the linear parameter, we could infer that there is a negative correlation between superior airway size and CPAP pressure.

Akahoshi et al. (2009)16 analyzed anthropometric, polysomnographic and cephalometric characteristics as predictors of therapeutic CPAP pressure. In the study, all of the patients had severe OSA and were moderately obese. When the cephalometric variables were combined with the polysomnographic parameters and the BMI, the therapeutic CPAP pressure was predicted more accurately, suggesting that craniofacial characteristics are critical to the pathogenesis of OSA in Asian populations.16

It should be noted that predictive equations cannot be a substitute for manual titration, but they could be used as an auxiliary procedure, helping to simplify the titration process and to decrease the number of changes of the pressure settings. However, predictive equations may be used as an alternative to manual titration in situations where a laboratory titration is either not possible (due to immobility, safety concerns, or critical health conditions) or when high costs, prolonged titration times, or lengthy waiting lists prevent or delay titration in sleep laboratories.

The study has some limitation, since the formula needs to be validated on a separate set of patients; we included only male patients and in this protocol, and an additional radiograph exam was required.