In developing countries, where patients do not enjoy reimbursement for sleep study, non-invasive mechanical ventilation and CPAP, majority of the patients admitted to the sleep clinics are in moderate to severe OSA stage and cardiovascular disease (Setareh, Mehrnia, & Mirabi, 2018).In these patient, device titration is recommended in sleep lab. Optimal device titration is a titration that cannot only remove respiratory events while having optimum leak but also lets the patient have a comfortable sleep and suitable oxygen saturation. This would be extremely helpful in sleep lab, specifically after midnight when only the technician can attend the patient. On the other hand, in such a situation, changing CPAP into BiPAP can be extremely expensive for the patients, so the more precisely we can decide on the type of device drawing on various parameters, the shorter is the time needed by the technician to make a logical decision. This would ultimately lead to a more reliable night time titration of device.
Two seminal studies in the field of Manual Titration Obstructive Apnea have been introduced according to AASM Guidelines (Berry et al., 2010; Kushida et al., 2008).The first study by Kushida addresses the OSA patients without comorbidities. The study takes into account variables such as the number of respiratory events, supine position, and REM Sleep, but does not provide any protocol for low SaO2 except during events. The second study by Berry is conducted on patients suffering from hypoventilation and OSA, and explains the protocol for utilizing BiPAP in low SaO2 situation during wake and sleep from the onset. However, in a number of patients such as those participating in the current study, Baseline SaO2 is 94.7% which declines to lower than 84.6% in NREM stage. As a result, this can contribute to the likelihood of changing CPAP to BIPAP (Fig. 2).
Studies on mechanism of cortical control of ventilation showed that the duration of breath-holding can be limited by several factors, such as sensitivity of peripheral chemoreflex, spirometry parameters, the arterial PCo2, PO2,previous maneuver (Bain et al., 2017; Trembach & Zabolotskikh, 2018).Ventilatory response to PCo2 and PO2 is considerably variable between patient (West & Luks, 2016).The duration of voluntary breath-holding doubled after breathing a hyperoxic mixture or after pre-hyperventilation. Those whose minimum post breath-holding SaO2 was lower, might have a higher likelihood of changing CPAP to BiPAP. This might be useful for technicians as a straight forward method to have an estimation of these ventilatory responses and help them to choose BiPAP over CPAP (Table 2). The borderline predictive role for minimums oxygen saturation in voluntary breath-holding maneuver in the study may be limited by variable peripheral chemoreflex and spirometry parameters in these patients in normal range (Bain et al., 2017).
At different levels of hypoxia, there is a nonlinear response to oxygen and most of the responses to level of PO2 are less than 50 mmHg. These physiological findings display considerable variability among individuals. Therefore, it seems that measuring oxygen desaturation during NREM sleep and minimum oxygen saturation during voluntary breath-holding maneuver are the predictors of a need of BiPAP during titration with a good trade-off between sensitivity and specificity shown through ROC curve analysis.
Some other studies attempted to find formulas specific for different races. Basoglu et al. proposed a new formula with the same variables among Turkish population (Basoglu & Tasbakan, 2012).Their formula utilized neck circumference (NC) and oxygen desaturation index (ODI). It can be easily concluded from every titration protocol that more severe OSA patients need BiPAP more frequently (Kushida et al., 2008). Our results were in concordance with more severe OSA and lower SaO2 during NREM sleep.
Camacho et al. in their systematic review of 26 studies on mathematical equations of CPAP prediction, reported BMI and mean oxygen saturation as the most heavily weighted variables along with BMI, AHI, and neck circumference as the most frequently used variables (Camacho et al., 2015).
The likelihood of changing from CPAP to BiPAP was almost doubled by higher BMI and neck circumference in our study population. Meanwhile, Hoffstein formula (Miljeteig & Hoffstein, 1993) is the most widely used CPAP prediction tool worldwide, which uses BMI, AHI, and neck circumference. Although it is widely used for many years and validated in many different studies, this study suggests that these parameters were not enough compared to other factors (Fig. 2).
OSA is also a well-known etiology of REM deprivation. OSA causes the reduction in REM sleep percentage. The more intense OSA, the shorter the length of REM (Kimoff, 1996; Wang et al., 2015).Therefore, a relative short REM can indicate an increase in the probability of need for BiPAP.
The present study was designed to evaluate BiPAP versus CPAP preference by anthropometric, polysomnographic, and other type of data on a clinical maneuver called voluntary breath-holding maneuver. We used decision tree analysis to find if there is any relationship between patient’s variables and the need to use BiPAP or CPAP, by using some of the most common polysomnographic and anthropometric variables. Although these results cannot render the titration procedure useless completely, it is one of the limitations of the study but can help the technicians to pay attention to special variables in order to save time, specifically during split-night titration and to choose the more suitable device as soon as possible.
The main focus in this study was using of logistic regression analysis as a parametric model to find an equation between independent variables with respect to its dependent variable. Technically by logistic regression application, it was estimated the effect of each independent variable on the odds of change of device therapy from BIPAP to CPAP.On the other side, we gained advantages of the decision tree to find out considerable overlap between the variables used in the logistic regression equation and the decision tree. It might be suspected that the two procedures are picking the same cases like what was seen about REM duration and NREM minimum SaO2.