INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: sleep apnea syndromes; quality of life; clinical trials; questionnaires

Sleep apnea results from repetitive narrowing or complete obstruction of the upper airway leading to oxygen desaturation. Apneic events are usually terminated by brief arousals from sleep with associated surges in blood pressure. The recurrent arousals from sleep, and possibly the oxygen desaturations, are thought to cause sleepiness, the most common daytime manifestation of sleep apnea. In previous treatment studies, the severity of sleep apnea and the extent of improvement have usually been evaluated by changes in physiologic measures such as the apnea-hypopnea index (AHI), and in sleepiness symptom scores. However, physiologic measures and symptom scales may fail to adequately measure the true impact of the disorder. Recently, sleep apnea has been shown to adversely affect patients' quality of life. Initially, this was shown with generic health status measures such as the Nottingham Health Profile and the Medical Outcome Survey Short Form (SF-36) (1, 2). However, these questionnaires contain no questions that are specific to sleep disorders such as sleep apnea, and may therefore be insensitive to important changes in quality of life that sleep apnea patients experience as a result of therapy.

Health-related quality of life questionnaires may be used to assess how much change there has been in a patient's condition over time, usually after some type of therapeutic intervention (an evaluative instrument). Alternatively, some health-related quality of life questionnaires are designed to distinguish between those patients with a better from those with a worse quality of life at a single point in time (a discriminative index) (3).

The Calgary Sleep Apnea Quality of Life Index (SAQLI) was developed as an evaluative instrument to measure within-subject change in response to a therapeutic intervention (4). During its development we demonstrated that the items in each of the four SAQLI domains of normal daily routine, social interactions, emotional functioning, and symptoms were similar across strata of disease severity and sex. We therefore constructed a single questionnaire to measure quality of life in adult patients with sleep apnea. The questionnaire was developed according to published guidelines (5) that had been successfully used to develop validated disease-specific quality of life questionnaires for disease states such as asthma (6), chronic obstructive pulmonary disease (7), heart failure (8), rhinoconjuctivitis (9), and inflammatory bowel disease (10). Like these questionnaires, the SAQLI uses a 7-point Likert scale ranging from 1 (maximal impairment) to 7 (no impairment). Domain scores are averaged by dividing by the number of questions answered, and the total score is averaged over the four domains so that all scores maintain a total range of 1 to 7. Unlike other quality of life questionnaires, the SAQLI includes a fifth domain, to capture some of the adverse consequences of currently available therapies for sleep apnea. This fifth domain, comprising treatment-related symptoms, is used after a therapeutic intervention, and is subtracted from the other four domains in determining the total SAQLI score. Details of the use of the questionnaire and its scoring have been previously published (4).

In a small cross-sectional study, we previously demonstrated that the SAQLI had a modest correlation with the vitality domain of the SF-36 (4). The current study was designed to assess and report a detailed analysis of the measurement properties of the SAQLI. Since the SAQLI was designed to be an evaluative instrument, we focused on determining the responsiveness and longitudinal construct validity of the questionnaire by evaluating a series of sleep apnea patients before and after a 4-wk trial of nasal continuous positive airway pressure (CPAP). Our preliminary work also suggested that the SAQLI had useful discriminative properties; we therefore further evaluated this feature in the same patient cohort before the initiation of CPAP.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

A total of 90 patients with diagnosed sleep apnea participated in the study trial. Sleep apnea was diagnosed and quantified with the use of a portable unattended monitor that has been validated against polysomnography (11). The monitor samples Sa_O2 at a frequency of 1 Hz and converts it to a digital signal, which is stored in computer memory until analyzed by a custom-designed, automated software program that determines the number of oxygen desaturations  4% from baseline, and reports them as respiratory disturbances. The details of the analysis algorithm have been previously published (12). The number of respiratory disturbances divided by the hours during which the oxygen probe is attached to the patient is defined as the respiratory disturbance index (RDI). All patients with an RDI  10 were invited to participate in the study. Patients were excluded from the study if they were under the age of 18 yr, had severe respiratory disease (FEV₁ < 50% predicted, arterial carbon dioxide tension [Pa_CO2] > 45 mm Hg, arterial oxygen tension [Pa_O2] < 55 mm Hg), cardiac ischemia (myocardial infarction in the previous 6 mo or unstable angina), or neuromuscular disease. The study was approved by the University of Calgary Conjoint Ethics Review Board, and all subjects provided informed consent.

Study Design

All patients had a baseline assessment of their sleep apnea severity by measurement of their RDI. The SAQLI, the SF-36, and the Ferrans and Powers Quality of Life Index were all administered by trained research assistants before and after a minimum of 4 wk of home CPAP use. A majority of patients completed the SAQLI on two other occasions prior to the CPAP trial, at 1 and 2 wk after the first administration, respectively, so that its reproducibility could be assessed. All patients completed an overnight study of Sa_O2 while receiving CPAP to determine their RDI and thus their physiologic response to therapy. Patients were included in the analysis of the change in SAQLI scores if they met the minimum definition of compliance with CPAP therapy.

Polysomnography and CPAP Titration

All patients underwent a complete night of CPAP titration during polysomnography. Patients were instrumented in a standard fashion. A pneumotachometer was placed in line proximal to the CPAP mask so that a continuous recording of flow could be evaluated throughout the CPAP titration study. The titration was performed by trained, experienced technicians who increased CPAP gradually so that all apneas, hypopneas, snoring, and periods of inspiratory flow limitation were eliminated.

CPAP Trial

Patients were instructed in the proper use of CPAP by experienced therapists, to whom the patients had easy access for 5 d per week so as to quickly resolve any problems that developed during their home trial of this therapy. All patients were provided with a CPAP machine (Tranquility Plus; Healthdyne Technologies, Marietta, GA) that had a pressure-activated recording of the time for which the machine was turned on. Patients were not informed about the covert monitoring of their compliance with CPAP until after the study was completed. All patients were asked to sign poststudy consent forms informing them of the covert compliance monitoring and requesting that their research data be kept and used in the analysis of the present study. No patient refused this request. Patients used the CPAP machines for a minimum of 4 wk. If any patient was having difficulty tolerating CPAP, all attempts were made to correct these problems, and some patients may have had the CPAP machine for up to 8 wk. Patients were judged to have been compliant with CPAP if they had used it for a minimum of 40 h during the final 2 wk of their home trial.

Quality of Life Assessment

Quality of life was assessed with four methods: the SAQLI, the SF-36 (13), a single global rating of quality of life (using a 10 point scale with 1 = worst possible and 10 = best possible), and the Ferrans and Powers Quality of Life Index (14). The Ferrans and Powers index has four domains; health and function, socioeconomic, psychological/spiritual, and family. Each domain and the total score have a range of 0 to 30, calculated by averaging all the item scores. The SF-36 was chosen as a comparison because of its widespread use in many different disease states, and because it has previously been shown to be responsive in patients with sleep apnea who are treated with CPAP (1). The Ferrans and Powers index was chosen because it is one of the few quality of life measures that asks patients for a judgment about satisfaction and importance in each domain, and because of its previous use in other sleep disorders such as narcolepsy (15).

Reliability

The reliability of the SAQLI was determined by administering it to stable subjects with sleep apnea on repeated occasions (at 1 wk and 2 wk after the first administration) before treatment with CPAP. The within-subject SD was calculated for each of the two repeat administrations, and a reliability (intraclass correlation) coefficient was calculated as the ratio of the between-subject variance to total variance (16).

Discriminative Properties

Construct validity was determined by correlating SAQLI scores at baseline with domains of the SF-36, the global quality of life score, and the Ferrans and Powers Quality of Life score. The observed correlations were compared with a priori predictions made by a sleep apnea specialist about the magnitude of the correlations that should be expected between the SAQLI and other measures of quality of life and a commonly used subjective index of sleepiness, the Epworth Sleepiness Scale (17).

Determining the Minimal Clinically Important Difference of the SAQLI

Changes in quality of life with therapy need to be interpretable and should be based on what is important to patients. We determined the minimum clinically important difference for the SAQLI based on methods used to ascertain this value for similarly constructed questionnaires (18, 19). A rating-of-change questionnaire was developed for the SAQLI, and has been previously described (4). In brief, for each item of the SAQLI, patients were asked if their rating of that item had changed for them, and if their response was yes, whether they had improved or worsened and by how much. The degree of change was rated on a 7-point Likert scale ranging from hardly improved (or worsened) to a very large improvement (or worsening), for a total of 15 possible responses (7 to +7). These scores were then averaged for each domain by dividing by the number of questions in the domain. A total rating-of-change score was calculated by summing the four domains and subtracting treatment-related symptoms before dividing by four, as has been described for the SAQLI. Each of the domain rating-of-change scores and the total rating-of-change scores were recoded as follows: No change (1 to 1); minimal important difference, whether better (1.1 to 3) or worse (1.1 to 3); moderate difference, whether better (3.1 to 5) or worse (3.1 to 5); and large difference, whether better (5.1 to 7) or worse (5.1 to 7).

Evaluative Properties

Responsiveness.A key property of an evaluative instrument is its responsiveness: the ability to measure small but clinically important within-subject change after an effective therapeutic intervention. Responsiveness can be viewed as a description of an instrument's signal-to-noise ratio. The signal can be measured in two ways. The first is the mean within-subject change after treatment (mean  treatment). However, since there is no way to know if this difference is clinically significant, the second and more meaningful method is to use the minimum clinically important difference (MCID) (20). The noise in the instrument is the variability of scores when the instrument is administered on two or more occasions to a cohort of stable patients (pretreatment). This estimate can be derived from the analysis of variance (ANOVA) model with the examination of repeated test observations, which can be used when calculating reliability (VAR  baseline) (20). We therefore calculated two indices of responsiveness: RESP_meantreat = mean  treat/VAR  baseline, and RESP_MCID = MCID/VAR  baseline.

Another method of determining how well an instrument detects change is the calculation of effect size. This is used in the situation in which repeated measurements at baseline have not been made, in which case a surrogate of measuring the non-treatment-related variability in the instrument is required. This can be done with the between-subject variability of the within-subject change in scores (posttreatment  pretreatment).

A paired samples t test was used to compare mean results at baseline and after CPAP treatment.

Validity

Construct validity was assessed with hypotheses based on how the within-subject change in SAQLI scores (posttreatment  pretreatment) in patients who had proven to be compliant with treatment correlated with the within-subject change in scores of the SF-36 domains, Ferrans and Powers Quality of Life Index, global quality of life scores, and improvement in RDI. The observed correlations were compared with predictions made a priori by a sleep apnea specialist.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The 90 patients who participated in the study were predominantly male (81%), with an age of 52.1 ± 10.4 (mean ± SD) yr (see Table E1 in the online data supplement). They had a wide range of RDI values (mean ± SD: 51.4 ± 32.5) and Epworth Sleepiness Scale scores (mean ± SD: 12.6 ± 5.2). Mean scores for the SAQLI and the global quality of life rating were in the middle of their scales. The scores for the SF-36 were below published general population values in all domains (21).

Discriminative Properties

To establish construct validity for the SAQLI as a discriminative measure, we determined correlation coefficients at baseline (pre-treatment) for the SAQLI versus RDI, subjective sleepiness, and other measures of quality of life (Table 1). The table compares observed correlation coefficients with predicted correlations. There was no correlation with the physiologic measure of sleep apnea severity as measured with the RDI. There was a weak correlation with the Epworth Sleepiness Scale and modest to strong correlations with the domains of the SF-36. The SAQLI was predicted to have a stronger correlation with the RDI than was observed, whereas stronger than predicted correlations were observed for most of the domains of the SF-36. It was predicted that the strongest correlation would be seen between the SAQLI and the vitality and social functioning domains of the SF-36, which was what was observed.

Reliability

Repeatability of the SAQLI at 1 and 2 wk is shown in Tables 2 and 3. There were no significant differences in domain or total scores of the SAQLI among its three different administrations (Table 2); however, there was variability in the scores when the SAQLI was readministered (Table 3). There was slightly more variability between the first and third week scores than between the first and second week scores. The reliability (intraclass correlation) coefficient was 0.92, and did not change appreciably between Week 1 and Week 2. The reliability coefficient was lowest for the symptoms domain.

Evaluative Properties

Of the 90 patients who completed the baseline assessment, 62 completed at least 4 wk of CPAP therapy and were proven to meet the minimum definition for compliance. There were no significant differences between completers and noncompleters at baseline in age, percentage of females, subjective sleepiness, RDI, or any of the quality of life scores. Comparison of the status of the compliant patients after the completion of CPAP with their baseline results showed significant changes in almost all variables (Table 4). Construct validity is shown in Table 5. Within-subject changes in the SAQLI were most highly correlated with changes in RDI, the global quality of life rating, the vitality and social functioning domains of the SF-36, and the health and functional domain of the Ferrans and Powers index. These results correspond closely with the predictions of how the SAQLI would relate to these other variables. Effect sizes calculated for the SAQLI as compared with the other quality of life indices are shown in Table 6. All of the SAQLI domains and the total score had very large effect sizes, whereas only the vitality domain of the SF-36 had a large effect size.

When patients rated aspects of their quality of life affected by sleep apnea as improved with CPAP by a small but important amount, the mean change in their domain and total scores ranged from 1.1 to 2.2 (see Table E2 in the online data supplement). As expected, the mean changes for moderate improvement versus a large improvement were correspondingly larger for the latter. These results indicate that the minimal important difference in the total SAQLI score is approximately 1, that a moderate difference is approximately 2, and a large difference is approximately 3.

The responsiveness index for the total SAQLI score, based on the minimum clinically important difference (RESP_MCID) of 1.0, was 1.9, and was identical with the index calculated on the basis of the mean within-subject change (RESP_meantreat). The estimate of the between-subject variability for the responsiveness index was based on the Week 2 versus baseline comparison of scores in stable subjects. The responsiveness indices (RESP_MCID) for domains A through D were 1.5, 1.7, 1.6, and 2.3, respectively.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The SAQLI has previously been shown to have high internal consistency, with Cronbach  scores ranging from 0.88 to 0.92 for both domain and total scores. In this study, we demonstrated that the SAQLI has excellent responsiveness with mean change scores of 1.0 in patients who were proven to use nasal CPAP over a minimum of 4 wk. The responsiveness index was very high for all domains and for the total score. Effect sizes were much higher for the SAQLI domains and total score than for any of the SF-36 domains other than the vitality domain. The values for responsiveness were higher for the SAQLI than for the Asthma Quality of Life Questionnaire (22).

The change in SAQLI scores with treatment were correlated, in accord with a priori predictions, with changes in most of the SF-36 domains, changes in a global rating of quality of life, and changes in the RDI. The greatest correlation was observed between the SAQLI total score and the vitality domain of the SF-36, the domain that would be expected to best reflect the effect of sleep apnea on quality of life. The vitality domain has previously been shown to have the largest effect size in sleep apnea patients undergoing treatment with CPAP (1). The correlations between SAQLI total score and the role physical and bodily pain domains of the SF-36, in accordance with a priori predictions, were low and not statistically significant. These results show strong evidence for construct validity, indicating that the SAQLI measures components of health-related quality of life that are important to sleep apnea patients.

The SAQLI was found to have some variability in each of its domains when it was administered on repeated occasions before CPAP treatment. The within-subject SDs of the mean differences were higher for the individual domains than that found in a similarly constructed questionnaire used for asthma (22). The reasons for this are not obvious, but may be related to whether subjects reviewed their baseline scores when completing the SAQLI at subsequent visits prior to treatment. In the Asthma Quality of Life Questionnaire, subjects reviewed their baseline answers before repeating the questionnaire (6). When subjects took a repeat administration of the SAQLI, they had their baseline scores available for reference if they so wished; however, fewer than 20% of patients referred to these scores. Although it is likely that this would increase the within-subject variability, we believe that it provides a closer reflection of the expected variability of baseline scores when the SAQLI is in general use. The reliability coefficient (intraclass correlation) was very high, indicating that in repeated administrations of the SAQLI, the within-subject variability in scores was low as compared with the between-subject variability in scores. The SAQLI shows a wide range of scores at baseline, without any obvious floor or ceiling effects.

The methods used to determine the minimum clinically important difference in SAQLI scores were similar to methods described for other disease-specific quality of life questionnaires. One notable difference was that in the present study, patients were asked to rate the degree of change on an item-by-item basis rather than by domain. We obtained a higher minimum important difference for the SAQLI than that found for similarly constructed questionnaires for other disease states (18, 19). This may have been due to methodologic differences; however, we suspect that it is more likely to be related to a fundamental difference in the disease states. Sleep apnea patients have often struggled with quality of life issues for years without recognizing what their cause may have been. Many patients with apnea initiate referrals to sleep clinics through their attending physician because of information they have heard from family, friends, and the media. Given this, we suspect that their expectations are probably higher than those of patients with other diseases such as chronic obstructive lung disease or asthma. Although sleep apnea patients' quality of life improves with treatment, it has been previously demonstrated that as a group, they do not return to general population norms (1). Thus, if they do not meet their expected degree of improvement, their rating-of-change scores may be lower. Another factor that may have influenced the minimum important difference in SAQLI scores is that although patients were provided an opportunity to refer to their baseline SAQLI scores when completing the rating-of-change questionnaire, only a small minority chose to do so. Therefore, it is possible that they had improved to a greater extent than they were able to remember. We believe that an increase in SAQLI scores of 1.0 or greater represents a clear and clinically meaningful improvement in the quality of life of sleep apnea patients, is well above the baseline within-subject variability, and is associated with a very large effect size. On the basis of our data, we recommended a conservative approach, with use of a minimal clinically important difference of 1.0 in SAQLI scores for the purpose of planning clinical trials and calculating sample sizes. Changes of 0.5 to 1.0 in SAQLI scores represent an indeterminate area in which the signal-to-noise ratio is likely to be poor. Further research utilizing different methods of determining the minimal clinically important difference in SAQLI score and assessing patients' expected improvements a priori, is needed to better define the meaning of a change in SAQLI total score of 0.5 to 1.0.

The SAQLI was able to discriminate patients with a better from those with a worse quality of life, as demonstrated by the high correlations with several of the SF-36 domains and the global quality of life rating, as was predicted a priori. The SAQLI did not correlate with the RDI at baseline, which would seem to suggest that it does not discriminate between patients with mild sleep apnea and those with severe sleep apnea as defined by a physiologic measurement. However, in this study there was also no correlation at baseline between the RDI and the SF-36 or the global rating of quality of life. There is no doubt that patients with sleep apnea may have other life issues that affect the quality of their lives and which would be captured by scales like the SF-36 and the SAQLI, and would reduce the correlations of these instruments' scores with the baseline RDI. However, there are numerous examples of the poor correlation between RDI and other outcome variables, such as daytime sleepiness (23), neuropsychologic functioning (24), and motor vehicle accident rates (25). We therefore believe that the RDI poorly characterizes the potential effect of sleep apnea on symptoms and quality of life, and that its lack of correlation with the SAQLI and the SF-36 reflects this well recognized problem and should not be misconstrued as a problem with these quality of life indices.

The moderate to strong correlations observed between the SAQLI and other quality of life measures is a clear indication of the ability of the SAQLI to discriminate between patients in whom sleep apnea has had a more adverse effect on the quality of life from those in whom the disorder has had relatively little effect.