Supplement. The impact of AHI as a predictor of prevalence of somatic syndromes, anxiety and insomnia

Introduction

The primary article to which this supplement is attached elucidated the roles of the BSQ score and of gender as predictors of the prevalence of the functional somatic syndromes, of anxiety and of insomnia. The primary article looked specifically at the following 6 somatic syndromes:

·  Restless Legs Syndrome (RLS)

·  Headaches

·  Irritable Bowel Syndrome (IBS)

·  Temporomandibular Joint Syndrome (TMJ)

·  Fibromyalgia

·  Chronic Fatigue Syndrome (CFS).

With respect to anxiety, the following 3 parameters were considered:

·  Anxiety disorder Y/N

·  Nightmares – measured on a 4-point scale and dichotomized as described in the primary article

·  Use of benzodiazepines Y/N.

With respect to insomnia 4 parameters were considered. Three of them were ratings in response to questions on an insomnia questionnaire that were dichotomized as described in the primary article:

·  Have trouble falling asleep

·  Lie awake with intense thoughts

·  Wake up during the night

·  Use of hypnotics.

The primary article showed a greater prevalence of somatic syndromes in females than in males, a well-established feature of the somatic syndromes. A greater frequency of anxiety was also seen in females. However, there was no difference between males and females with respect to prevalence of insomnia. With increasing BSQ score, significant increases in prevalence were shown for most of the somatic syndromes, and anxiety and insomnia parameters.

In 2003, Gold and associates (ref) published a study of 75 patients split into 3 groups, 25 UARS patients, 25 mild/moderate OSAH patients and 25 moderate/severe OSAH patients, looking at the prevalence of a number of signs and symptoms. They reported statistically significant trends of increasing prevalence as AHI progressively decreased (i.e. negative correlation with AHI) for the following signs and symptoms:

·  Sleep onset insomnia

·  Headaches

·  IBS

·  Alpha-delta sleep

Thus, as one progressed from moderate/severe OSAH to UARS, prevalence of these signs and symptoms progressively increased. A significant negative correlation was not seen for rhinitis, depression, GERD, asthma, orthostatic syncope or hypothyroidism.

The overall pattern of negative correlation in the 4 signs/symptoms above was seen within each gender, but the patient counts were generally too small to test for statistical significance.

In this supplement, AHI is investigated as a co-predictor of the prevalence of somatic syndromes, of anxiety and of insomnia in conjunction with gender and BSQ score. Thus, this Supplement extends the primary article to which it is attached, which considered only gender and BSQ, and also extends the analysis from the 2003 paper, which considered only gender and AHI.

Analytical Methods and Preliminaries

As a first step, logistic regression models were fit to the prevalence data for the various disorders and parameters considered in the primary article within the categories:

·  Somatic syndromes – 6 disorders,

·  Anxiety – 3 parameters,

·  Insomnia – 4 parameters.

Prevalence of each disorder or parameter was modeled as a function of gender, BSQ score and AHI. For this purpose, patients were categorized according to the 4 AHI groups on which the patient population was originally constructed, 152 UARS patients, 50 mild OSAH patients (10≤AHI<30), 50 moderate OSAH patients (30≤AHI<60) and 50 severe OSAH patients (AHI>60). These groups were assigned the scores 0, 1, 2, 3 respectively and these values were entered into the model. Gender was therefore the only purely categorical predictor.

For each somatic syndrome and each anxiety or insomnia parameter, gender, BSQ score and AHI group were tested for statistical significance as predictors of prevalence, and all 2-way interactions between these 3 terms were tested as well. If, for a given parameter, AHI group is found to be a significant predictor of prevalence, the impact of AHI and any other significant predictors of prevalence is further illustrated by showing crude prevalence rates within subgroups determined based on the significant predictors. If AHI group is not found to be a significant predictor of prevalence, then no further results are shown for that parameter because the purpose of this supplement is to illustrate the impact of AHI.

In order to define subgroups and compute crude prevalence rates within them, BSQ score was categorized, as in the primary article, as high (BSQ>35), medium (25<BSQ≤35) or low (≤25).

Results

The covariates:

First we consider the relationships between BSQ, gender and AHI group. Table A1 shows the relationships between the three variables.

Table A1. The relationships between gender, BSQ score and AHI group

AHI Group
UARS / MILD OSAH / Moderate OSAH / Severe OSAH
Number of Patients / 152 / 50 / 50 / 50
% Female / 50 / 24 / 16 / 16
Mean BSQ Score – Females / 32.5 / 26.5 / 41.3 / 33.8
Mean BSQ Score - Males / 27.3 / 25.5 / 26.6 / 27.7

As noted in the primary article, UARS patients were split evenly among males and females, whereas males were over-represented in the OSAH groups. Among males, mean BSQ score in each AHI group was roughly 26-27. Among females, mean BSQ score varied around 33, more tightly in some AHI groups, more loosely in others. The tendency of females to have higher BSQ scores than males by 6-7 points, already noted in the primary article, holds up within AHI groups.

Thus we see that gender is correlated with AHI group (fewer women in higher AHI groups) and with BSQ score (women have higher BSQ scores), but that within each gender there is no correlation between AHI group and BSQ score.

Somatic syndromes:

For each of the somatic syndromes, Table A2 shows the statistical significance of the various candidate predictors of prevalence as determined by logistic regression. The p-values for the interaction terms shown in Table A2 come from the model with the 3 main effects and the 3 two-way interactions. The only statistically significant 2-way term was the interaction of BSQ score and AHI group as a predictor of prevalence of IBS (p=0.02). Given the lack of statistically significant interaction terms overall, and the relatively weak significance of the one significant term, the model was re-run with only the 3 main effects (gender, BSQ score, AHI group) and the p-values for the main effects in Table A2 come from this reduced model.

Table A2. P-values for gender, BSQ score and AHI group and their interaction terms as predictors of prevalence of somatic syndromes

Somatic Syndromes / Main effects / Interactions
Gender / BSQ / AHI / Gender
*BSQ / Gender
*AHI / BSQ
*AHI
Fibromyalgia / NS / 0.009 / 0.046 / NS / NS / NS
CFS / 0.03 / 0.01 / 0.04 / NS / NS / NS
Headache / <0.0001 / NS / NS / NS / NS / 0.02
IBS / 0.03 / 0.02 / 0.03 / NS / NS / NS
TMJ / 0.004 / NS / NS / NS / NS / NS
RLS / NS / <0.0001 / NS / NS / NS / NS

NS = Not significant.

P-values for main effects are from a logistic model that included only the main effects.

CFS = chronic fatigue syndrome; IBS = irritable bowel syndrome; RLS = restless legs syndrome; TMJ = temporomandibular joint syndrome.

AHI group was a significant predictor of prevalence of IBS, CFS and fibromyalgia. For CFS and IBS, gender, BSQ score and AHI group were all statistically significant predictors of prevalence. Therefore Table A3 shows the crude prevalence rates of CFS and IBS in the 24 subgroups determined by gender, BSQ score and AHI group. For fibromyalgia, only BSQ and AHI were significant predictors of prevalence. Therefore Table A3 shows prevalence rates for the 12 categories resulting from crossing BSQ score and AHI group, pooled over males and females.

Table A3. Prevalences of selected somatic syndromes in subgroups determined by gender, BSQ score and AHI group

Somatic Syndromes / Gender / BSQ Score / UARS / OSAH / All OSAH
Mild / Mod / Sev
CFS (%) / Male / ≤25 / 5.1 / 4.6 / 0 / 4.6 / 3.1
26-35 / 0 / 0 / 0 / 0 / 0
>35 / 16.7 / 33.3 / 0 / 0 / 6.7
Female / ≤25 / 18.2 / 0 / - / 0 / 0
26-35 / 16.1 / 12.5 / 0 / 0 / 6.7
>35 / 26.1 / - / 25.0 / 0 / 14.3
IBS (%) / Male / ≤25 / 12.8 / 4.6 / 0 / 9.1 / 4.6
26-35 / 16.0 / 23.1 / 5.9 / 0 / 9.5
>35 / 33.3 / 0 / 25.0 / 0 / 6.7
Female / ≤25 / 22.7 / 0 / - / 50 / 16.7
26-35 / 25.8 / 12.5 / 50.0 / 33.3 / 26.7
>35 / 43.5 / - / 25.0 / 0 / 14.3
Fibromyalgia (%) / Pooled / ≤25 / 6.6 / 3.9 / 0 / 0 / 1.4
26-35 / 8.9 / 4.8 / 4.8 / 6.7 / 5.3
>35 / 22.9 / 33.3 / 0 / 0 / 4.6

- No data

Table A4 is an important accompaniment to Table A3, showing the number of patients in each subgroup determined by gender, BSQ score and AHI, and also pooled over males and females.

Table A4. Sizes of subgroups determined by gender, BSQ score and AHI group

Gender / BSQ Score / UARS / OSAH / All OSAH
Mild / Mod / Sev
Male / ≤25 / 39 / 22 / 21 / 22 / 65
26-35 / 25 / 13 / 17 / 12 / 42
>35 / 12 / 3 / 4 / 8 / 15
Female / ≤25 / 22 / 4 / 0 / 2 / 6
26-35 / 31 / 8 / 4 / 3 / 15
>35 / 23 / 0 / 4 / 3 / 7
Pooled / ≤25 / 61 / 26 / 21 / 24 / 71
26-35 / 56 / 21 / 21 / 15 / 57
>35 / 35 / 3 / 8 / 11 / 22

Table A4 indicates that there are relatively few females with OSAH, so that prevalences among females in those categories are quite variable and conclusions should be drawn only from the pooled group of females with OSAH, with 28 members. Additionally there are relatively few OSAH patients with BSQ score >35, so that there is considerable variability in those values as well.

Table A3 shows that as AHI group increases from UARS to severe OSAH, the prevalence of CFS tends to decrease. Among males, with lower prevalence in general, the highest risk group, with BSQ score >35 and UARS, has a prevalence of 16.7%, whereas males with OSAH show a prevalence of 5% or less. Among females, with higher prevalence overall, the highest risk group, with BSQ score >35 and UARS, has a prevalence of 26.1%, whereas females with OSAH show a prevalence of approximately 10%.

Correspondingly, as AHI group increases from UARS to severe OSAH, the prevalence of IBS tends to decrease. Among males, with lower prevalence in general, the highest risk group, with BSQ score >35 and UARS, has a prevalence of 33.3%, whereas males with OSAH show a prevalence of 5%-10%. Among females, with higher prevalence overall, the highest risk group, with BSQ score >35 and UARS, has a prevalence of 43.5%, whereas females with OSAH show a prevalence of approximately 15%-25%.

As AHI group increases from UARS to severe OSAH, the prevalence of fibromyalgia also tends to decrease. The highest risk group, those with BSQ score >35 and UARS, has a prevalence of 22.9%. The lowest risk groups, those with OSAH, have a prevalence of 1%-5%.

In summary, of the 6 somatic syndromes studied, AHI was found to significantly predict prevalence of 3, CFS, IBS and fibromyalgia, among the more severe of the 6 somatic syndrome. In all 3, as AHI increased, the prevalence of the somatic syndrome decreased, even after controlling for differences in gender and BSQ score.

Anxiety parameters:

For each anxiety parameter, Table A5 shows the statistical significance of the various candidate predictors of prevalence as determined by logistic regression. The p-values for the interaction terms shown in Table A5 come from the model with the 3 main effects and the 3 two-way interactions. The only statistically significant 2-way term was the interaction of gender and AHI group as a predictor of prevalence of nightmares (p=0.03). Given the lack of statistically significant interaction terms overall, the relatively weak significance of the one significant term, and the absence of significance in the main effects making up the significant interaction term, the model was re-run with only the 3 main effects (gender, BSQ score, AHI group). The p-values for the main effects in Table A5 come from this reduced model.

Table A5. P-values for gender, BSQ score and AHI group and their interaction terms as predictors of prevalence of anxiety parameters

Somatic Syndromes / Main effects / Interactions
Gender / BSQ / AHI / Gender
*BSQ / Gender
*AHI / BSQ
*AHI
Anxiety disorder / NS / <0.0001 / NS / NS / NS / NS
Nightmares / NS / <0.0001 / NS / NS / 0.03 / NS
Uses benzodiazepines / 0.03 / NS / NS / NS / NS / NS

In general, AHI is not a significant predictor of anxiety.

Insomnia parameters:

For each insomnia parameter, Table A6 shows the statistical significance of the various candidate predictors of prevalence as determined by logistic regression. The p-values for the interaction terms shown in Table A6 come from the model with the 3 main effects and the 3 two-way interactions. There were no statistically significant 2-way terms. Therefore the model was re-run with only the 3 main effects (gender, BSQ score, AHI group). The p-values for the main effects in Table A6 come from this reduced model.

Table A6. P-values for gender, BSQ score and AHI group and their interaction terms as predictors of prevalence of insomnia parameters

Somatic Syndromes / Main effects / Interactions
Gender / BSQ / AHI / Gender
*BSQ / Gender
*AHI / BSQ
*AHI
Have trouble falling asleep / NS / 0.0002 / NS / NS / NS / NS
Lie awake with intense thoughts / NS / <0.0001 / NS / NS / NS / NS
Wake up during the night / NS / 0.003 / NS / NS / NS / NS
Uses hypnotics / NS / NS / 0.02 / NS / NS / NS

Table A7 shows a clear trend towards more use of hypnotics as AHI group decreases from severe OSAH to UARS. This is consistent with the trend seen in somatic syndromes of a higher frequency of sufferers among UARS patients.