Learning Center
Plans & pricing Sign in
Sign Out



									BMI 546                                  Gangnon                                       July 2, 2010




In this exercise, you will be introduced to a broad spectrum of study designs in clinical and
epidemiologic investigation. In particular, you will evaluate the advantages and disadvantages of
different study designs and the measures of risk and association used in such studies. You will
also learn how to assess whether an observed association is likely to be causal and how to deal
with the concept of confounding.

After completing this Exercise, students will be able to:

   1. Understand design features, strengths and limitations of different study designs in clinical
      epidemiologic investigation.

   2. Calculate the different measures of risk and association used in epidemiologic studies.

   3. Understand sources of bias in observational and experimental studies.

   4. Understand the concept of confounding and a way to address this.

The specific area of research that is used to illustrate these issues and concepts is the
hypothesized association between sleep apnea and hypertension. This is an area of increasing
concern because of the potentially large clinical and public health significance of such an
association, should it be proven to be causal. It is important to note, however, that research in
this area has evolved only recently (in the last 30 years or so) and that some of the implications
of this research remain controversial still today.

The specific studies used to illustrate the different study designs options in this exercise were
picked based on convenience (type of data fitting the specific concept that was to be illustrated)
and also because they follow a chronological sequence more or less parallel to the evolution of
the rationale supporting the hypothesis. These studies, however, do not necessarily represent the
―best‖ or most representative studies in this area.

The following section provides a brief description of the pathophysiology of sleep apnea.

                                            1 of 18
BMI 546                                 Gangnon                                      July 2, 2010

PART 1: Background

Sleep apnea (aka ―Sleep Disorder Breathing‖ or ―SDB‖) is the common term for a spectrum of
disorders characterized by repeated breathing pauses during sleep. The main clinical
manifestations of this syndrome are snoring and excessive daytime sleepiness. The breathing
pauses may be complete (no air flow, apnea) or partial (reduced airflow, hypopnea). The most
common form of this syndrome, known as ―obstructive sleep apnea‖ (or ―OSA‖) is caused by
increased resistance and partial or total collapse of the upper airway due to loss of tone in the
pharyngeal muscles. The vulnerability to collapse may be due to increased fat deposition in the
neck, but the mechanism is not known. A less common form of the syndrome, ―central sleep
apnea,‖ results from lack of brain signal to respiratory muscles usually secondary to other
disorders with neurological damage, such as stroke. This exercise will focus on obstructive sleep
apnea, a disorder we now know to be quite prevalent (but often undiagnosed) in the general
population, particularly among overweight persons (see below). Appendix 1 contains a recent
review article (1) describing the epidemiology and pathophysiology of obstructive sleep apnea.
A brief summary follows.

Pathophysiology of obstructive sleep apnea

With the onset of sleep, there is a relaxation of the tone of the upper airway. This happens in
everyone, but in some people, the upper airway develops more compliance (becomes "floppy"),
and as air is drawn through the airway in trying to breathe, resistance increases. This makes the
airway vulnerable to collapse. (Think of trying to drink water through a floppy straw-the harder
you try, the more the straw collapses.) During an apnea, the respiratory muscles work harder and
harder, to increase the force in trying to breathe against a closed airway, resulting in an even
more tightly collapsed airway; the resulting sound of air going through a collapsing airway is
heard as snoring.

Significant breathing pauses (apnea or hypopnea) are those lasting 10 seconds or more and result
in a number of profound physiologic changes that can be assessed using polysomnography
(PSG). PSG is a multiple channel recording of the patient‘s sleep that includes measures of brain
and muscle activity (to characterize sleep stage), electrocardiogram, breathing (air flow and
respiratory movements), blood pressure, and blood oxygen levels (see figure 2 in Appendix 1).
One immediate consequence of the breathing pause is a momentary drop in blood oxygen levels
(hypoxemia or hypoxia). This drop in blood oxygen saturation may be just a few percentage
points (e.g., 98% to 94%) or might go as low as 60%. This hypoxia is detected by physiological
sensors (chemoreceptors and baroreceptors) that signal an alarm to the brain—here is a choice
between sleeping and breathing: the result is a brief arousal from sleep, during which time the
awake tone of the airway returns, and clears the apnea. Usually, the person is not even aware of
the arousal, and goes back to sleep only to go on and have another apnea, arousal, apnea, and on
and on. With each event, there is an increase in sympathetic nerve activity, and this too has
consequences. There are fluctuations in heart rate and rhythm, and blood pressure drops and
then spikes as much as 120 mm Hg.

One consequence of the sleep arousal is that it fragments the continuity of sleep. The arousals
disturb the normal pattern of going from light to deeper sleep and rapid eye movement (REM)

                                           2 of 18
BMI 546                                  Gangnon                                       July 2, 2010

sleep, in a few orderly cycles, over the night. When these arousals occur frequently over the
course of the night (i.e., in cases of severe OSA) the affected person is never able to go into deep
sleep. This may cause extreme daytime sleepiness, and the inability to fight off drowsiness and
unwanted sleep. In addition, when these events occur several times an hour, night after night, the
continuous burst of sympathetic activity and resulting hemodynamic and cardiopulmonary
consequences are likely to have adverse effects that may impair daytime or overall health.

Clinical diagnosis and basic epidemiology of OSA

A common metric of the severity of sleep apnea is the average number of pauses per hour of
sleep (known as the apnea-hypopnea index, or AHI) and is clinically assessed using a PSG exam.
The number of breathing pauses may range from zero to >100 per hour of sleep. The occurrence
of occasional apneas/hypopneas is considered normal, but an average of >5 per hour is usually
considered physiological evidence of sleep apnea. AHI>15 and AHI>30 are cut-points used to
define the presence of moderate or severe sleep apnea, respectively.

OSA was considered a rare disorder until fairly recently. Since the breathing pauses occurred
during sleep, struggles to breath were usually not noticed—and snoring or daytime sleepiness
were not considered signs of a disorder, but rather something to joke about. Thus, only severe
cases were seen, and treatment was tracheotomy (certainly an unpopular treatment!). This all
changed in 1982 with the invention of nasal continuous positive air pressure (nCPAP)—a nasal
mask that delivers pressurized room air and keeps the airway open during sleep. When an
effective and acceptable treatment became available, clinicians were more likely to look for OSA
in their patients, and the National Institutes of Health became more interested in finding out more
about the burden of OSA in the population.

The first large population study was begun in 1987, the Wisconsin Sleep Cohort. The study was
designed to take a random sample of the population, and bring them into a sleep laboratory for a
PSG exam and repeat this every four years. In this way, the prevalence could be estimated at
baseline, and over time, incidence could be determined. With the longitudinal data, people with
and without OSA could be followed to compare the incidence of health consequences. One of
the first findings, published in 1993 in the New England Journal of Medicine (2), was that
contrary to belief, OSA was quite prevalent in adults, and there was a wide severity spectrum.
Most importantly, only a small (<7%) of cases had ever been diagnosed—leaving 93% of the
total burden undiagnosed. The study showed that 4% of women and 9% of men had an AHI of
15 or greater. The significant proportion of women with OSA was even more unexpected than
the overall prevalence. In clinics, the ratio of men to women with OSA was 9:1, but in the
population, the ratio was about 2:1, showing that there was selection bias for men to be
diagnosed with OSA.

Studies over the past few decades have identified some risk factors, such as aging, male gender,
overweight, abnormal facial structure, and nasal congestion. Importantly, clinical and
epidemiologic studies have shown that OSA is likely to have significant health consequences
such as behavioral morbidity associated with cognitive impairment and excessive sleepiness:
depression, occupational injuries, automobile crashes, as well as cardio- and cerebro-vascular
disorders. The latter includes hypertension, the subject of this exercise.

                                            3 of 18
BMI 546                                 Gangnon                                      July 2, 2010

PART 2: Early clinical observations

Reports concerning breathing disorders during sleep are remarkably absent from the medical
literature up to around the beginning of the 1970s. One of the cardinal, but non-specific,
symptoms of the syndrome, snoring, was at the time not considered a medical problem but rather
an annoyance to bed partners and other family members of the affected persons. The other main
symptom, excessive daytime sleepiness, was often perceived as a sign of ―laziness.‖ As a result,
patients with this disorder were not only undiagnosed and untreated, but often laughed at and
constantly embarrassed; the sleepiness tendency was often associated with mood disturbances
that frequently resulted in a diagnosis of ―depression‖ with patients sustaining anti-depressive
treatments for the rest of their lives.

In the late 1960s and early 1970s, and thanks in part to the development of polysomography
(PSG) techniques (see Appendix A), a few clinical observations were published that described
some of the profound psychological and physiological disturbances associated with this yet
unrecognized syndrome.

Guilleminault et al: Arch Intern Med 1977

Among these early publications, Christian Guilleminault and co-authors from Stanford
University, published a paper in Archives of Internal Medicine in 1977 describing the clinical
features of 25 patients with ―sleep apnea syndrome‖ (3). These patients were referred to the
Sleep Disorders Clinic for either excessive daytime sleepiness or combination of loud snoring
with hypertension, headaches, or abnormal behavior during sleep. All patients were male with
age ranging from 25 to 65 years (mean 44.3 years) and a PSG exam determined that they all
suffered from ―obstructive‖ sleep apnea.

Using the 1959 Metropolitan Life Insurance Company statistical tables to standardize by age and
height, the weight distribution of these patients was as follows:

             Table 1 – Distribution of relative weight categories of sleep
             apnea patients (n=25) according to 1959 Metropolitan Live
             Insurance Company standards; Guilleminault et al, 1977
              Weight category                               Number
              Normal                                        5 (20%)
              5% to 15% Overweight                          4 (16%)
              16% to 39% Overweight                         8 (32%)
              40% to 100% Overweight                        8 (32%)

The following were among the symptoms and signs reported in the manuscript:

-   Loud snoring (occasionally interrupted by silences of 20 seconds or longer)
-   Excessive daytime sleepiness (sudden, excessive drowsiness at inappropriate times of day)
-   Personality changes (including depression, anxiety, irritability, hostility, as reported by
    family members)
-   Morning headaches (frontal and occasionally diffuse)

                                           4 of 18
BMI 546                                  Gangnon                                       July 2, 2010

-   Hypertension (systolic blood pressure higher than 150 mm Hg or diastolic blood pressure
    higher than 95 mm Hg)

Table 2 presents the occurrence of each of these clinical characteristics (with ‗+‘ indicating
presence and ‗–‘ indicating absence) in the 25 patients, as reported in the manuscript.

     Table 2 – Symptoms in sleep apnea patients (n=25); Guilleminault et al, 1977
                           Excessive Personality      Morning
       Patient   Snoring sleepiness     changes      headaches Hypertension
          1         +          +            +            +              +
          2         +          +            +            +              +
          3         +          +            +            -              +
          4         +          +            +            -              +
          5         +          +            +            -              +
          6         +          +            +            +              -
          7         +          +            +            -              -
          8         +          +            +            +              -
          9         +          +            +            -              +
         10         +          +            +            -              +
         11         +          +            +            +              _
         12         +          +            +            +              _
         13         +          +            +            +              _
         14         +          +            +            -              _
         15         +          +            -            +              +
         16         +          +            -            +              +
         17         +          +            -            -              +
         18         +          +            -            -              +
         19         +          -            -            +              +
         20         +          -            +            +              -
         21         +          +            -            -              +
         22         +          +            +            -              -
         23         +          -            -            -              +
         24         +          -            -            -              -
         25         +          +            -            -              -

Question 1: Calculate the estimated prevalence of each of these five clinical characteristics in
this patient population. Find a 95% confidence interval for the true population prevalence of
each of these five clinical characteristics based on these data.

                                            5 of 18
BMI 546                                Gangnon                                    July 2, 2010

Question 2: The prevalence of hypertension in the general adult (18-74) male U.S. population
based on the 1976-1980 National Health and Nutrition Examination Study was 33%. How
does the prevalence of hypertension in these patients with sleep apnea compare to the
prevalence in the general population?

Question 3. Would you conclude from these data that an association exists between sleep
apnea and hypertension? Why or why not?

In the ―Comments‖ section of the manuscript, Guilleminault et al wrote ―Our Sleep Disorders
Clinic has an obvious bias because most patients are referred for a ‗sleep disorder‘.‖

Question 4: What do you think the authors are referring to with this statement? What other
types of bias may have affected these data?

                                          6 of 18
BMI 546                                        Gangnon                                              July 2, 2010

Guilleminault et al conclude their manuscript describing the diverse therapeutic interventions in
these 25 patients, including diet, medications, and surgery; 5 patients refused treatment. Surgical
treatment included adenoidectomy, resection of part of the soft palate or, most frequently (8
cases), tracheostomy and positioning of a permanent tracheal valve. The latter was reported as
the most effective treatment in this series.

Lavie et al: Am Heart J 1984

In a subsequent study conducted in Israel, Lavie et al examined the prevalence of sleep apnea
among patients with hypertension (4). The series included 50 patients (10 women and 40 men)
attending Ramban University Hospital outpatient hypertension clinic that were diagnosed with
―essential hypertension‖ (blood pressure >160 mm Hg systolic and 95 mm Hg diastolic with no
known cause of hypertension). According to the authors, ―all patients approached agreed to be
interviewed; this precludes the possibility that the sample of 50 patients included a
disproportionate number of patients with more severe sleep disorders, who expected some
benefits from the sleep recordings.‖

Question 5: What kind of bias are the authors concerned about? Do you agree that the 100%
participation eliminates such bias?

Among the 50 patients that were interviewed, 16 (12 men and 4 women) were selected for a PSG
examination based on the presence of at least three of the following complaints: 1) excessive
daytime sleepiness; 2) loud snoring; 3) excessive motility in sleep; 4) multiple awakenings from
sleep; 5) frequent headaches; and 6) chronic fatigue. Sleep apnea was diagnosed based on the
―apnea index (AI),‖* the average number of apneas (lasting more than 10 seconds) per hour of

Among the 16 patients who underwent PSG examination, the mean AI was 21.6 (SD, 18.4).
Thirteen of the 16 patients had AI>5, 11 patients had AI>10, and 8 patients had AI>30. In the
Discussion section of the paper (page 375), the authors state that the estimated prevalence of

 In the early years of research on this subject, hypopneas were often not considered and thus the Apnea Index was
commonly used instead of the previously described Apnea-Hypopnea Index.

                                                   7 of 18
BMI 546                                 Gangnon                                      July 2, 2010

sleep apnea (defined as AI>5) ―is 26% of the initial sample of 50 patients [which] is considerably
higher than the estimate of 1.26% of sleep apnea syndrome in men older than 21.‖

Question 6: How was the 26% estimate calculated? Do you think this represents an accurate
estimate of the prevalence of sleep apnea among hypertensives? Why or why not?

Question 7: Find a 95% confidence interval for the prevalence of sleep apnea among
hypertensives. How does the prevalence of sleep apnea among hypertensives compare to the
prevalence in the general population?

Question 8: Do these results help you conclude that sleep apnea and hypertension are
associated? Why or why not?

                                           8 of 18
BMI 546                                  Gangnon                                       July 2, 2010

PART 3: Observational epidemiologic evidence

More rigorous observational studies on this topic were published between the mid 1980s and the
1990s. These studies varied widely in terms of study population and study design. This section
describes some salient results emerging from a few of these studies.

Norton and Dunn: Br Med J 1985

In a study conducted in four family practices in Toronto, Canada (5), 2001 subjects visiting these
practices (about 85% of those approached) agreed to respond to a questionnaire that included
basic demographic data about themselves and questions on snoring and medical conditions in
members of their households. The definition of snoring was ―a general one of noise produced
while sleeping, but it was left to the reporters to interpret this as they saw fit.‖ Medical records
were not reviewed. The 2001 reporters (691 of whom were men) provided data on a total of
2,629 subjects (1,411 men, 1,211 women, and seven with no recorded gender); the prevalence of
snoring in this group was 42%. Table 3 displays the reported prevalence of several medical
conditions and characteristics according to snoring history.

     Table 3 – Demographic characteristics and medical conditions according to
     snoring (values are percentages); Norton & Dunn, 1985
                                                           Those who     Those who
                                              Occasional snore nearly snore every
      Condition         All     Non-snorers      snorers   every night      night
                     (n=2629)     (n=1379)      (n=638)     (n=213)       (n=254)

      Male sex            53.7          45.4            64.2           69.0           73.2
      Overweight           7.2           2.3             8.9           16.9           22.8
      Smoking             17.9          10.3            24.0           32.9           33.9

      Depression          0.8           0.2             1.3            2.8             2.0
      Asthma              2.9           2.0             5.0            3.8             2.0
      Diabetes            1.5           0.8             2.0            2.3             3.5

      Heart disease       4.2           1.7             5.5             9.9           11.8
      Hypertension        6.2           2.3             8.8            12.2           18.5

Question 7: Using „non-snorers‟ as the reference category, calculate the prevalence odds ratio
of hypertension and associated 95% confidence interval for each of the other snoring
categories. Describe and interpret your findings. (Please, comment on how the prevalence
varies according to increasing snoring frequency.)

                                              9 of 18
BMI 546                                Gangnon                                     July 2, 2010

Question 8: Describe the main limitations of the preceding analysis.

Because the above conditions occur more frequently in older individuals, Norton & Dunn also
conducted analyses restricted to individuals older than 40 years of age. The following table
displays results from more detailed analysis of hypertension prevalence in this older group.

     Table 4 – Prevalence of hypertension (in %) in study participants ≥40 years old
     according to snoring, stratified by gender, age and other characteristics; Norton
     & Dunn, 1985
                                            Men                      Women
                                   Snorers    Non-snorers     Snorers     Non-snorers
          40-49 y                      7            0            12              0
          50-59 y                     22           13            10             12
          60-69 y                     28           14            15              7
          70-79 y                     27           13            40             29
          80-89 y                     12           11            40             40

      Non-smokers, non-obese          19               8          20              7
      Smokers (non-obese)             15              11          10             37
      Obese (non-smokers)             29              12          39             22
      Smokers and obese               35               0          33              0

Question 9: Why do you think the authors showed these stratified data?

                                           10 of 18
BMI 546                                 Gangnon                                     July 2, 2010

Question 10: Calculate the prevalence odds ratio for hypertension within each strata. Are the
prevalence odds ratios consistent across strata? Is there any additional information that you
would need to answer this question more definitively?

Question 11: Was this a representative sample of the population? Why or why not? Is it
essential for the purpose of the study that the sample be a fully representative one? Why or
why not?

Question 12: Compared to the studies described in the preceding section, what features of this
study add to the evaluation of a possible association between sleep apnea and hypertension?
What are the main limitations affecting the conclusions from this study?

Fletcher et al: Ann Intern Med 1985

In another study conducted at the Houston Veterans Administration Medical Center, Fletcher et
al examined the prevalence of undiagnosed sleep apnea in patients with essential hypertension
and controls (6). The study population consisted of 46 men with essential hypertension (systolic
blood pressure >140 mm Hg, or diastolic >90 mm Hg if age <45 years and >95 mm Hg if age
>45 years, with no identifiable renal or endocrine abnormalities) and 34 normotensive men as
controls. Sleep apnea was defined as more than 10 apneas per hour of sleep measured by PSG.

Hypertensive men were recruited from the hypertension, medical, and dermatologic clinics and
from hospital employees; normotensive controls consisted of outpatients with minor
dermatologic problems or healthy hospital employees. According to the authors, ―men were

                                          11 of 18
BMI 546                                  Gangnon                                       July 2, 2010

selected without bias to physical habitus, except that efforts were made to recruit control and
hypertensive persons of equivalent age and weight.‖

Question 13: What type of study design was this? Why were the investigators concerned
about age and weight? Were these concerns properly addressed?

The following table presents a comparison of characteristics of hypertensive patients and

   Table 5 – Comparison of characteristics of hypertensive patients and controls;
   Fletcher et al, 1985*
                                        Controls     Hypertensives     P-Value

    Number                                      34              46

    Age, years                                52.4±1.5        53.9±1.2
    Percentage of ideal body weight          111.9±3.0       117.2±3.9
    Apnea index                                3.3±0.7        10.0±2.3

     * Values are mean and standard error.

Question 14: Calculate p-values for the comparisons of age, percentage of ideal body weight
and apnea index between controls and hypertensives. Interpret the results..

                                           12 of 18
BMI 546                                 Gangnon                                                       July 2, 2010

Figure 1 represents the apnea index in the
46 hypertensive and 34 control patients as
presented in the published paper.

Question 15: Based on the data shown in
figure 1, calculate the odds ratio of sleep
apnea comparing hypertensive patients and
controls along with the associated confidence




                                                        f apneas per hour of sleep for hypertensive subjects (n=46) and
                                                        controls (n=34). Horizontal line at 10 indicates the level above
                                                        which the person is considered to have sleep apnea. Different
                                                        bullet shapes in the hypertension group indicates the type of
                                                        antihypertensive regiment in each patient.

Based on these observations, Fletcher et al concluded that ―undiagnosed sleep apnea syndrome
may be associated with systemic hypertension in many middle- and older-aged men. In some,
sleep apnea syndrome could be the cause of hypertension, and in others it may contribute to
hypertension of another cause.‖

Question 16: How does this study add to the evaluation of a possible association between
sleep apnea and hypertension? Are you persuaded by the authors‟ conclusions regarding
“causality”? What are the limitations affecting the conclusions from this study?

                                             13 of 18
BMI 546                                    Gangnon                                    July 2, 2010

Nieto et al: JAMA 2000

The Sleep Heart Health Study (SHHS) conducted a PSG exam in participants in several ongoing
federally-funded cohort studies of cardiovascular disease (7). Participants in these cohort studies
had been recruited from the community in 10 field centers across the US and had been subject to
extensive evaluation of cardiovascular risk factors, including repeat blood pressure
measurements. A total of 6,123 members of these cohorts underwent a full unattended overnight
PSG exam in their home using a portable PSG monitor. Based on theses recordings, sleep apnea
was defined based on the average number of apneas or hypopneas (partial apnea) per hour of
sleep (the Apnea-Hypopnea Index, AHI). Table 6 presents the number of SHHS participants in a
cross-tabulation of AHI and hypertension categories (hypertension defined based as systolic
blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, or use of antihypertensive
medication, at the cohort exam closest to the time when the PSG exam was done.

Question 17: What type of study design was this?

       Table 6 – Hypertension according to apnea-hypopnea index (AHI)
       categories, the Sleep Heart Health Study; Nieto et al, 2000
              AHI*                Yes               No             Odds ratio

               <1.5                 738                 953            1.0 (Reference)

              1.5-4.9               835                 763              _________

             5.0-14.9               1051                700              _________

              15-29.9               451                 268              _________

                ≥30                 254                 119              _________

       *See text for definitions.

Question 18: Using the lowest AHI category (<1.5) as reference for comparison, calculate the
odds ratio of hypertension for all the other categories along with the corresponding 95%
confidence intervals.

                                            14 of 18
BMI 546                                        Gangnon                                              July 2, 2010

There are several different definitions for ―confounding.‖ Generally it occurs when a variable,
related to the exposures and outcome of interest in an observational study, is not properly
accounted for (by statistical adjustment, stratification or matching), thus leading to an incorrect
association to be measured between exposure and outcome. Empirically, confounding is
detected when, after adjusting for a potentially confounding variable, the association (odds ratio,
relative risk, correlation, etc.) between exposure and outcome substantially changes (e.g. the
association is attenuated). Table 7 shows the odds ratio of hypertension adjusting for
demographic characteristics (age, sex, and race) as well as body mass index (BMI) † as reported
in the original manuscript (7).

        Table 7 – Crude and adjusted odds ratio of hypertension according to
        apnea-hypopnea index (AHI) categories, the Sleep Heart Health Study;
        Nieto et al, 2000
                                          Odds ratio of Hypertension*
             AHI*         Unadjusted (from Adjusted for age,       Adjusted for age,
                           previous table)      sex, and race     sex, race, and BMI*

               <1.5           1.0 (Reference)          1.0 (Reference)          1.0 (Reference)

              1.5-4.9            _________                    1.25                     1.12

             5.0-14.9            _________                    1.57                     1.28

             15-29.9             _________                    1.73                     1.32

                ≥30              _________                    2.27                     1.60

        *See text for definitions.

Question 19: Place the odds ratios you calculated in the previous question in the first column
of Table 7. Compare these odds ratios with the adjusted odds ratios in the other two columns.
Interpret the differences you observe.

 Body mass index (BMI): a measure of body weight relative to height, calculated as the ratio of weight (in
kilograms) divided by the square of height (in meters). BMI≥25 is usually considered to indicate ―overweight;‖ a
BMI≥30 is used to define ―obesity.‖

                                                  15 of 18
BMI 546                                  Gangnon                                          July 2, 2010

Question 20: What are the potential strengths of this study as compared with the previous
ones? What are its main limitations?

Peppard et al: N Engl J Med 2000

The Wisconsin Sleep Cohort Study is an ongoing prospective study of the occurrence, causes
and consequences of sleep disorders (particularly sleep apnea) run by Professor Terry Young of
the UW School of Medicine and Public Health. The study follows >1000 men and women
selected from a working population in the late 1980s when participants were between 30 and 60
years old. Measures of PSG (attended, in-laboratory), blood pressure (average of three seated
blood pressure readings), body habitus, and a variety of health history and behavior data are
assessed at 4-year intervals. Peppard et al (8), described the results of analyses that examined the
association of sleep apnea at baseline studies and the subsequent risk of developing hypertension
4 years later. Table 8 summarizes the findings:

Table 8 – Hypertension status at 4-year follow-up by baseline sleep apnea category among
515 baseline normotensives; the Wisconsin Sleep Cohort Study, Peppard et al, 2000
Baseline                                       Risk of
             Normotensive Hypertensive                     Absolute
sleep                                       hypertension               Relative     Odds
                at 4-year      at 4-year                     Risk
apnea                                         at 4-year                 Risk        Ratio
               follow-up      follow-up                   Difference
category                                     follow-up
                   121              15
AHI=0                                             ______        0.0 (ref.)   1.0 (ref.)     1.0 (ref.)
                   260              48
0<AHI<5                                           ______        _______       ______         ______
                    37              18
5<AHI<15                                          ______        _______       ______         ______

AHI>15              11               6            ______        _______       ______         ______

Question 21: Calculate the risk of developing hypertension at 4-year follow-up for each
baseline sleep apnea category. Then, calculate the absolute risk difference, the relative risk
and the odds ratios of developing hypertension for each of the 3 higher sleep apnea categories
relative to participants with baseline AHI=0. Write your results in Table 8.

                                           16 of 18
BMI 546                                     Gangnon                                   July 2, 2010

Because of the potential for confounding, analyses controlling for additional variables were
conducted. The results are shown in table 9.

Table 9 – Relative risks for baseline sleep apnea category predicting hypertension at 4-year
follow-up; potential confounding factors (“adjustment variables”) are progressively added
to the model; the Wisconsin Sleep Cohort Study, Peppard et al, 2000
                                                Age, sex, and body       Age, sex, body
 Adjustment variables
                             Age and sex       habitus (BMI, neck &    habitus, alcohol &
       (cumulative)
                                                    waist girth)         smoking habits
Sleep apnea category        Relative Risk          Relative Risk          Relative Risk

AHI=0                          1.0 (ref.)               1.0 (ref.)              1.0 (ref.)
0<AHI<5                           1.6                      1.3                     1.3
5≥AHI<15                          2.5                      1.7                     1.7
AHI≥15                            3.9                      2.3                     2.3
P value for trend               0.001                     0.03                    0.03

Question 22: Examining the results in Table 8, after adjusting for age and sex, which
additional adjustment variables (body habitus—BMI, waist & neck girth; or alcohol/cigarette
use habits) most clearly confound the association between sleep apnea and hypertension?
How do you know?

Question 23: How do the findings from Peppard et al (2000) affect your judgment of the
association of sleep apnea and hypertension?

                                             17 of 18
BMI 546                               Gangnon                                     July 2, 2010


1.    Caples SM, Gami AS, Somers VK. Obstructive sleep apnea. Ann Intern Med

2.    Young T., Palta M., Dempsey J., Skatrud J., Weber S., Badr S. The occurrence of sleep-
      disordered breathing among middle-aged adults [see comments]. N Engl J Med

3.    Guilleminault C, Eldridge FL, Tilkian A, Simmons FB, Dement WC. Sleep apnea
      syndrome due to upper airway obstruction: a review of 25 cases. Arch Intern Med

4.    Lavie P, Ben-Yosef R, Rubin AE. Prevalence of sleep apnea syndrome among patients
      with essential hypertension. Am Heart J 1984;108:373-6.

5.    Norton PG, Dunn EV. Snoring as a risk factor for disease: an epidemiological survey. Br
      Med J (Clin Res Ed) 1985;291:630-2.

6.    Fletcher EC, DeBehnke RD, Lovoi MS, Gorin AB. Undiagnosed sleep apnea in patients
      with essential hypertension. Ann Intern Med 1985;103:190-5.

7.    Nieto FJ, Young TB, Lind BK, et al. Association of sleep-disordered breathing, sleep
      apnea, and hypertension in a large community-based study. Sleep Heart Health Study.
      Jama 2000;283:1829-36.

8.    Peppard PE, Young T, Palta M, Skatrud J. Prospective study of the association between
      sleep-disordered breathing and hypertension. N Engl J Med 2000;342:1378-84.

                                        18 of 18

To top