Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
15 Sleep Disturbance Study Plan The following paragraphs describe the two suggested research protocols. These are followed by the procedures related to the selection of measurement sites, the selection of study participants, acoustical measurements, and supplementary data gathered will be outlined. These procedures do not differ between protocols #1, Polysomnography and #2, Actigraphy plus ECG. 15.1 The Two Protocols 15.1.1 Research Protocol #1: Polysomnography This protocol will facilitate polysomnography (i.e., the simultaneous measurement of the electroencephalogram [EEG â brain activity], electrooculogram [EOG â eye movement], and electromyogram [EMG â skeletal muscle tone]) for the measurement of sleep. According to specific conventions (International 10-20-system), electrodes are attached to the scalp and the skin of the face of the subject. The electrical potentials generated by the brain, chin muscles and eye movements are amplified, converted into digital signals and stored on digital media. The signals are later analyzed by trained personnel according to specific conventions [4,5]. Polysomnography is considered the gold standard for measuring sleep, and it permits detection of subtle changes in sleep physiology induced by aircraft noise. The use of polysomnography will assure comparability to a series of studies on the effects of aircraft noise and rail traffic noise on sleep performed in the European Union.[6,7] At the same time, due to the high methodological expense, at a given level of funding, it will be possible to investigate only a single or a limited number of airports relative to research protocol #2, and the external validity will thus be limited. The advantages and disadvantages of polysomnography are summarized here (see Basner et al. [2]): Advantages of Polysomnography: Polysomnography is the gold standard for measuring sleep, the evaluation of sleep structure and the degree of sleep fragmentation. It is a method that covers most physiological aspects of sleep (with the exception of conscious awakenings, as we cannot tell with certainty from the polysomnogram whether a subject regained waking consciousness or not). It is a very sensitive method that will detect even subtle changes in sleep physiology. Also, the method itself is very well standardized. Disadvantages of Polysomnography: EEG, EOG, and EMG electrodes and leads are somewhat disruptive, may influence sleep, and thus at least one night is usually required for adaptation[8]. The measurement instruments are expensive and fragile. The instrumentation and de-instrumentation of subjects is cumbersome and has to be done by trained personnel. EEG and EMG electrodes are sometimes affected by movements or excessive sweating of the subjects, which may render the analysis of (part of) the data gathered during the night impossible. Finally, sleep stage classification requires trained personnel and is known to have high inter- and intra-observer variabilities [9,10,11]. Automated sleep stage classification systems exist, but so far validation studies reached contradictory conclusions [12]. 15.1.2 Research Protocol #2: Actigraphy plus ECG This protocol will facilitate the simultaneous measurement of actigraphy (skeletal muscle movement) and heart rate (ECG). Actigraphs measure acceleration of body movements (in one or more dimensions), are the size of a watch, and are worn like wrist-watches (usually on the wrist of the non-dominant arm). Some products have additional features, for example, light sensors measuring environmental light intensity (sometimes in different spectra), body position sensors, an event marker button (e.g., to signal lights out), or a display (e.g., for displaying clock time).Actigraphy is a well-established method in research on the effects of aircraft noise on sleep. It was used in studies around Heathrow [13], Amsterdam [14], and Cologne-Bonn Airport [6]. Therefore, using actigraphy ensures comparability of the results of a US field study with those of the above mentioned European studies. It is suggested that higher data storage rates 42
are used than those commonly applied (1-2 samples/min) in order to allow for an event-related analysis. Newer equipment can continuously sample and store raw data at 30 -100 Hz for several days to weeks. The advantages and disadvantages of actigraphy are summarized here (see Basner et al.[2]): Advantages of Actigraphy: Actigraphs are inexpensive and comparatively robust. After an initial orientation, subjects can wear the device for several days and nights unsupervised (i.e., the methodological expense is low). The movement activity data gathered with actigraphy are the measure of interest, so there is no need to visually score data. Actigraphs are less disturbing than the sensors applied for polysomnography, and it is unlikely that actigraphs substantially influence normal sleep. Disadvantages of Actigraphy: Although actigraphs are an accepted measure to determine rest- activity cycles [15], more subtle physiological changes cannot be detected by actigraphy. Unfortunately, the degree of standardization overall is relatively low. Different models (i.e., hardware) will give slightly different results, there are several methods to determine activity counts (time above threshold, zero crossing, digital integration)[15], and each company has its own algorithm to differentiate wake from sleep periods. Therefore, it is not surprising that the results of comparisons between polysomnography and actigraphy vary widely [15-21]. Although CNS activations and body movements often occur simultaneously, both may occur independently from each other, and thus one cannot expect a 1:1 agreement. Rather, some misclassifications are obvious: For example, someone lying awake and not moving but trying to fall asleep would be misclassified as being asleep by actigraphy. The ECG offers a unique opportunity to measure both subtle and more obvious changes in sleep physiology with less disruptive and less expensive methods than polysomnography. Self-instrumentation and automatic data analysis make this an inexpensive and objective method. Nocturnal vegetative activations may play an important role in the genesis of cardiovascular disease, and therefore the analysis of heart rate information alone delivers important insights. An ECG-based algorithm for the automatic identification of cortical arousals was developed [22] and validated [23] by Basner et al. This algorithm is currently adapted (to better match EEG awakenings) and extensively validated with polysomnographic data gathered around Frankfurt airport within PARTNER Project 25B. The current version of the algorithm shows almost perfect agreement with EEG awakenings (kappa > 0.8). This methodology will deliver meaningful data while being much more cost-effective. Therefore, it will be possible to investigate several US airports at the same level of funding that would be needed to measure a single airport with polysomnography. Preferentially, both actigraphy and the ECG will be recorded with the same device. This avoids data synchronization problems. The advantages and disadvantages of the ECG follow (see Basner et al.[2]): Advantages of ECG: Similar to actigraphy, devices measuring the ECG are relatively inexpensive and robust. After an initial orientation, subjects can attach and detach the ECG electrodes themselves and (depending on storage capacity) can wear the device for several days and nights unsupervised (i.e., the methodological expense is low). The data are scored automatically by the algorithm described above, so there is no need to visually score data. The ECG is less disruptive than the sensors applied for polysomnography, and it is unlikely that the ECG alone substantially influences normal sleep. Repeated noise induced autonomic activations may play a key role in the genesis of hypertension and associated cardiovascular diseases, and therefore measuring autonomic activations may be an advantage from a conceptual standpoint. In the recent past, the utility of specific aspects of the ECG signal (like heart rate variability [24] or cardiopulmonary coupling [25]) for sleep research has been acknowledged in the field. For this reason alone it will be worthwhile to sample the ECG in a field study on the effects of noise on sleep. Disadvantages of ECG: A certain period throughout the night is spent awake, and it is unclear how to interpret heart rate increases during wakefulness (the same is true for actigraphy, see above). Basner 43
et al.[23] discuss this the following way: "Situations where the subject was already awake before playback of the ANE started (10.3% of all events) were excluded from the analysis in this study â¦. Comparable to actigraphy, the ECG algorithm is not able to differentiate between wake and sleep unless polysomnography is performed simultaneously. If the ECG is sampled alone, cardiac activations during wakefulness may be misinterpreted as awakenings, potentially overestimating the number of traffic noise induced awakenings. However, in situations where the subject is already awake traffic noise may nevertheless adversely affect sleep by preventing the subject from falling asleep again, and therefore prolonging spontaneous or noise induced awakenings [26]. In these situations, noise induced cardiac activations may indicate an increased state of arousal and, therefore, a decreased likelihood of falling asleep again. Hence, although cardiac activations during wake periods may overestimate the number of EEG awakenings, they may nevertheless be a useful indicator of noise induced sleep disturbance. Further analyses on the association of cardiac activations during wakefulness and the time needed to fall asleep again should be performed in the future." 15.2 Measurement Sites A US field study on the effects of aircraft noise on sleep should be performed at least at one airport with nocturnal air traffic and one control airport without aircraft noise exposure. Generalizability of the findings and exposure-response relationships will increase with the number and representativeness of the airports studied. An approach similar to the one recently described for the FAA sponsored aircraft noise and annoyance study could be adopted. For that project, HMMH has been asked to investigate twenty airports, including at least one from each of the eight FAA Regions located within the contiguous United States. It would also be valuable to investigate airports with traffic curfews or ones that have recently experienced a significant increase or decrease in traffic volume (e.g., opening of a new runway), but including such changes is not necessary for the success of the project. Runway use depends on wind direction. Sites that are exposed to aircraft noise independent of wind direction (i.e., either by aircraft taking off or by aircraft approaching) will be preferentially chosen as measurement sites (as opposed to sites that are only exposed to aircraft noise under certain wind conditions). This assures that a high number of noise events per subjects will be measured (which increases the statistical power of the study and the precision of the exposure-response relationship), and that the likelihood of subjects receiving no aircraft noise at all during a measurement period decreases. The choice of study regions around the airport will reflect varying degrees of aircraft noise exposure (i.e., high exposure regions in close proximity to the runways and low exposure regions farther away from the runways). Study regions will be identified using Lnight contours using the INM or equivalent. The number of aircraft contributing to Lnight and the expected maximum sound pressure level LAS,max at the exposure site will be used as supplementary criteria for site selection. Subjects will be sampled in equal parts from regions with Lnight > 55 dB (high degree of sleep disturbance according to WHO,27 more than one additional awakening per night according to Basner et al.28) and from regions with Lnight between 40 dB and 55 dB (moderate degree of sleep disturbance according to WHO,27 less than one additional awakening per night according to Basner et al.28). Subjects living in regions with Lnight <40 dB will be ineligible for study participation, as no relevant degree of aircraft noise-induced sleep disturbance is expected. 44
Figure 12 Lnight contours 40 to >55 dB, based on historical data from Newark International Airport. The sampled noise regions will also be classified according to additional criteria (e.g., average family income per household) on zip-code level. Control sites will be selected to reflect the distribution of these additional criteria at the exposure sites. Measurement site selection will also ensure that exposure to road and rail traffic noise at the control sites is comparable to that at the exposure sites. For many communities, maps showing traffic noise exposure levels already exist. If not, the degree of traffic noise exposure can be estimated from the following variables: type of road (number of lanes, cul-de-sac, etc.), distance of most-strongly exposed façade from the roadway/railway, and bedroom window facing road/rail (yes/no). For final subject recruitment blocks within designated exposure and control areas will be randomly selected. A selected aircraft noise exposure block will be matched to a selected control block that has similar road traffic noise. Recruiters will then go from door to door within each block and leave flyers to recruit subjects. 45
15.3 Assessment of the Consequences of Aircraft Noise-Induced Sleep Disturbance In order to minimize methodological expense (and thus maximize response rates and generalizability of results), the assessment of the consequences of aircraft noise-induced sleep disturbance will be restricted to brief morning questionnaires (see Appendix I). However, as sleep fragmentation has been shown to cause transient increases in blood pressure during the night, which may over time contribute to more long-term effects, blood pressure will be measured twice (once during the distribution and a second time during the collection of the measurement equipment). Based on prior field studies on the effects of aircraft noise on sleep, a relevant change in cognitive performance due to the aircraft noise exposure is not expected, and thus the increase in methodological expense due to cognitive performance tests is not justified. 15.4 Assessment of the Acoustical Environment It is suggested that actual sounds inside the bedroom are continuously recorded along with noise levels with class-1 noise level meters. Also, it is suggested that the study be done in cooperation with the airport so that detailed information on flight operations with a high temporal resolution can be collected. The combination of interior sound recordings and flight operations data should be sufficient for the identification of aircraft noise events. If flight operations data cannot be obtained, the recording of outdoor sounds may be necessary to correctly identify aircraft noise events. If simultaneous measurements are being conducted at sites that are within a close vicinity it may be sufficient to record outdoor sounds at one central site. If this is not possible, outdoor measurements at each site should be made. 15.5 Data Synchronization It is suggested that actigraphy and the ECG be recorded with the same device. If it is not feasible to use wireless technology, the internal clocks of all measuring devices should be synchronized immediately before the start of the measuring period and the data corrected for the time drift of each individual device (that would be established before the start of the study), in order to assure synchronization between acoustical and physiological variables. 15.6 Assessment of Non-Acoustical Extrinsic Factors Influencing Sleep Temperature, humidity and light intensity as potential confounders should be continuously measured in the bedroom during the field study, in order to be able to control for the effects of these variables on sleep in the statistical analysis. Sampling should be alternated between exposure and control sites, so that exposure and control groups will be measured during the same season of the year. 15.7 Subject Selection Criteria and Sample Size 15.7.1 Selection Criteria As few selection criteria as possible will be used in order to increase response rates and the generalizability of results (but with representation of both sexes and a wide age range). However, it will be possible to adjust for some of the standard selection criteria in the analysis phase of the study. The following eligibility criteria should be applied: ï§ Subject is at least 21 years old. ï§ Subject does not use hearing aids during the day or ear plugs during the night. ï§ Subject understands and is able to speak/write the English language. 46
ï§ Subject has no active alcohol or drug addiction. ï§ Subject has no history of cardiac arrhythmia (ECG algorithm not validated in arrhythmia). ï§ Subject has no history of and is not treated for obstructive or central sleep apnea. ï§ Subject does not consume sleeping medication on a chronic basis (more than twice per week). ï§ Measurement equipment can be securely stored in the subject's home. After subjects have been selected for the study, they will fill out a general questionnaire (see 0), a Health Survey (SF-36),29 the Pittsburgh Sleep Quality Index (PSQI),30 and the Horne-Ostberg Morningness- Eveningness Questionnaire31 to determine their circadian preference. 15.7.2 Sample Size Calculations The power of the study and the precision of the exposure-response relationship depend on both the number of investigated subjects and the expected cumulative number of noise events per subject. The latter will depend on the traffic volume at the study site. Therefore, at busy airports it may be sufficient to investigate subjects for a single or a few nights, whereas at airports with low traffic volumes or traffic curfews it may be necessary to measure for several nights. Different combinations of "number of subjects" and "number of noise events per subject" can lead to the same power/precision, see Figure 13 below. Figure 13 shows results of Monte Carlo simulation based sample size calculations for a study on the effects of aircraft noise on sleep. Dots show simulation results and best fit regression lines are presented. The left panel shows the precision of 95% confidence intervals surrounding exposure-response relationships for response probability depending on number of investigated subjects and number of expected noise events per subject. Exposure-response relationships are based on random subject effect logistic regression analyses with maximum sound pressure level as the only explanatory variable. The right panel shows the Statistical Power 1-β (i.e., probability to detect a statistically significant effect if in reality there is an effect) of a study on the effects of aircraft noise of maximum sound pressure level on awakening probability depending on number of investigated subjects and number of expected noise events per subject. This work was performed within PARTNER Project 25B. 47
Figure 13 Effects of Subject and Cumulative Event Numbers on Confidence Interval and Statistical Power For this proposed study, the preference should be given to increase the number of subjects, as we are more interested in getting precise information on a representative group of subjects than very precise information on a smaller group of subjects. The investigated number of subjects and number of nights per subject should be chosen in a way that at least 80% power is achieved even with some attrition or a lower than expected number of noise events per night. If a single airport and a single control site are investigated, we suggest sampling at least 40 subjects per site (better: 60 subjects) with a target of gathering, on average, reactions to 60 aircraft noise events per subject at the aircraft noise exposure site. This would result in an average precision (i.e., width of the 95% confidence interval surrounding the exposure-response function) of 4.3% and a power of 80.5% to detect a statistically significant effect (3.5% and 91.7% for 60 subjects, respectively). If more than one airport is investigated, the power for the pooled data to detect a significant aircraft noise effect will be >99.9% and the precision for the pooled exposure-response function will be â¤3%. If individual airports are of lesser interest and the focus is on the pooled exposure-response function, sample sizes at individual airports could be further reduced. 48
