lntraindividual Peak Flow Variability*

Transcription

1 lntraindividual Peak Flow Variability* Matthew]. Hegewald, MD; Robert 0. Crapo, MD, FCCP; and Robert L. jensen, PhD Objectives: To quantify intraindividual variability in peak expiratory flow (PEF) measured with peak flowmeters and to define factors affecting PEF variability. Methods: Three hundred one healthy subjects (aged 4 to 84 years) were recruited from sites at sea level (n=220) and at 1,400 m altitude (n=81). All testing was done with the same model peak flowmeter. Each subject was actively coached through five to eight successive PEF maneuvers. Three meters of the same model were tested using a mechanical waveform simulator at three different flows at both testing altitudes (sea level and at 1,400 m). Results: Excluding outliers, the mean PEF was 523 L/min, mean standard deviation (SD) was 22 L/min, and mean coefficient of variation (CV) was 4.6%. The upper 95th percentile for CV was 8% for adults and 10% for youths. Analyzing only the three highest peak flows for each subject, the mean PEF was 539 L/min, mean SD was 12 L/ min, and mean CV was 2.4%. The upper 95th percentile for CV was 6% for adults and 9% for youths. Linear regression analysis revealed a small but statistically significant correlation (p<0.01) between mean peak flow and CV. In adults, SD correlated with sex (p<0.01) but neither CV nor SD was correlated with age, height, weight, or altitude. Meter variability defined with the mechanical waveform simulator was small. Standard deviation varied from 1.5 to 4.2 L/min and CV from 0.4 to 1.6%. When the three largest peak flows for each subject were used, 5.5% of intraindividual variance was explained by meter variance. Conclusions: These estimates of intraindividual variability in healthy subjects are generally lower than those previously reported. Meter variability accounts for only a small part of total intraindividual variability. The 95th percentile data suggest that a fall in PEF of 6 to 8% in adults and 9 to 10% in youths would be statistically significant. (Chest 1995; 107:156-61) ATS=American Thoracic Society; CI=confidence interval; CV=coefficient of variation; ERS=European Respiratory Society; FEVg=forced expiratory volume in 1 s; NAEP= National Asthma Education Program; PEF=peak expiratory flow Key words: asthma; intraindividual variability; peak expiratory flow; pulmonary function peak expiratory flow (PEF), the maximum flow generated during forced exhalation, is dependent on effort, expiratory muscle strength, airway caliber, lung volume, 1 and the length of pause time at total lung capacity prior to initiation of forced exhalation. 2 Peak expiratory flow can be measured with standard spirometric equipment and with simple, inexpensive, portable devices. It correlates well with forced expiratory volume in 1 s (FEY 1). 3 The primary clinical use of PEF monitoring has been to follow the course of asthma and its response to therapy. 4-7 Several other uses of PEF have been proposed, including identification of provocative factors for asthma, 8 9 diagnosing asthma, 5.l 0 11 determining bronchodilator response,12.13 and determining steroid response in chronic obstructive pulmonary disease. 14 Because clinical decisions are often based on changes in PEF, it is important to define what changes are significant. Significant change must be greater than the sum of biologic intraindividual *From the Pulmonary Division and the Department of Internal Medicine, LDS Hospital and the University of Utah, Salt Lake City. Supported by a grant from Health Scan Inc, Cedar Grove, NJ. Manuscript received December 7, 1993; revision accepted May 17, variability and device variability to overcome the "noise" in the measurement. The National Asthma Education Program (NAEP) defines a 20% reduction in PEF from baseline as clinically significant. 4 Previous estimates of PEF intraindividual variability measured with a pneumotachograph and expressed as coefficient of variation (CY) have ranged from 5.0 to 12.9% Intraindividual PEF variability measured with a hand-held peak flowmeter is less well defined with CYs ranging from 3.8 to 8.2%_17 19, 2 0 Intraindividual variability is greater for PEF than for FEY and is greater in patients with respiratory diseases than in normal subjects Peak expiratory flow intraindividual variability has been shown not to be influenced by age, sex, height, and weight. 22 Previous estimates of intrainstrument variability for a specific PEF meter (Assess, Health Scan Inc, Cedar Grove, NJ) measured on a mechanical waveform simulator and expressed as CY ranged from 0.70 to 2.15%. 23 Reproducibility criteria for consecutive PEF measurements are not well defined. Burge 24 has proposed that the best two of at least three trials be within 20 L / min of each other. A preliminary recommendation from the European Respiratory Society (ERS) suggests that two of at least three readings must be lntraindividual Peak Flow Variability (Hegewald, Crapo, Jensen)

2 within 5% or 20 L/min of one another, whichever is greater. 25 Dahlquist et al 20 propose a less stringent criterion of two measurements within 10% of each other. The purpose of this study is to further define intraindividual PEF variability in healthy subjects tested with a portable, hand-held peak flowmeter. Other objectives were to identify factors affecting intraindividual PEF variability, to define reproducibility criteria for successive measurements, to evaluate the utility of the current NAEP recommendation of three repeated efforts as the standard approach to PEF assessment, 4 and to compare our mean PEF values with those obtained using the reference equations present in the NAEP document. 4 MATERIALS AND METHODS The study population consisted of 301 healthy volunteers between ages 4 and 84 years (141 male, 160 female) recruited for a "normal values" study from three cities located near sea level (Atlanta, mean barometric pressure 740 mm Hg; Boston, mean barometric pressure 740 mm Hg; and Hartford, Conn, mean barometric pressure 760 mm Hg) and one located at an altitude of 1,400 m (Salt Lake City, Utah, mean barometric pressure 642 mm Hg). Recruited subjects were mostly hospital employees, their families, and acquaintances. The subjects' age, height, weight, and a brief medical history were recorded. Exclusion criteria were as follows: (1) greater than 1 pack-year smoking history or a history of regular smoking in the past 6 months; (2) history of chronic cough, chronic sputum production, wheezing, or dyspnea; (3) history of prior significant heart or lung disease; ( 4) use of medications that could affect lung function; (5) history of occupational exposure to lung toxins; (6) presence of severe obesity (based on visual inspection, no weight criteria were used); and (7) presence of a thoracic cage disorder. Twelve peak flowmeters (Assess) were used at each of the four sites. The NAEP statement on technical standards for peak flowmeters requires individual meters to be accurate over their full range within ± 10% when tested on nine multiples of American Thoracic Society (ATS) waveform 24. Different meters of the same brand must be reproducible within ± 5% of each other. 26 All 48 peak flowmeters met these standards when tested on a mechanical waveform simulator by the manufacturer. Measurements were made with the subjects standing, without noseclips, and the PEF meter scale held in the vertical position. A technician enthusiastically coached each subject to blow as hard and as fast as possible. The results were read to the nearest 5 L/min. All measurements were recorded. If necessary, the technician coached the subject after each maneuver. A minimum of five measurements were made by all subjects. If the five measurements were within ±50 L/ min of each other, no further measurements were made. If this criterion was not met, additional measurements were made until the criteria were met or until a total of eight trials were completed. Disposable mouthpieces were changed after each subject. The devices were washed and air-dried at the end of each day. To quantify the variability of the PEF meter itself and to determine if instrument variability is dependent on absolute peak flow or altitude, three peak flowmeters (Assess) were tested with a mechanical waveform simulator using A TS waveform 24 as previously described. 23 Each meter was tested 10 times at target peak flows of 200 L/min, 400 L/min, and 600 L/ min both at sea level and at 1,400 m altitude. Statistical Methods The data were analyzed using statistical software packages (Abstat, Anderson-Bell Corp, Parker, Colo, and SPSS, SPSS Corp, Chicago). The average of the three highest PEF values for each subject was calculated (high three mean). Every PEF value for each subject was then divided by the high three mean and a histogram of all measurements was created and analyzed. Based on this analysis, outliers for each individual were identified as those values deviating from the high three mean by more than 20% and were removed from the data set. Outlier exclusion was important to determine an accurate measure of variability because the subjects were untrained and every trial was recorded. Using these criteria, 4.5% of the total measurements were excluded from further analysis (4.4% below 20% and 0.1 % above 20%). Every subject had three to eight acceptable measurements after outlier removal. From this modified data set, the mean, SD, and CV were calculated for each individual in two ways: first, from all PEF values remaining after outlier exclusion (mean PEFI, SDI, and CV Il and second, from the three highest PEF values for each subject (mean PEFz, SD2, and CV2). SDI was calculated in the standard manner, whereas SD2 was calculated using the small samples technique of Dixon and Massey. 27 The 95% confidence intervals for mean CV I and mean CV 2 were calculated as ± 1.96 SEM. The 95% confidence intervals for SD1 and SDz were estimated using the standard techniques described by Dixon and Massey. 27 The upper 95th percentile values for CV1, CV2, SD1, and SD2 were determined. The study population was divided into two age groups based on visual examination of mean PEF 1 vs age graphs that revealed a break point at approximately 21 years. Group 1 consisted of adults older than 20 years. Group 2 consisted of youths 20 years old or younger. The data were also analyzed after dividing the study population into four separate traditional age groups: pediatric (age <12 years), adolescent (age 12 to 17 years), adult (age 18 to 64 years), and elderly (age >64 years). Variability, expressed as CV I and CV 2, was compared between the various age groups by two-tailed independent t tests. Variability, expressed as variance (SD1 2 and SD2 2 ), was compared between the various age groups by an f test of the ratio of the variances. Regression analysis was performed to determine correlations between CV and SD measurements and mean PEF, height, weight, age, altitude, and sex. Because multiple comparisons were made, p values less than 0.01 were deemed significant. We also compared our mean PEF 2 with predicted PEF obtained using the prediction equations present in the NAEP document applied only to our sea-level subjects. 4 The prediction equations of Godfrey et al 28 were used for the subjects less than age 19 years and the prediction equations of Nunn and Gregg 29 were used for the subjects greater than age 18 years. To our knowledge, there are no published normal values for PEF measured with peak flowmeters at altitudes above sea level. From the instrument testing using the mechanical waveform simulator, mean PEF, SD, and variance were calculated for each meter at each target flow and altitude (3 meters were tested at two separate altitudes and three target peak flows). A pooled SD for each flow and altitude was calculated as the square root of the pooled variance. An overall mean peak flow was calculated for each flow at each altitude by averaging the 30 individual measurements. Average peak flows were expressed as a percentage of target peak flow to determine accuracy. The pooled variances for each flow were compared between the two altitudes for a total of three comparisons. The effect of the magnitude of peak flow on variance was evaluated by comparing the pooled variances for each of the target peak flows at a given altitude for a total of six comparisons (three at each altitude). Overall pooled variances for each altitude were also compared. All variances were CHEST / 107 / 1 I JANUARY,

3 Table!-Peak Flow Means and Intraindividual Age <21 yr PEFt. L/ min PEF2, L/min SD 1, L/ min SD2, L/ min CVt. % cv2,% Age >20 yr PEFt. L/min PEF2, L/min SDt, L/ min SD2, L/min cv],% cv2, % Variability Data* Mean (SD) (122.3) (127.3) 19.6 (9.1) 12.6 (10.7) 6.3 (2.4) 4.0 (3.5) (167.4) (170.2) 23.0 (14.3) 11.9 (13.7) 4.1 (2.4) 2.1 (2.3) Upper 95th Percentile *PEF 1 =mean p eak flow calculated from all acceptable trials; PEF2=mean peak flow calculated from the three highest trials; SD 1 =standard deviation calculated from all acceptable trials; SD2=standard deviation calculated from the three highest trials; CV 1 =coefficient of variation calculated from all acceptable trials; CV 2=coefficient of variation calculated from the three highest trials. compared using an f test of the ratio of the variances. Statistically significant differences were acepted when p<o.ol because multiple statistical comparisons were made. RESULTS lntraindividual Variability Three hundred one subjects were tested (141 male, 160 female). All were white. Eighty-one subjects were studied at 1,400 m altitude and 220 were studied near sea level. Sixty-nine subjects were younger than 21 years old, including 36 younger than 12 years old. There were 232 subjects aged 21 years or older, including 29 elderly subjects (age >64 years). Using all acceptable values, mean population PEF 1 was L /min, mean SD 1 was 22.2 L / min (95% confidence interval [CI], 20.6 to 24.1), and mean CV 1 was 4.6% (95% CI, 4.3 to 4.9%). When only the high three values for each subject were used, mean PEF2 was L / min, mean SD2 was 12.1 L / min (95% CI, 11.2 to 13.2), and mean CV2 was 2.4% (95% CI, 2.2 to 2.7%). Mean values for peak flow, indices of variability, and their upper 95th percentiles for adults and youths are listed in Table l. Variability, expressed as CV, for youths (CV1 =6.3%, CV2=4.0%) was greater than for adults (age >20 years) (CV1=4.1%, CV2=2.1%) (two-tailed independent t test p<0.001 for both CV 1 and CV 2). When variability was expressed as variance (SD 1 2 and SD 2 2 ), there was no difference between youths and adults by variance ratio testing (p>o.ol). There was no significant difference between the pediatric (age <12 years), adolescent (age 12 to 17 years), or youth (age <21 years) groups for both CV 1 and CV 2 (two-tailed independent t test p>0.01), and SD 1 2 and SD2 2 by variance ratio testing (p>o.ol). There was also no significant difference in variability, expressed as CV 1, CV 2, SD 1 2, or SD2 2, between the adult (age 18 to 64 years) and the elderly (age >64 years) groups. The mean values for PEF2, SD2, and CV2 using the pediatric, adolescent, adult, and elderly age categories by sex are listed in Table 2. Age, height, weight, and altitude did not correlate significantly with either SD1, SD2, CV 1, or CV 2 in subjects over 20 years of age (p>0.01). There was a statistically significant negative correlation of CV 1 with mean PEF 1 in both adult men and women and CV2 with mean PEF2 in adult women (p<0.01). However, these correlations were weak; the largest r 2 was 0.11 and the largest regression coefficient was % L/ min. Standard deviation in adults correlated only with sex; women had significantly lower SD1 (18.7 vs 27.8) and SD2 (9.4 vs 13.9) than men (two-tailed independent t test p<0.001 for both comparisons). The correlation of SD with sex was not explained by differences in mean PEF as SD1 and SD2 did not significantly correlate with mean PEF 1 or mean PEF2 (p>0.11 for both comparisons). Comparisons between mean PEF2 at sea level and predicted PEF using the regression equations of Godfrey et al 28 and Nunn and Gregg 29 are listed in Table 3. Mean PEF 2 differed from mean predicted PEF using the Godfrey et al equations by up to 27% and by up to 21 % using the Nunn and Gregg equations. Most subjects achieved representative values in their first three trials; 67% of subjects achieved their highest or second highest PEF value in their initial Table 2-PEF 2, SD 2, and CV2 by Sex Using Expanded Age Groups Age, yr n PEF2, L/ min SD2, L/ min CV2, % M F M F M F M F lntraindividual Peak Flow Variability (Hegewald, Crapo, Jensen)

4 Table 3-Comparison of Sea-Level Mean PEF 2 With Currently Used PEF Prediction Equations n Mean PEFz, L/ min %Predicted M F M F M F Godfrey et al 28 Age <12 yr Age yr Nunn and Gregg 29 t Age yr *Godfrey et al 28 regression equations: male PEF (L/ min)=5.288*height (cm) ; female PEF (L/ min)=5.278*height (cm) tnunn and Gregg 29 regression equations: male log. PEF (L/ min)=0.544 log. age *age-74.7/ height (cm)+5.48; female log. PEF (L/ min)=0.376 log. age *age-58.8/ height (cm) three efforts. However, 33% of subjects recorded their highest or second highest PEF in efforts 6 through 8, and 63% of subjects required more than five trials to achieve the required level of reproducibility (five values within 50 L/ min of each other). Children (age <12 years) and the elderly (age >64 years) exhibited greater learning effect with 72% of children and 93% of elderly subjects requiring more than five trials to achieve the desired reproducibility. When compared with reproducibility criteria available in the literature, 12.6% of subjects failed to meet the criteria of Burge 24 (best two values within 20 L/ min of each other), 9.6% failed to meet the ERS criteria 25 (best two values within 5% or 20 L/ min, whichever is greater), and 3.0% failed to meet the criteria of Dahlquist et al 20 (best two values within 10% of each other). Instrument Variability The 3 peak flowmeters tested on the waveform simulator were both accurate and precise (Table 4). Mean pooled PEF values were within 0.2 to 5.4% of target. The pooled SDs ranged from 1.52 to 4.23 Table 4-Instrument Accuracy and Variability Measured With a Mechanical Waveform Simulator Flow Rate, L/ min Sea Level 1,400 m p Value* 200 Mean <0.001 SD Mean/ target, % Mean <0.001 SD Mean/ target, % Mean <0.001 SD Mean/ target, % All flows SD *p values for comparison of means were determined by paired t test; 27 p values for comparisons of variance (SD 2 ) were determined using ratio of variances and f test. 27 L/ min and pooled CVs ranged from 0.37% to 1.55%. There were no significant differences (p<0.01) between the pooled variances by variance ratio testing across the two altitudes (Table 4). At a given altitude, variability was generally independent of the magnitude of peak flow; only one of the six comparisons had a p value less than 0.01 by variance ratio testing (sea level 200 L/ min target PEF compared with 600 L/ min target PEF, p=0.0037). Measured PEF was 5.3 to 6.9% lower at 1,400 m altitude than at sea level for the same peak flow generated by the mechanical waveform simulator (Table 4). This is consistent with previous reports that peak flowmeters significantly underestimate measured peak flow at elevations above sea level DISCUSSION This study was limited to a single model of peak flowmeter. However, reproducibility among several of the available brands of hand-held peak flowmeters has been shown to be less than ± 5% at flow rates greater than 100 L/ min. 23 This study was also limited to normal subjects; variance in patients with obstructive lung disease has been reported to be 64% greater than in normal subjects 21 and the percentage of change in PEF measured by spirometry required for significance in patients with obstructive lung disease has been reported to be approximately twice that of normal subjects.l 8 This may not be applicable to PEF meters used in monitoring asthma where a high frequency of measurements is likely to improve reproducibility. Only one component of overall variability was measured: the reproducibility of multiple trials measured during the same sitting on a single day. The goal was not to address diurnal variability or between-day variability, although the latter has been shown to be similar to variability from a single session in normal subjects. 15 Mean PEF and intraindividual variability were calculated first using all acceptable values and then using only the three highest values for each subject. Variability obtained by the latter method is likely to be more representative of well-trained individuals CHEST / 107 / 1 / JANUARY,

5 and the difference between the two is probably a reflection of learning effect. Others have shown similar reductions in CV for PEF measured on a spirometer when the first three efforts were compared with the best three of five efforts. 20 PEF 2, SD2, and CV 2 likely provide a more accurate measure of mean PEF and variability when patients are trained and making frequent measurements and are more applicable for general use. Comparisons between our mean PEF 2 values at sea level and predicted PEF using the reference equations present in the NAEP document 4 demonstrate large differences between measured PEF 2 and predicted PEF (exceeding 27% for some age groups). This difference is not unexpected as the previous studies used a standard peak flowmeter (Wright) and different study designs. New normal values and regression equations using the newer generation peak flowmeters are needed. Our best estimate of CV for trained adults was 2.1% ( CV 2). This is lower than those previously reported (3.8 to 8.2%) that correspond better to our CV 1, which included training effect. Our lower CV 2 may be due to the five to eight measurements recorded for each subject, the inclusion of only healthy subjects, the vigorous coaching, and the criteria used to remove outliers. In adults, our measures of CV correlated only with mean PEF and these correlations were weak. Our finding that CV did not correlate with age, sex, height, weight, or altitude corresponds to earlier studies.l 8 22 Standard deviation correlated only with sex. Current asthma management guidelines use percent change in PEF to guide management. 4 We believe this is reasonable as CV is sex independent and although it has a minor correlation with mean PEF, the correlation is of little clinical significance because of the small coefficient of determination and regression coefficient. The upper 95th percentiles for CV 2 were 5.8% for adults and 9.4% for youths. Therefore, PEF must fall or rise by 5.8% in healthy adults and 9.4% in healthy youths to represent a statistically significant change. These values are considerably less than the 20% change that the NAEP guidelines suggest is clinically important. However, if patients with obstructive lung disease have approximately twice the variability of normal subjects, the NAEP guideline is close to the "noise" expected in the measurement for youths. Further study of intraindividual variability in patients with obstructive lung disease is warranted. Instrument variance accounted for only 5% of the overall variance for adults (calculated from SD2), indicating that most variability was due to the subject and not the instrument. At increased altitude (1,400 m), PEF meters underestimate PEF by 5.3 to 6.9%. Underreading of PEF at altitude occurs because peak 160 flowmeters are sensitive to air density. 30 Our results are similar to previous studies that found that some peak flowmeters (mini-wright) underread PEF by 6.8% per 100 mm Hg drop in barometric pressure Overall instrument precision was not significantly affected by altitude or by the magnitude of peak flow. As the NAEP document suggests, 4 three trials are adequate for most untrained people. Two thirds of our subjects achieved their highest or second highest PEF measurement in their first three efforts. However, 16% produced their highest PEF reading in efforts 6 to 8 and 63% required more than five efforts to achieve the desired level of reproducibility. It is important to properly train individuals to use peak flowmeters and to demonstrate adequate reproducibility prior to allowing independent use. This is particularly important in young and elderly individuals who exhibit more learning effect. We arbitrarily defined adequate reproducibility as five readings within 50 L/ min of one another for this study. A better criterion is that suggested by Dahlquist et at2 two highest readings within 10% of each other. Only 3% of our coached subjects failed to achieve this level of reproducibility in five to eight trials, whereas 9.6% failed to meet the ERS criteria 25 and 12.6% failed to meet the suggested reproducibility criteria of Burge. 24 We quantified intraindividual variability in healthy subjects and determined the change in PEF that represents statistically significant change (6% in adults and 9% in youths). The vast majority of overall variability was due to the subject and not the instrument. We propose using CV as the measure of intraindividual variability as it is sex independent and only minimally affected by the magnitude of peak flow. We also advocate the use of the reproducibility criteria of Dahlquist et al 20 (two highest readings within 10% of each other) when performing successive measurements. ACKNOWLEDGMENT: We thank Mark Cassidy, Mt. Sinai Hospital, Hartford, Conn, Debbie Crews, LDS Hospital, Salt Lake City, Utah, Marilyn Helgesen, Emory University Hospital, Atlanta, and Domenic Misiano, Massachusetts General Hospital, Boston, for their technical support. REFERENCES l Quanjer PH, Tammeling GF, Cotes JE, et al. Lung volumes and forced ventilatory flows: official statement of the European Respiratory Society. Eur Respir J 1993; 6(suppl16): D 'Angelo E, Prandi E, Milic-Emili J. Dependence of maximal flow-volume curves on time course of preceding inspiration. J Appl Physiol1993; 75: Kelly CA, Gibson GJ. Relation between FEV 1 and peak expiratory flow in patients with chronic airflow obstruction. Thorax 1988; 42: National Asthma Education Program (NAEP). Guidelines for the diagnosis and management of asthma. Bethesda, Md: National Institute of Health, August 1991 (OHMS publication No. lntraindividual Peak Flow Variability (Hegewald, Crapo, Jensen)

6 ) 5 Cross D, Nelson HS. The role of the peak flow meter in the diagnosis and management of asthma. J Allergy Clin Immunol 1991; 87: Nowak RM, Pensler MI, Sarkar DD, eta!. Comparison of peak expiratory flow and FEY 1 admission criteria for acute bronchial asthma. Ann Emerg Med 1982; 11: Woodhead M, ed. Guidelines on the management of asthma. Thorax 1993; 48(suppl):s1-s24 8 Burge PS. Single and serial measurements of lung function in the diagnosis of occupational asthma. Eur J Respir Dis 1982; 63(suppl123): Perrin B, Lagier F, L' Archeveque J, et al. Occupational asthma: validity of monitoring of peak expiratory flow rates and non-allergic bronchial responsiveness as compared to specific inhalation challenge. Eur Respir J 1992; 5: Ryan G, Latimer KM, Dolovich J, Hargreave FE. Bronchial responsiveness to histamine: relationship to diurnal variation of peak flow rate, improvement after bronchodilator, and airway calibre. Thorax 1982; 37: Higgins BJ, Britton JR, Chinn S, Cooper S, Burney PG, Tattersfield AE. Comparison of bronchial reactivity and peak expiratory flow variability measurements for epidemiologic studies. Am Rev Respir Dis 1992; 145: Dekker FW, Schrier AC, Sterk PJ, Dijkman JH. Validity of peak expiratory flow measurement in assessing reversibility of airflow obstruction. Thorax 1992; 47: Editorial. Reversibility of airflow obstruction: FEY 1 vs peak flow. Lancet 1992; 340: Mitchell DM, Gildeh P, Dimond AH, Collins JV. Value of serial peak expiratory flow measurements in assessing treatment response in chronic airflow limitation. Thorax 1986; 41: McCarthy DS, Craig DB, Cherniak RM. Intraindividual variability in maximal expiratory flow-volume and closing volume in asymptomatic subjects. Am Rev Respir Dis 1975; 112: Leeder SR, Swan A V, Peat JK, Woolcock AJ, Blackburn CRB. Maximum expiratory flow-volume curves in children: changes with growth and individual variability. Bull Eur Physiopathol Respir 1977; 13: Lebowitz MD, Knudson RJ, Robertson G, Burrows B. Significance of intraindividual changes in maximum expiratory flow volume and peak expiratory flow measurements. Chest 1982; 81: Nickerson BG, Lemen RJ, Gerder CB, Wegmann MJ, Robertson G. Within-subject variability and percent change for significance of spirometry in normal subjects and in patients with cystic fibrosis. Am Rev Respir Dis 1980; 122: Pistelli R, Fusol L, Muzzolon R, et al. Comparison of the performance of two mini peak flow meters. Respiration 1989; 56: Dahlquist M, Eisen EA, Wegman DH, Kriebal D. Reproducibility of peak expiratory flow measurements. Occup Med 1993; 8: Vaughan TR, Weber RW, Tipton WR, Nelson HS. Comparison of PERF and FEY 1 in patients with varying degrees of airway obstruction: effect of modest altitude. Chest 1989; 95: Groth S, Dirksen A, Dirksen H, Rossing N. Intraindividual variation and effect of learning in lung function examinations: a population study. Bull Eur Physiopathol Respir 1986; 22: Gardner RM, Crapo RO, Jackson BR, Jensen RL. Evaluation of accuracy and reproducibility of peak flow meters at 1,400 meters. Chest 1993; 101: Burge PS. Use of serial measurements of peak flow in the diagnosis of occupational asthma. Occup Med 1993; 8: Quanjer PH, Lebowitz MD, Gregg I. Peak expiratory flow: conclusions and recommendations of a working party of the European Respiratory Society (draft, July 1992) 26 Statement on technical standards for peak flow meters. Bethesda, Md: National Asthma Education Program, National Heart, Lung and Blood Institute, Feb 4, Dixon WS, Massey FJ. Introduction to statistical analysis. New York: McGraw-Hill, 1983; 109, , 517, Godfrey S, Kamburoff PL, Nairn JR. Spirometry, lung volumes and airway resistance in normal children aged 5 to 18 years. Br J Dis Chest 1970; 64: Nunn AJ, Gregg I. New regression equations for predicting peak expiratory flow in adults. BMJ 1989; 298: Thomas PS, Harding RM, Milledge JS. Peak expiratory flow at altitude. Thorax 1990; 45: Forster P, Parker RW. Peak expiratory flow rate at high altitude [letter]. Lancet 1983; 2:100 CHEST /1 07 /1 I JANUARY,