Reproducibility of Nasal Peak Inspiratory Flow Among Healthy Adults

Reproducibility of Nasal Peak Inspiratory Flow Among Healthy Adults* Assessment of Epidemiologic Utility Sung-Il Cho, MD; Russ Hauser, MD, SeD, MPH; and David C. Christiani, MD, MPH, FCCP

Study objective: To assess the reproducibility of nasal peak inspiratory flow (PIFn). Participants: Twelve healthy nonsmoking volunteers were studied. Methods: Repeated measurements of PIFn and oral (PIFm) peak inspiratory flow were performed for 5 consecutive days. Two methods of inhalation were compared. In the residual volume (RV) method, the forced maximal inspiratory maneuver was initiated from the end of a maximal expiration, while in the functional residual capacity (FRC) method, the maneuver was from the end of a tidal breath. Reproducibility was assessed by the intraclass correlation coefficient. Time trend for the 5 days was assessed by random effect models adjusting for different baseline for each subject. Results: The intraclass correlation coefficient (ICC) of PIFn was 0.89 (lower limit of one-sided 95% confidence interval is 0.80) by the RV method and 0.78 (95% lower limit is 0.63) by the FRC method, suggesting that both methods have good reproducibility. These were similar to the ICCs of PIFm by each method. The FRC method did not show a significant time trend for PIFn. The RV method had a small, but significant, decreasing time trend of a magnitude considered inconsequential for the purpose of epidemiologic study. Conclusion: PIFn, measured from either RV or from FRC, showed good reproducibility and can be employed in epidemiologic studies investigating the upper airways' response to air pollutant exposure. Further studies of the relationships between PIFn and signs and symptoms of rhinitis are needed to evaluate the utility of this test for clinical and epidemiologic use. (CHEST 1997; 112:1547-53) Key words: intraclass correlation ; nasal peak inspiratoty flow; reproducibility Abbreviations: CI =confidence interval; FRC=functional residual capacity; NOI =nasal obstruction PIFm=peak inspiratory flow via mouth ; PIFn = nasal peak inspiratory flow; RV = residual volume

responses of the respiratory tract to various air Thepollutants have been studied extensively. Although most studies have focused on the lower respiratory tract, re cently, upper respiratory segments, including the nose and sinuses, have drawn attention either as a primmy target organ of airborne pollutants or as an easily accessible part of the respiratory tract that may reflect changes in the lower tract. 1-3 The recent interest in the upper respiratory tract stems from the fact that its structure and physiologic function are designed to provide the *From the Occupational Health Program, Departm ent of Environmental Health, Harvard School of Public Health , Boston. Supported by NIH grants ES0.5947 and ES00002, and NIOSH grant U60/CCU109979 Manuscript received January 23, 1997; revi sion accepted May 8 , 1997. Reprint requests: David C. Christiani, MD, Dept of Endronm.ental Health, Occupational Health Program, Haruard School of Public Health, 665 Huntington A~,;e, Boston, MA 02115

index;

initial mechanisms for clearing pollutants from inhaled air, that its respiratory mucosa is similar to that in the lowe r tract, and that certain types of pollutants tend to be deposited or absorbed in the nose to a greater extent than other types of pollutants. These characte1istics imply that the upper respiratory responses may be a reasonably good indicator of the earliest adverse health effect resulting from exposure to certain air pollutants. Air pollutants can induce several physiologic responses in the upper airway. These responses include irritation, alteration of airflow resistance, impairment of mucociliary clearance, impairment of immune defenses, and direct damage to cells.4 Although various biomarkers are available to measure th ese responses,5 their use in epidemiologic studies is limited b ycost and subject compliance.6 For these reasons, the measurement of air flow by nasal peak inspiratory flow (PIFn) is of interest; it is easily CHEST I 112 I 6 I DECEMBER, 1997

1547

performed, noninvasive, and inexpensive, allowing for application to population studies. 7-9 The validity of PIFn bas been evaluated in several studies by assessing the correlation between PIFn measurements and other parameters of nasal air flow such as nasal resistance measured by rhinomanometry, L0-12 nasal peak expiratory flow,13 or the subjective sensation of nasal patencyY These studies consistently showed strong correlations between PIFn and the other measurement techniques. Many recent studies applying PIFn measurement in clinical settings found it useful for objectively comparing nasal patency behveen different treatment groups in both pediatric and adult populations. 14-16 In particular, the validity of PIFn, as compared to rhinomanometry, drew investigators' attention because rbinomanomeh)', a standardized technique for the measurement of nasal airflow and pressure,17 which has an established role in clinical research and patient evaluation, is labor intensive, time consuming, and expensive for field applications in occupational or community population studies of environmental exposures.6·10·11 However, some authors have raised concerns that nasal peak flow measurements might be poorly reproducible.l 8 A result or measurement is said to be reproducible when the repetition of it gives the same results .L 9 Reproducibility is an essential characteristic of a test for utility in epidemiologic studies of inhaled pollutants. However, there has been little objective evaluation of the reproducibility of PIFn measurements. In this study, we performed serial measurements of PIFn and oral peak inspiratory flow among healthy volunteers to evaluate the reproducibility of PIFn measurement. \
Subjects and Design

We recruited 14 healthy nonsmoking subjects from among the students and faculty at the Harvard School of Public Health. Informed consent was obtained and the study was approved by the Institutional Review Board of the school. At least 4 days before the study began, all subjects had a training session to familiarize themselves with the technique of nasal and oral peak flow. \Ve utilized a repeated measure design to assess not only the reproducibility but also the time tre nd of the measurements, which is often of interest to detect cumulative effects of air pollutants. The subjects were tested over 2 weeks, 5 consecutive days each week. Each subject was asked to keep th e same time of the day for the testing throughout the whole study to minimize the effect of diurnal valiation in nasal resistance. At least three satisfactory maximal inspirations were done on each day and the maximum value of those was used for the analysis. 1548

Two methods of forced maximal inhalation were used, one beginning from the end of full expiration (residual volume [RV] method), the other beginning From the end of normal expiration (functional residual capacity [FRC] method). If the FRC method were comparable to the RV method in reproducibility. it would be the more comfortable way to perform the nasal peak inspiratOJy maneuver. Each subject performed the test with both methods. To minimize any influence by carryover effect from the earlier procedure to the late r procedure, each m ethod of inhalation was used in a diffe rent week. Therefore, the 14 subjects were randomized into two groups, assigning 7 subjects to each group. One group perform ed the tests by RV method during the first week and FH.C method duling the second week, th e other group performed the testing in reverse order, FH.C during th e first week and RV during the second week. Twelve of th e 14 subjects completed the testing and were included in the analysis. One subject dropped out because of an acute upper respirat01y tract infection; another subject was unable to correctly perform the maneuver. Since one subject from each randomized group dropped out, six subjects were left in each group for the analysis. Among the 10 test days for each subject, two subjects had 1 missing day, and another subject had 2 missing days. All others finished the complete testing protocol. Nasal Peak Flow Measurement Procedure

A portable microspirometer (Micro Plus; London, UK) was used for th e measurement of peak inspiratory flows. For PIFn , a modified nasal continuous positive airway pressure mask was attached to the spirometer. The masks were chosen to fit tightly on each subject's face without touching the nose, and were cleaned with alcohol and dlied before and after usage. Since it has been shown that taking the maximum among three acceptable blows minimizes the effort dependent variation 20 in the case of peak bronchial expirat01y flow measure ment, we adopted a similar protocol using three forc ed maximal inspirations. PIFn measurements were taken under supervision with at least a 30-s intetval between each inspiration. Another three forced maximal inspirations were also measured through the mouth using a mouthpiece, to measure the peak inspiratory flow via mouth (PIFm). Subjects were encouraged to inhale as hard and fast as they could through the nasal mask or mouthpiece. If the maneuver was not satisfactmy in te rms of th e effort, leakage, or body position, a nadditional blow was performed. The additional fourth inspiratmy maneuve r was needed on 2 days for one subject and 3 days for another subject. Questionnaire

Upon enrollment in the study, each subject completed a short questionnaire that included questions on age, height, gender, and history of respiratmy illness and medication use. Thereafter, on each test day, another questionnaire on environmental exposures during the 24 h preceding testing was completed. This questionnaire also contained questions on medications and caffeinated beverage intake. Subjects were asked to stay indoors and at rest at least 30 min prior to the tests. One of the investigators (S.C. or H..H. ) interviewed the subject on each day's visit with a checklist of upper and lower respiratory symptoms to screen for respiratory infection. Any subject who showed such symptoms was withdrawn from the study. Statistical Analysis

A nasal obstruction index (NOI) was calculated by the following formula: Clinical Investigations

NOI= (PIFm- PIFn)I PIFm

RESULTS

where PIFm is peak inspiratory flow through mouth and PIFn is peak inspiratmy flow through nose. This index reflects the fraction of the reduction by the nasal ailway segment of the flow into the trachea through mouth. This index is 1 if the nasal passage is completely obstructed, and 0 if the nasal airway presents no resistance at all. Similar indexes to represent nasal patency are "blocking index" for nasal peak expiratory flow21 or "nasal patency index" for forced inspiratory volume in 0.5 s through the nose.22 To evaluate the reproducibility of measurements, the intraclass correlation coefficient was calculated for PIFn, PIFm, and NOI for each inhalation method, using one-way random effect models. 23 The intraclass correlation coefficient represents the proportion of between-subject variability among the total variability. The total variability of peak flow measurements consists of the portion resulting from within-subject variability among repeated measurements as well as the portion resulting from betweensubject differences. A coefficient of 0 indicates that the measurement is random and completely nonreproducible, and a coefficient of 1 means that the measurement is the same for any time for a certain subject, representing perfect reproducibility. Thus, a larger coefficient indicates greater reproducibility, which is more desirable to detect the effect of a particular exposure on peak flow values, and generally a coefficient bigger than 0.80 is considered good reproducibility.24 A longitudinal random effects model25 was applied to the PIFn and PIFm values with software (SAS PROC MIXED) 26 to examine whether the re is a trend over time. In this model, the variance of the random intercept represents the between-person variation of the measure ments, and the fixed slope of the linear time trend represents the instability over time in the measurements. For subject i (i= 1, 2,... , n) at timeS (j= 1, 2, 3, 4, 5), the observation (Y) is expressed as follows:

Twelve subjects were included in the analysis. There were five women and seven men; eight were white and four were Asian. In the baseline questionnaire, none of the participants had a history of atopy, allergic rhinitis, or any seasonal or chronic respiratory illness. None of the subjects had current exposure to occupational chemicals, or environmental tobacco smoke at home or school, or used medication. Nine subjects performed the testing approximately at the same time of the day within a 2-h range. Each of the other three subjects had ::::;3 days with different timing for the testing. The distributions of age, height, PIFn, PIFm, and NOI are shown in Table l. Since each subject had five daily measurements for peak flow over a week, the average of the five measurements was taken, and the distribution of those averages are presented in Table l. Peak flow values and NOis for FRC and RV methods were similar, with FRC being slightly lower (Table 1). Paired t tests on the 5-day means for each individual did not show a significant difference between FRC and RV for PIFn (p=0.6), but a moderately significant difference for PIFm (p=0.04). Figure 1 shows the daily peak flow values as the percentage of the mean value for the individual for the 5-day consecutive measurements. The plots indicate that there is no noticeable increasing or decreasing trend, although PIFn from RV seems minimally declining on the average. In general, PIFn appears to be more variable than PIFm, and the FRC method appears to be more variable than the RV method. This is also reflected in the coefficients of variation shown in Table l. Intraclass correlation coefficients for PIFn, PIFm, and the NOI measured by the FRC and RV methods are shown in Table 2. Both FRC and RV methods gave coefficients of :::::0.78, showing good reproducibility. Since coefficients for PIFn are similar to those of

where 130 is the fixed baseline value adjusted to the middle of the week (the third day) ( t3 = 0), 131 is the fixed time slope, e, is the measure of the random effect of subject i which represents between-subject variability, and ey is an exchangeable correlated error. This model estimates the underlying common time slope (13 1 ) among the subjects adjusting for the difference in the baseline values in each individual. The time value was set to zero on the third day so that the inte rcept 13 0 represe nts the mean of all measurements among all individuals.

Table !-Distribution of Age, Height, PIFn, PIFm, and NOI Measurements* CV,%

Distribution

Age, yr Height, e m PIFn: FRC, Us PIFn: RV, Us PIFm: FRC, Us PIFm: RV, Us NOI: FRC, Us NOI: HV, Us

Mean 1 (SD)

Range

Mean

Range

31.83 (5.91) 168.08 (9.20) 3.33 (0.87) 3.4 (1.00) 6.47 (1.48) 6.93 (1.32) 0.48 (0.09) 0.51 (0.10 )

21-43 157-188 1.67-4.93 1.94-5.15 4.48-9.10 4.63-9.41 0.36-0.66 0.28-0.68

12.1 10.1 8.2 4.4 14.3 11.3

3.4-26.2 3.0-25.3 2.7-24.9 1.7-7.8 5.0-25.5 2.4-37.7

*CV=coefficient of variation. 1 Five-day mean for 12 subj ects. CHEST I 112 I 6 I DECEMBER, 1997

1549

:::j ..>.<:

C1> C1>

:;::: Q)

£

0

c:::

""

PIFn from

PIFn from

RV

150

: : 1~

125 ..>.<: Q)

~ Q)

£

c:::

""

E

50 1

'~j 75

0

Q)

Q.>

E

~

'"'[_----

FRC

~

sol 25

25 ~

3

2

4

5

1

PIFm from

3

2

4

5

DAY

DAY

PIFm from

FRC

RV

175 150 125

0

m E

75

75

50

25

501

i

25

0 ~"·~~~~~~~~~~~~~~--~~~

0 ~~~~~~~~~~~~--~~~~~

3

2

4

5

1

3

2

5

4

DAY

DAY

FIGURE l. Measurements of PIFn and PIFm represented as the percentage of the 5-day mean for each subject.

PIFm, the reproducibility of PIFn is nearly as good as that of PIFm, suggesting that the nasal airway segment does not cause significant reduction in reproducibility. The NOI shows poorer reproducibility than PIFn, suggesting that using NOI gives no advantage over using the nasal peak flow measurement itself. An analysis using a random effects model was used to evaluate the time trend after accounting for individual difference in the baseline value. This model also makes efficient use of data containing missing values. The data did not significantly violate the assumption of normality in the error and in the random between-subject variability. Since the same subjects were tested with two

different methods, separate analyses by the methods were done. A model with a random intercept and a fixed slope was fitted (Table 3). The time slope was not significant for the FRC method, whereas it was marginally significant for the RV method (p=0.05), indicating a decline of 0.06 Us/d (95% confidence interval [CI], -0.001, -0.12). This decline accounts for approximately a 2% change in the peak flow value. Table 4 shows the estimated means ofPIFn, PIFm, and NOI at the midweek (on day 3) and the corresponding 95% Cis. Small differences in the mean values from the model and from the crude data (Table 1) result from the missing values. Cis of the estimated means for FRC

Table 2-Intraclass Correlation Coefficient (R) and One-Sided 95% CI (95% LL) for PIFn, PIFm, and NOI, at FRC and RV PIFm

PIFn

R

95% LL

1550

NOI

FRC

RV

FRC

RV

FRC

RV

0.78 0.63

0.89 0.8

0.79 0.64

0.94 0.89

0.49 0.28

0.75 0.58 Clinical Investigations

Table 3-Coefficients From Random Effect Models for PIFn and PIFm

PIFn, Us (FRC) PIFn, Us (RV) PIFm, Us (FRC) PIFm, Us (RV)

Parameter

Estimate

SE

p Value

Intercept* Time slope Intercept Time slope Intercept Time slope Intercept Time slope

3.34 0.07 3.41 - 0.06 6.47 -0.01 6.93 0.04

0.25 0.04 0.29 0.03 0.43 0.07 0.38 0.03

0.00 0.10 0.00 0.05 0.00 0.84 0.00 0.23

*Intercept represents the predicted mean on the midweek day (day 3).

and RV methods closely overlap, suggesting that the two methods are not markedly different after accounting for the differences in the baseline value among the subjects.

DISCUSSION

The reproducibility of a test can be evaluated by repetition of the test in the same subject. This repetition may be conducted over minutes, over days, or over even longer intervals. Since the main utility of PIFn would be to measure acute reversible effects of air pollutants occurring in the environment, the reproducibility over days is the most relevant dimension to assess the test performance. This measure of reproducibility reflects the combination of measurement error and biological variability over the relevant period. Our study shows that PIFn measurements on healthy nonexposed subjects possess good reproducibility with an intraclass correlation coefficient of 0.89 when measured from RV and 0.78 when measured from FRC. There was no apparent instability over time with the FRC method, although a small and marginally significant decline appeared with the RV method. The reason for this decline may be due to chance, although we cannot exclude the possibility of maneuver-induced nasal congestion as a result of the forceful inhalation from RV. However, this de-

Table 4-Estimated Means and 95% Cis for the Centered Data*

PIFn (FRC) PIFn (RV) PIFm (FRC) PIFm (RV) NOI (FRC) NOI (RV)

Mean, Us

95% CI

3.34 3.41 6.47 6.93 0.48 0.51

2.84, 3.83 2.85, 3.97 5.63, 7.31 6.18, 7.68 0.43, 0.53 0.45, 0.57

*Obtained from random effects models for PIFn, PIFm, and NOI.

cline is about 2% of the peak flow value and should be inconsequential for practical purposes. Our data reveal that nasal peak flow measurements have better reproducibility than previously believed. Intraclass correlation coefficients for PIFn are comparable to that of PIFm in our study. It is also comparable to the results from the study by Pavord and coworkers 27 in which reproducibility of various measures of bronchial peak expiratory flow were examined. In that study, the log-transformed values of mean daily amplitude (maximum-minimum), amplitude percent highest ([maximum-minimum]/maximum), amplitude percent mean ([maximum-minimum]/mean), and other measures showed intraclass correlation coefficients ranging from 0.79 to 0.86. In our study, the RV method had a higher intraclass correlation coefficient but showed a slightly declining slope. However, the FRC method is more comfortable for the subject to perform and may therefore be the preferred method for field applications. The coefficients of variation shown in Table 1 are comparable to the study by Shelton and coworkers28 in which the mean intrasubject coefficient of variation from 10 baseline measurements among six subjects was 9.8% (range, 4 to 39%) for PIFn. They also reported coefficients of variation of 26% (range, 8 to 53%) for the nasal airway resistance measured by posterior rhinomanometry and 6% (range, 1.4 to 14%) by anterior rhinomanometry. In a recent study, Shelton 29 observed that mean intrasubject coefficients of variation were 16% for posterior rhinomanometry and 12% for anterior rhinomanometry, from eight consecutive measurements in each normal subject. Our result shows that PIFn from RV has a mean coefficient of variation of 10.1 %, which falls between the two previous estimates for anterior rhinomanometry. Recently, Clarke and Jones30 compared PIFn and rhinomanometric resistance in the nasal airway after histamine challenge and after decongestion with xylometazoline, an a-adrenergic agonist.3l With hisCHEST /112 /6/ DECEMBER, 1997

1551

tamine challenge, PIFn showed a smaller percent change than nasal resistance, and with xylometazoline decongestion, PIFn showed a smaller percent change than nasal resistance. The authors concluded that PIFn was less sensitive than the rhinomanometric measurements of nasal resistance. They suggest that these findings may be because of the combined effects of nasal alar collapse and "downstream" resistance in the small intrapulmonary airways, both of which limit nasal inspiratory airflow. In an earlier study, Bridger and Proctor32 found that nasal obstructive symptoms are related to the maximum inspiratory flow and to the collapsibility of the "nasal valve" rather than to the upstream or downstream resistances. Our data (Tables 1 and 4) show that PIFm is about twice PIFn, suggesting that the major limiting factor of the PIFn is in the nasal segment rather than downstream airway. Pertuze and colleagues33 share this view by stating that even though obstruction in the airway below the nasopharynx or a gross failure to generate a significant driving pressure could give a misleadingly low PIFn, this value could be checked by comparing values of PIFn with those of PIFm. Thus, downstream ai1way resistance, if any, should not be an important problem in the practical application of PIFn. Pertuze and coworkers33 evaluated the factors limiting PIFn by applying thin-walled stenting to prevent alar collapse and topical xylometazoline to reduce resistance due to mucosal vasculature. Their results demonstrate that narrowing of the alar vestibule and the state of mucosal vasculature both influence PIFn independently. With combined intervention of the alar stenting and nasal vasoconstriction, maximum flows via the nose both on inspiration and expiration approached flows via the mouth. Thus, nasal air flow is limited by two major components, the mucosal component and the nasal valve, and not so much by other factors such as downstream resistance. Between the two components, the mucosal component is the potential target of air pollutant exposure and therefore the component of interest as far as PIFn measurement is concerned. The utility of PIFn measurements in epidemiologic studies depends on the following: (1) the reproducibility of PIFn measurements; (2) the validity of PIFn measurements; and (3) the sensitivity of PIFn to detect a response. Our study investigated the reproducibility of PIFn measurements and serves as a pilot study for a larger epidemiologic application. Despite the small sample size, this study showed good reproducibility of PIFn measurements. In conclusion, the reproducibility shown in our study, extending previous studies on the validity of PIFn, supports the utility of PIFn measurements to evaluate the upper airway response to ambient ex1552

posures among healthy nonsmoking adults. Further human population research on the sensitivity of PIFn to detect upper airway responses after various environmental exposures is needed. ACKNOWLEDGMENT: The authors are grateful to the volunteers who participated in this study as the subjects. We also wish to thank Drs. Hang Lee and Louise Ryan in the Department of Biostatistics, Harvard School of Public Health, for their valuable advice on statistical analysis.

REFERENCES 1 Koenig JQ, Pierson WE. Nasal responses to air pollutants. Clin Rev Allergy 1984; 2:255-61 2 Koren JQ, Hatch GE, Graham DE. Nasal lavage as a tool in assessing acute inflammation in response to inhaled pollutants. Toxicology 1990; 60:15-25 3 Witek TJ. The nose as a target for adverse effects from the environment. Am J Ind Med 1993; 24:649-57 4 Leopold DA. Pollution: the nose and sinuses. Otolaryngol Head Neck Surg 1992; 106:713-19 5 Leopold DA. Nasal toxicity: end points of concern in humans. In: Miller F , ed. Nasal toxicity and dosimetry of inhaled xenobiotics: implications for human health. Washington, DC: Taylor & Francis, 1995; 23-39 6 Lund VJ. Toxicity of nasal respiratory mucosa in humans. In: Miller, ed. Nasal toxicity and dosimetry of inhaled xenobiotics: implications for human health. Washington, DC: Taylor & Francis, 1995; 277-87 7 Youlten LJF. The peak nasal inspiratory flow meter: a new instrument for the assessment of the response to immunotherapy in seasonal allergic rhinitis. Allergol Immunopathol 1980; 8:335 8 Jones AS, Lancer JM. Rhinomanometry. Clin Otolaryngol 1987; 12:233-36 9 Fairley JW, Durham LH, Ell SR. Correlation of subjective sensation of nasal patency with nasal inspiratory peak flow rate. Clin Otolaryngol1993; 18:19-22 10 Jones AS, Viani L, Phillips D, eta!. The objective assessment of nasal patency. Clin Otolaryngol1991; 16:206-11 11 Gleeson MJ, Youlten LJF, Shelton DM, eta!. Assessment of nasal airway patency: acomparison of four methods. Clin Otolarygol 1986; 11:99-107 12 Holmstrom M, Scadding GK, Lund VJ, e t a!. Assessment of nasal obstruction: a comparison between rhinomanometry and nasal inspiratory peak flow. Rhinology 1990; 28:191-6 13 Wihl JA, Maim L. Rhino manometry and nasal peak expiratory and inspiratory flow rate. Ann Allergy 1988; 61:50-55 14 Scadding GK, Darby YC, Austin CE. Effect of short-term treatment with fluticasone propionate nasal spray on the response to nasal allergen challenge. Br J Clin Pharmacal 1994; 38:447-51 15 Shaikh WA. Ephedrine-saline nasal wash in allergic rhinitis. J Allergy Clin Immunol 1995; 96:597-600 16 Jordana G, Dolovich J, Briscoe MP, et a!. Intranasal fluticasone propionate versus loratadine in the treatment of adolescent patients \vith seasonal allergic rhinitis. J Allergy Clin Immunoll996; 97:588-95 17 Clement PAR. Committee report on standardization of rhinomanometry. Rhinology 1984; 22:151-55 18 Druce HM, Schumacher MJ. Nasal provocation challenge. J Allergy Clin Immunol 1990; 86:261-64 19 Last JM . A dictionary of epidemiology. 3rd ed. New York: Oxford University Press, 1995 20 American Thoracic Society. Standardization of spirometryClinical Investigations

1987 update. Am Rev Respir Dis 1987; 136:1285-98 21 Taylor G, Macneil AR. Assessing degree of nasal patency by measuring peak expiratory flow rate through the nose. J Allergy Clin Immunol 1973; 52:193-98 22 Davies HJ. Measurement of nasal patency using a vitalograph. Clin Allergy 1978; 8:517-23 23 Fleiss JL. The design and analysis of clinical experiments. New York: John Wiley & Sons, 1986 24 Chinn S. The assessment of methods of measurement. Stat Med 1990; 9:351-62 25 Laird NM , Ware JH. Random-effects models for longitudinal data. Biometrics 1982; 38:963-74 26 SAS Institute. The mixed procedure. Cary, NC: SAS Institute, 1992 27 Pavord I, Richards K, Wisniewski A, et al. Repeatability of

28 29 30 31 32 33

measure of peak expiratory flow variability for community studies. Thorax 1992; 47:230 Shelton DM , Youlten LJF, Chan TB, eta!. A comparison of three methods of assessing nasal airway response to topical histamine. Thorax 1985; 40:216 Shelton DM. Evaluation of active anterior and posterior rhinometry in normal subjects. Clin Otolaryngol 1992; 17:178-82 Clarke RW, Jones AS. The limitations of peak nasal flow measurement. Clin Oto!a,yngol1994; 17:502-04 Clarke RW, Jones AS, Richardson H. Peak nasal inspiratory flow- the plateau effect. J Laryngol Otol 1995; 109:399-402 Bridger GP, Proctor DF. Maximum nasal inspiratory flow and nasal resistance. Ann Otol 1970; 79:481-88 Pertuze J, Watson A, Pride NB. Maximum airflow through the nose in humans. JAppl Physiol 1991; 70:1369-76

CHEST I 112 I 6 I DECEMBER, 1997

1553