0
Clinical Investigations: ASTHMA |

The Utility of Peak Flow, Symptom Scores, and β-Agonist Use as Outcome Measures in Asthma Clinical Research* FREE TO VIEW

Frank T. Leone, MD, MS; Elizabeth A. Mauger, PhD; Stephen P. Peters, PhD, MD, FCCP; Vernon M. Chinchilli, PhD; James E. Fish, MD, FCCP; Homer A. Boushey, MD; Reuben M. Cherniack, MD; Jeffrey M. Drazen, MD, FCCP; John V. Fahy, MD; Jean Ford, MD, FCCP; Elliot Israel, MD, FCCP; Stephen C. Lazarus, MD; Robert F. Lemanske, MD; Richard J. Martin, MD; Stephen J. McGeady, MD; Christine Sorkness, PharmD; Stanley J. Szefler, MD; for the Asthma Clinical Research Network of the National Heart Lung and Blood Institute
Author and Funding Information

Affiliations: *From Thomas Jefferson University (Drs. Leone, Peters, Fish, and McGeady), Philadelphia, PA; Milton S. Hershey Medical Center (Drs. Mauger and Chinchilli), Hershey, PA; University of California at San Francisco (Drs. Boushey, Fahy, and Lazarus), San Francisco, CA; National Jewish Medical and Research Center (Drs. Cherniack, Martin, and Szefler), Denver, CO; Brigham and Women’s Hospital and Harvard Medical School (Drs. Drazen and Israel), Boston, MA, The Harlem Hospital (Dr. Ford), Harlem NY; and University of Wisconsin (Drs. Lemanske and Sorkness), Madison, WI. ,  Additional participating investigators are listed in Appendix.

Correspondence to: Frank T. Leone, MD, MS, Jefferson Medical College, 1025 Walnut St, Room 805, Philadelphia, PA 19107; e-mail: frank.leone@mail.tju.edu



Chest. 2001;119(4):1027-1033. doi:10.1378/chest.119.4.1027
Text Size: A A A
Published online

Study objectives: Several methods of utilizing peak expiratory flow (PEF) and other markers of disease activity have been suggested as useful in the management of asthma. It remains unclear, however, as to which surrogate markers of disease status are discriminative indicators of treatment failure, suitable for use in clinical trials.

Design: We analyzed the operating characteristics of 66 surrogate markers of treatment failure using a receiver operating characteristic (ROC) curve analysis.

Participants: Information regarding FEV1, symptoms, β2-agonist use, and PEF was available from 313 subjects previously enrolled in two Asthma Clinical Research Network trials, in which 71 treatment failures occurred (defined by a 20% fall in FEV1 from baseline).

Interventions: None.

Measurements and results: None of the measures had an acceptable ability to discriminate subjects with a ≥ 20% fall in FEV1 from those without, regardless of the duration of the period of analysis or the criteria for test positivity employed. Areas under the ROC curves generated ranged from 0.51 to 0.79, but none were statistically superior. Sensitivity and specificity combinations were generally poor at all cutoff values; true-positive rates could not be raised without unacceptably elevating false-positive rates concurrently.

Conclusions: Studies that seek to detect treatment failure defined by a significant fall in FEV1 should not use such individual surrogate measures to estimate disease severity.

Figures in this Article

Since 1994, the National Heart, Lung, and Blood Institute-sponsored Asthma Clinical Research Network (ACRN) has endeavored to design and execute clinical trials to evaluate both established and emerging therapies. One of its first major studies employed a time-to-event model, in which subjects started from a state of optimal asthma control, and were allowed to experience a decline in their lung function after withdrawal of inhaled corticosteroid therapy.1To capitalize on the operational advantages of such a study design, ACRN investigators recognized the need to define clinically meaningful study end points that captured subjects in“ treatment failure” at a critical point before they would otherwise require significant medical intervention. Prior studies29 have supported a relationship between a falling FEV1 and other markers of disease activity, such as worsening symptoms, increasing use ofβ 2-agonists, and various mathematical indexes of peak expiratory flow (PEF) variability. Attempts have been made to understand the adequacy and validity of these methods of assessing asthma control,,914 but few studies have estimated the comparative accuracy of these measures in an asthmatic population.1517

For the initial ACRN studies, the experimental criteria empirically chosen to define this critical point of decline included the following: (1) a ≥ 20% fall in FEV1 from the subject’s baseline values; (2) a fall in FEV1 to ≤ 40% of the predicted value; (3) a fall in the postbronchodilator PEF to≤ 65% of the baseline value on two of three consecutively scheduled measurements; (4) an increase in as-neededβ 2-agonist use to 8 puffs over the baseline value, or ≥ 16 puffs total, per 24-h period, sustained > 48 h; and (5) refusal to continue with study drugs because of lack of satisfaction with the treatment regimen. Subjects who met any of these criteria were termed “treatment failures” and had control of their asthma reestablished. Analysis of outcome data indicated that a fall in FEV1 of ≥ 20% was the most commonly identified measure defining treatment failure, and reliably identified subjects at a safe and appropriate point on the asthma continuum.1 Because few subjects were classified as treatment failures by the remaining criteria, it remained unclear exactly how the other experimental markers of asthma status related to the decline in FEV1 and to each other.

Because subject-recorded daily measurements may allow for more timely and perhaps safer recognition of treatment failure, information about the operating characteristics of subject-derived data would be valuable for optimizing definitions of treatment failure in future research projects. This study evaluates the experience of the ACRN in this regard. It also defines the operating characteristics of various self-reported measures of asthma with regard to their ability to identify a fall in FEV1 of ≥ 20% from baseline. Further, we attempted to identify the diary-derived measure with the best diagnostic capabilities within each of three measurement categories: peak flow, symptom score, andβ 2-agonist use. Finally, the optimum operating point of the most efficient measure within each category was identified where possible.

Subjects

Diary entry and pulmonary function information was collected from 326 subjects enrolled in two ACRN studies between 1994 and 1996. The subject population of these two studies had variable baseline severity of disease when evaluated as a whole. The 255 subjects enrolled in the β2-Agonists in Mild Asthmatics study18 had a baseline FEV1 of at least 70% of predicted, required a provocative concentration of methacholine to produce a 20% fall in FEV1 of no more than 16 mg/mL, used no more than 56 puffs of albuterol per week, and had not required inhaled corticosteroid therapy for a period of at least 6 weeks preceding recruitment. The 71 subjects enrolled in the Colchicine in Moderate Asthma (CIMA) study,1 had moderately severe asthma with baseline FEV1 ≥ 40% of predicted despite treatment with both inhaledβ 2-agonists and inhaled corticosteroids. During both source studies, subjects recorded disease-related information daily including morning and evening PEF, rescueβ 2-agonist use, and a symptom score from zero (no symptoms) to 3 (incapacitating symptoms) in each of five domains: shortness of breath, chest tightness, wheeze, cough, and mucus production. Both protocols required routine visits with measurement of FEV1 at prescribed intervals. Unscheduled visits with spirometry were performed if subjects met any of the self-reported treatment failure criteria at any time during their study.

For this analysis, a fall in FEV1 was used as the outcome indicator. Subjects were classified as disease positive if their FEV1 fell to 80% of baseline at any point during data collection. All other subjects were classified as disease negative. The date on which they either met disease-positive criteria or, for disease-negative subjects, the date on which data collection was terminated was designated as the study end date. Data were considered complete if daily symptom scores,β 2-agonist use, and both morning and evening PEF values were available for at least 7 of the 10 days preceding the study end date. Only the 313 subjects with a complete data set (96% of the total study population) were included in this analysis.

Test Definition

Each disease marker, or “test,” was defined by its unique combination of three variables: the measure to be analyzed, the length of time preceding study end included in the analysis, and the number of times the subject’s values must cross the imposed threshold (the positivity criteria). Each test then could be represented as the combination of these three variables and its sensitivity and specificity calculated across the range of cutoff values (κ; Table 1 ). In this way, the effect of each of the three variables on the discriminative capacity of the test as a whole could be accurately evaluated.

Analysis

We evaluated and compared these diagnostic tests by plotting receiver operating characteristic (ROC) curves and comparing the areas under each.1920 Curves were derived by plotting the true-positive rate (sensitivity) and false-positive rate (1-specificity) along the vertical and horizontal axes, respectively, in the standard fashion. Calculations of area under the curves, standard errors, and comparisons were performed using software (ROC Analyzer, 6.0; Centor and Keightley, 1990) and a desktop personal computer. In addition, confirmatory hand calculations of area were performed for each curve using the standard geometric method,20and were compared using the more conservative nonparametric method for evaluating multiple ROC curves acquired from the same population.21 Of the 66 surrogate measures, nine specific two-sided comparisons were planned a priori, and three were added post hoc. After Bonferonni correction for multiple comparisons, p values ≤ 0.004 were considered significant.

Population Characteristics

Of the 313 subjects who met inclusion criteria, 71 subjects (23%) were classified as disease positive. The two groups were similar in terms of age, gender, and race. The disease-positive group had a statistically greater degree of airway obstruction at baseline; however, both groups had subjects within a similar range of airway function at the start of their source studies (Table 2 ). The disease-positive group sustained a statistically significant fall in FEV1 over the course of the study (mean FEV1 at baseline, 2.72 L, vs 2.02 L at study end; p < 0.0001 by two-tailed paired t test), while the disease-negative control subjects did not (3.10 L at baseline vs 3.08 L at study end; p = 0.33). The groups did not cluster near the 20% fall in FEV1 cutoff.

Discriminative Capacity

The tests of exacerbation evaluated in our study all had an overall inadequate discriminative capacity. Areas under the ROC curves range from 0.51 (SE, 0.039) to 0.79 (SE, 0.034), with no curves attaining both sensitivity and specificity of ≥ 80% at any cutoff value. Visual inspection of each of the 66 curves generated suggested that none of the tests had a clear inflection point at which a favorable combination of sensitivity and specificity could be identified. Curves within and between groups are strikingly similar, regardless of measure employed, period analyzed, or positivity criteria used.

Individual Curve Comparisons

No index of PEF displayed significantly superior discriminative capacity over any other. Equally unexpected was the lack of effect of varying the period of observation or the positivity criteria (Table 3 ). Expressing peak flow as a percentage of baseline or of predicted PEF conferred no discriminative advantage over the absolute drop in peak flow value. Focusing on morning PEF alone did not improve the area under the ROC curves. Indexes of peak flow variability could not distinguish disease-positive subjects from control subjects with a higher degree of accuracy than any other measure of peak flow. Comparison of areas under the peak flow variability ROC curves suggests that there is no significant difference in the discriminative capacities of the indexes of variation tested, regardless of whether one chooses morning, evening, mean, or maximum PEF values as the reference. Measures of symptom scores, the number of additional puffs of β2-agonist, or the percentage of high rescueβ 2-agonist–use criteria had no discriminative advantage or disadvantage compared to peak flow expressed as any measure.

Additional Observations

After developing the ROC curves, reviews of individual test operating characteristics at clinically meaningful cutoff values were performed to fully evaluate the performance of the tests under usual circumstances. Table 4 lists the sensitivities and specificities of tests of interest, at selected cutoff points. Requiring the peak flow to drop to < 80% of the baseline morning value on two of three consecutive readings over 3 days proved to be an ineffective marker of FEV1 deterioration. While the false-positive rate is very low at this level, only 17% of disease-positive subjects were correctly identified. The cutoff used in the source studies (65% of baseline) did have excellent specificity, but the true-positive rate was only 8%. By relaxing the cutoff value of PEF to 100% of baseline, 81% of disease-positive subjects were correctly identified; this means 19% of subjects with a drop in FEV1 had no fall in PEF compared to baseline measurements. Moreover, 9% of disease-positive subjects had an improvement in PEF of ≥ 10% when compared to baseline, despite a 20% fall in FEV1.

Variability with respect to morning PEF of 30% was highly specific, falsely classifying disease-negative patients only 2% of the time. However, < 20% of disease-positive subjects were correctly classified using this criterion. Attempts to elevate sensitivity by changing the cutoff to 5% variability raises the false-positive rate to 46% while still only correctly classifying 74% of subjects in exacerbation. Twenty-six percent of subjects with a fall in FEV1 of ≥ 20% had ≤ 5% variability in PEF preceding exacerbation.

In the source studies, subjects were classified as treatment failures if their rescue β2-agonist use was either 8 puffs above baseline use or ≥ 16 puffs total over 24 h, for 2 consecutive days, an arbitrary cutoff that was individually determined. The specificity of this criterion was excellent (99%), but it correctly identified only 3% of the subjects with FEV1 ≤ 80% of baseline. Relaxing the individual’s criteria by 50% only improved the sensitivity to 39% with a false-positive rate of 10%. Further reductions yielded near complete lack of specificity. Because using a preset index of rescueβ 2-agonist use as an indicator of worsening asthma appeared en face to be insensitive, a second analysis was performed to evaluate the discriminative capacity of the absolute number of puffs above baseline required. This index proved similarly disappointing. Average symptom scores were found to be equally indiscriminate.

In clinical practice, it is often useful to employ diagnostic cutoffs that increase specificity and minimize false-positives. In research, however, it is critically important to accurately identify subjects with the outcome of interest and to accurately discriminate those with the outcome from those without. Because such clinical measurements have a very different purpose in the research setting, it is critically important that we gain an understanding of their inherent capability to differentiate subjects accurately. Daily diary measurements can be described as tests with continuous outcome values; therefore, the most appropriate way to evaluate their discriminative capacity is by means of ROC curve analysis.

It is generally accepted that tests evaluated for any purpose should have areas under their respective ROC curves of approximately 0.9 in order to be considered highly discriminative. While tests with larger areas are in general more discriminative than those with smaller areas, little systematic information is available about the benefit of small increments in area under the ROC curve. Comparison of the ROC curves generated by the 66 individual tests in our analysis suggests that no one index of disease severity appears superior to the rest (Fig 1 ). In fact, compared to tests described for other purposes, peak flow, symptom score, and β2-agonist indexes of asthma status appear to have rather poor discriminative capacity.22Given the intensive quality assurance measurements built into the source protocols, the lack of discrimination of peak flow does not appear to be because of technical variables.23

For a test to have excellent clinical utility, it should generally have a combination of sensitivity and specificity of at least 80% each, at some cutoff value. Lacking these attributes, the ROC curve of a test should at minimum have an identifiable “inflection point” on visual inspection. That is to say, a cutoff point should exist at which the true-positive rate is maximized despite an increasing false-positive rate. None of the ROC curves generated in this analysis had such an attribute.

We recognize some limitations in our approach to this question. First, standard methods for comparing ROC curves assume that sensitivity and specificity data for each curve are derived from independent populations of subjects. Deriving and comparing multiple ROC curves from the same pool of subjects requires an adjustment for the degree of correlation between test results.21 Instead, a sensitivity analysis was performed assuming both maximal correlation and no correlation at all. No comparison achieved significance at either the 0.004 or 0.05 level, regardless of the degree of correlation assumed. Second, standard ROC analysis assumes that the curves are generated by markers that are measured only once per patient. Concern exists regarding the effect of longitudinally repeated measurements on the position of the “null,” noninformative curve.24 Briefly, if subjects with positive and negative underlying responses differ in the number of marker measurements per subject, the null line can be bowed above or below the expected diagonal position (unity). In order to minimize this effect, the number of measures used in the analysis for each group was made equal. Though we believe that our approach is useful in evaluating the performance of these tests under precisely controlled circumstances such as clinical research, there may be difficulties associated with extending our findings into clinical practice, where surrogate measures are recorded longitudinally, and perhaps with differing frequencies.

The ROC method specifically evaluates the utility of the test in identifying the subjects’ cumulative risk of a 20% fall in FEV1, but does not give an adequate estimate of the subjects’ instantaneous risk of disease. Additionally, the ROC analysis cannot accurately assess the effect of interactions between variables. Even though none of the measures appear discriminative individually, it is conceivable that some combination of variables may in fact be predictive. In order to accurately assess instantaneous risk of disease for use in the clinical setting, and to account for the combined effect of these variables, regression modeling is currently in progress. Nevertheless, we believe our study design is reasonably representative of the way in which surrogate measures are employed during the course of clinical research, and that our analysis accurately assesses the performance of such measures under these circumstances.

Finally, the choice of reference standard must be questioned. Unlike a clinical “asthma exacerbation,” which may be difficult to quantify objectively by a single measure, data collected during the CIMA study indicated that a fall in FEV1 of 20% may represent an important point on the disease spectrum. Over 70% of subjects enrolled in the CIMA trial were designated as treatment failures because of a fall in FEV1 of ≥ 20%. Intervention at this point was believed to be timely in 68% of these subjects, as judged by the subject, the study coordinator, or both. No adverse outcomes were reported in this group.1 Choosing the 20% fall in FEV1 criterion was reasonable for this analysis because our two cohorts had significantly different FEV1 at the conclusion of their source studies, were well differentiated by this criterion, and did not appear to cluster near this cutoff.

Perhaps it is not surprising that indexes of peak flow, symptoms, andβ 2-agonist use are rather inadequate surrogate markers of airway obstruction as measured by spirometry. However, what is surprising is the magnitude of the inaccuracy. All of the tests evaluated in our study had poor operating characteristics and could not reliably discriminate between subjects in and out of treatment failure. Operating characteristics of the tests are likely to degrade further in situations with less rigid structure. Studies that seek to detect treatment failures defined by a significant fall in FEV1 should not use such surrogate measures individually to estimate disease severity.

Additional Contributing Investigators:

Christopher V. Chambers, James Diamond, and Kenneth R. Epstein, Thomas Jefferson University, Philadelphia, PA; Donna Baker and Susan J. Kunselman, Milton S. Hershey Medical Center, Hershey, PA; Monica Kraft, National Jewish Medical and Research Center, Denver, CO; Diane McLean and Sami Nachman, Harlem Hospital, Harlem, NY; Suzanne Hurd, National Heart, Lung, and Blood Institute, Bethesda, MD.

Abbreviations: ACRN = Asthma Clinical Research Network; CIMA = Colchicine in Moderate Asthma; PEF = peak expiratory flow; ROC = receiver operating characteristic

Supported by National Institutes of Health grants U10 HL-51810, U10 HL-51834, U10 HL-51831, U10 HL-51823, and U10 HL-51845.

Table Graphic Jump Location
Table 1. Test Characteristics Evaluated*
* 

Description of test categories evaluated. Each individual test is also defined by the number of days of observation included in analysis as well as the criteria required to define test positivity. PF = peak flow.

 

κ = cutoff value.

Table Graphic Jump Location
Table 2. Population Characteristics
* 

χ2 test of proportions.

 

Two-sided unpaired Student’s t test.

Table Graphic Jump Location
Table 3. Comparison of ROC Curve Areas
* 

Post hoc comparisons.

Table Graphic Jump Location
Table 4. Sensitivities, Specificities, and False-Positive Rates of Selected Tests at Clinically Important Cutoff Values
Figure Jump LinkFigure 1. Representative ROC curves of several surrogate markers of disease severity. The diagonal line represents the ROC curve of a test with no discriminative capacity (ie, random results), while a theoretical test with perfect discriminative capacity would have a ROC curve that follows the left-uppermost boundaries of the plot (heavy lines).Grahic Jump Location

We are indebted to Ms. Trina Armstrong (Milton S. Hershey Medical Center, Hershey, PA) for her flexibility, diligence, and computing expertise. Without her, this project would not have been possible.

Fish, J, Peters, S, Chambers, C, et al (1997) The effects of colchicine in moderate asthma.Am J Respir Crit Care Med156,1165-1171. [PubMed]
 
Quakenboss, J, Lebowitz, M, Kryzanowski, M The normal range of diurnal changes in peak expiratory flow rates.Am Rev Respir Dis1991;143,323-330. [PubMed]
 
Higgins, B, Britton, J, Chinn, S, et al The distribution of peak expiratory flow variability in a population sample.Am Rev Respir Dis1989;140,1368-1372. [CrossRef] [PubMed]
 
Gern, J, Eggleston, P, Schuberth, K, et al Peak flow variation in childhood asthma: a 3-year analysis.J Allergy Clin Immunol1994;93,706-716. [CrossRef] [PubMed]
 
Lebowitz, M, Knudson, R, Robertson, G, et al Significance of intraindividual changes in maximum expiratory flow volume and peak expiratory flow measurements.Chest1982;8,566-570
 
Henderson, A, Carswell, F Circadian rhythm of peak expiratory flow in asthmatic and normal children.Thorax1989;44,401-414
 
Hetzel, M, Clark, T Comparison of normal and asthmatic circadian rhythms in peak expiratory flow rates.Thorax1980;35,732-738. [CrossRef] [PubMed]
 
Apter, A, ZuWallack, R, Clive, J Common measures of asthma severity lack association for describing its clinical course.J Allergy Clin Immunol1994;94,732-737. [CrossRef] [PubMed]
 
Boulet, L, Milot, J, Turcotte, H Relationship between changes in diurnal variation of expiratory flows, lung volumes and respiratory symptoms after acute asthma.Respir Med1991;85,487-493. [CrossRef] [PubMed]
 
Morris, N, Abramson, M, Strasser, R Adequacy of control of asthma in a general practice.Med J Aust1994;160,68-71. [PubMed]
 
Bellia, V, Visconti, A, Insalaco, G, et al Validation of morning dip of peak expiratory flow as an indicator of the severity of nocturnal asthma.Chest1988;94,108-110. [CrossRef] [PubMed]
 
Bellia, V, Cibella, F, Visconti, A, et al Peak flow records in asthma: evaluation of an algorithm for interpretation of patterns.Eur Respir J1989;2(suppl 6),532s-535s
 
Gibson, P, Wlodarczyk, J, Hensley, M, et al Using quality-control analysis of peak expiratory flow recordings to guide therapy for asthma.Ann Intern Med1995;123,488-492. [PubMed]
 
Harm, D, Kotses, H, Creer, T Improving the ability of peak expiratory flow rates to predict asthma.J Allergy Clin Immunol1985;76,688-694. [CrossRef] [PubMed]
 
Reddel, H, Salome, C, Peat, J, et al That index of peak expiratory flow is most useful in the management of stable asthma?Am J Respir Crit Care Med1995;151,1320-1325. [PubMed]
 
Jamison, J, McKinley, R Validity of peak expiratory flow rate variability for the diagnosis of asthma.Clin Sci1993;85,367-371. [PubMed]
 
Toogood, J, Andreou, P, Baskerville, J A methodologic assessment of diurnal variability of peak flow as a basis for comparing different inhaled steroid formulations.J Allergy Clin Immunol1996;98,555-562. [CrossRef] [PubMed]
 
Drazen, J, Israel, E, Boushey, H, et al Comparison of regularly scheduled with as-needed use of albuterol in mild asthma.N Engl J Med1996;335,841-847. [CrossRef] [PubMed]
 
Clarke, J, O’Donnell, T A scientific approach to surgical reasoning.Theor Surg1991;6,45-51
 
Hanley, J, McNeil, B The meaning and use of the area under the receiver operating characteristic (ROC) curve.Radiology1982;143,29-36. [PubMed]
 
Hanley, J, McNeil, B A method of comparing area under receiver operating curves derived from the same cases.Radiology1983;148,839-843. [PubMed]
 
Baker, A, Bowton, D, Haponik, E Decision making in nosocomial pneumonia: an analytic approach to the interpretation of quantitative bronchoscopic cultures.Chest1995;107,85-95. [CrossRef] [PubMed]
 
Irvin, C, Martin, R, Chinchilli, V Quality control of peak flow meters for multicenter clinical trials.Am J Respir Crit Care Med1997;156,396-402. [PubMed]
 
Murtaugh, P ROC curves with multiple marker measurements.Biometrics1995;51,1514-1522. [CrossRef] [PubMed]
 

Figures

Figure Jump LinkFigure 1. Representative ROC curves of several surrogate markers of disease severity. The diagonal line represents the ROC curve of a test with no discriminative capacity (ie, random results), while a theoretical test with perfect discriminative capacity would have a ROC curve that follows the left-uppermost boundaries of the plot (heavy lines).Grahic Jump Location

Tables

Table Graphic Jump Location
Table 1. Test Characteristics Evaluated*
* 

Description of test categories evaluated. Each individual test is also defined by the number of days of observation included in analysis as well as the criteria required to define test positivity. PF = peak flow.

 

κ = cutoff value.

Table Graphic Jump Location
Table 2. Population Characteristics
* 

χ2 test of proportions.

 

Two-sided unpaired Student’s t test.

Table Graphic Jump Location
Table 3. Comparison of ROC Curve Areas
* 

Post hoc comparisons.

Table Graphic Jump Location
Table 4. Sensitivities, Specificities, and False-Positive Rates of Selected Tests at Clinically Important Cutoff Values

References

Fish, J, Peters, S, Chambers, C, et al (1997) The effects of colchicine in moderate asthma.Am J Respir Crit Care Med156,1165-1171. [PubMed]
 
Quakenboss, J, Lebowitz, M, Kryzanowski, M The normal range of diurnal changes in peak expiratory flow rates.Am Rev Respir Dis1991;143,323-330. [PubMed]
 
Higgins, B, Britton, J, Chinn, S, et al The distribution of peak expiratory flow variability in a population sample.Am Rev Respir Dis1989;140,1368-1372. [CrossRef] [PubMed]
 
Gern, J, Eggleston, P, Schuberth, K, et al Peak flow variation in childhood asthma: a 3-year analysis.J Allergy Clin Immunol1994;93,706-716. [CrossRef] [PubMed]
 
Lebowitz, M, Knudson, R, Robertson, G, et al Significance of intraindividual changes in maximum expiratory flow volume and peak expiratory flow measurements.Chest1982;8,566-570
 
Henderson, A, Carswell, F Circadian rhythm of peak expiratory flow in asthmatic and normal children.Thorax1989;44,401-414
 
Hetzel, M, Clark, T Comparison of normal and asthmatic circadian rhythms in peak expiratory flow rates.Thorax1980;35,732-738. [CrossRef] [PubMed]
 
Apter, A, ZuWallack, R, Clive, J Common measures of asthma severity lack association for describing its clinical course.J Allergy Clin Immunol1994;94,732-737. [CrossRef] [PubMed]
 
Boulet, L, Milot, J, Turcotte, H Relationship between changes in diurnal variation of expiratory flows, lung volumes and respiratory symptoms after acute asthma.Respir Med1991;85,487-493. [CrossRef] [PubMed]
 
Morris, N, Abramson, M, Strasser, R Adequacy of control of asthma in a general practice.Med J Aust1994;160,68-71. [PubMed]
 
Bellia, V, Visconti, A, Insalaco, G, et al Validation of morning dip of peak expiratory flow as an indicator of the severity of nocturnal asthma.Chest1988;94,108-110. [CrossRef] [PubMed]
 
Bellia, V, Cibella, F, Visconti, A, et al Peak flow records in asthma: evaluation of an algorithm for interpretation of patterns.Eur Respir J1989;2(suppl 6),532s-535s
 
Gibson, P, Wlodarczyk, J, Hensley, M, et al Using quality-control analysis of peak expiratory flow recordings to guide therapy for asthma.Ann Intern Med1995;123,488-492. [PubMed]
 
Harm, D, Kotses, H, Creer, T Improving the ability of peak expiratory flow rates to predict asthma.J Allergy Clin Immunol1985;76,688-694. [CrossRef] [PubMed]
 
Reddel, H, Salome, C, Peat, J, et al That index of peak expiratory flow is most useful in the management of stable asthma?Am J Respir Crit Care Med1995;151,1320-1325. [PubMed]
 
Jamison, J, McKinley, R Validity of peak expiratory flow rate variability for the diagnosis of asthma.Clin Sci1993;85,367-371. [PubMed]
 
Toogood, J, Andreou, P, Baskerville, J A methodologic assessment of diurnal variability of peak flow as a basis for comparing different inhaled steroid formulations.J Allergy Clin Immunol1996;98,555-562. [CrossRef] [PubMed]
 
Drazen, J, Israel, E, Boushey, H, et al Comparison of regularly scheduled with as-needed use of albuterol in mild asthma.N Engl J Med1996;335,841-847. [CrossRef] [PubMed]
 
Clarke, J, O’Donnell, T A scientific approach to surgical reasoning.Theor Surg1991;6,45-51
 
Hanley, J, McNeil, B The meaning and use of the area under the receiver operating characteristic (ROC) curve.Radiology1982;143,29-36. [PubMed]
 
Hanley, J, McNeil, B A method of comparing area under receiver operating curves derived from the same cases.Radiology1983;148,839-843. [PubMed]
 
Baker, A, Bowton, D, Haponik, E Decision making in nosocomial pneumonia: an analytic approach to the interpretation of quantitative bronchoscopic cultures.Chest1995;107,85-95. [CrossRef] [PubMed]
 
Irvin, C, Martin, R, Chinchilli, V Quality control of peak flow meters for multicenter clinical trials.Am J Respir Crit Care Med1997;156,396-402. [PubMed]
 
Murtaugh, P ROC curves with multiple marker measurements.Biometrics1995;51,1514-1522. [CrossRef] [PubMed]
 
NOTE:
Citing articles are presented as examples only. In non-demo SCM6 implementation, integration with CrossRef’s "Cited By" API will populate this tab (http://www.crossref.org/citedby.html).

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging & repositioning the boxes below.

CHEST Journal Articles
CHEST Collections
PubMed Articles
Guidelines
Asthma in pregnancy.
American College of Obstetricians and Gynecologists | 9/19/2008
Adapting your practice: treatment and recommendations for homeless patients with asthma.
National Health Care for the Homeless Council, Inc. | 6/13/2008
  • CHEST Journal
    Print ISSN: 0012-3692
    Online ISSN: 1931-3543