Reducing the number of daily measurements results in poor estimation of diurnal variability of peak expiratory flow in healthy individuals.
D Gupta, AN Aggarwal, S Chaganti, SK Jindal
Department of Pulmonary Medicine, Postgraduate Institute of Medical Education and Research, Chandigarh- 160 012, India. , India
Department of Pulmonary Medicine, Postgraduate Institute of Medical Education and Research, Chandigarh- 160 012, India.
AIM: To determine the effect of reducing number of daily measurements on estimation of diurnal variability (DV) of peak expiratory flow (PEF). SUBJECTS AND METHODS: PEF was recorded five times daily for three days in 152 healthy adults. Amplitude percent mean (A%M) and standard deviation percent mean (SD%M) were calculated on third day from five, four, three and two daily readings. Proportion of variability explained by partial schedules was calculated and limits of agreement derived to assess if these methods could be used interchangeably. RESULTS: Four, three and two measurements explained 90-95%, 70-82% and 55% DV respectively using A%M. All schedules of partial measurement using SD%M explained >90% DV. Limits of agreement for A%M and SD%M widened as number of measurements were reduced. CONCLUSIONS: DV obtained by fewer daily measurements agrees poorly with results obtained from five measurements. SD%M is a better alternative if DV is assessed from fewer readings.
|How to cite this article:|
Gupta D, Aggarwal A N, Chaganti S, Jindal S K. Reducing the number of daily measurements results in poor estimation of diurnal variability of peak expiratory flow in healthy individuals. J Postgrad Med 2000;46:262-4
|How to cite this URL:|
Gupta D, Aggarwal A N, Chaganti S, Jindal S K. Reducing the number of daily measurements results in poor estimation of diurnal variability of peak expiratory flow in healthy individuals. J Postgrad Med [serial online] 2000 [cited 2020 May 30 ];46:262-4
Available from: http://www.jpgmonline.com/text.asp?2000/46/4/262/262
A diurnal variability in peak expiratory flow (PEF) is well recognised in humans. The commonest pattern is that of a nadir in early morning, followed by a progressive rise towards late afternoon and a slight fall at night. This variability is exaggerated in patients with bronchial hyperreactivity, and may serve as a useful tool in the diagnosis and monitoring of patients with asthma.
The optimum number of daily measurements while recording PEF variability is not clearly defined. Most practitioners monitor variability by twice daily measurements. There are some data to suggest that diurnal variation may be grossly underestimated by such protocols., We conducted this study to find the minimum number of daily measurements for assessment of diurnal variability in PEF in healthy young adults. This information is important for any meaningful interpretation of PEF records used for home monitoring by patients with bronchial asthma.
The study was carried on 152 apparently healthy non-smoker volunteers. Details regarding enrolment and demographic characteristics of these subjects, and protocols for measuring PEF and PEF variability, have been previously described. Subjects were instructed to record PEF five times a day: on waking, between 9 and 11 A.M., between 2 and 4 P.M., between 6 and 8 P.M., and at bedtime (subsequently referred to as W, 9, 2, 6 and B respectively) for three consecutive days. The first two days were considered as learning days and these measurements were not included in the final analysis. Diurnal variation in PEF for each subject on the third day was calculated using two indices: amplitude percent mean (A%M) and standard deviation percent mean (SD%M). Both A%M and SD%M were calculated separately for five (W926B), four (W92B, W96B, W26B), three (W9B, W2B, W6B) and two (WB) readings. Readings taken on waking in morning and at bedtime were always included to represent the first and last possible readings during waking hours. Population mean and SD for each index was calculated and paired t-test used to evaluate if differences between two sets of data were statistically significant. A ratio between each index and the corresponding value from all five measurements was calculated to assess the proportion of variability explained by each schedule of partial measurement. Mean of differences between indices obtained by partial and complete schedules of measurements were calculated and limits of agreement derived to evaluate if these methods could be used interchangeably.
Mean A%M for the study population using all five readings on the third day of measurement was 7.23 ? 3.48. This value decreased progressively as the number of daily measurements was reduced; all these differences were highly significant in two-by-two comparisons [Table:1]. Analysis of ratio of A%M obtained by partial and complete schedules of measurements showed that four, three and two measurements respectively picked up approximately 90-95%, 70-82% and 55% of variability in comparison to five measurements [Table:1]. There was considerable lack of agreement between values obtained by five and less than five daily readings. As the numbers of measurements were reduced, the limits of agreement widened [Table:1]. Using only two daily measurements (waking and bedtime), for example, gave values that were up to 8.98 lower or up to 1.50 higher than those obtained using five daily measurements.
The effect of reducing the number of PEF measurements on diurnal variation expressed as SD%M was found to be different [Table:1]. Overall, the effect of reducing the number of daily measurements was less marked in comparison to that seen with A%M, and using even two daily readings could explain more than 90% variability in PEF. No significant differences were observed when W92B, W96B or W6B schedules were used, but there was still a considerable lack of agreement with values obtained from a complete profile [Table:1]. Using W26B, W2B or W6B schedules paradoxically increased SD%M values in majority of subjects. Using only waking and bedtime readings led to a worse agreement as compared to other schedules of incomplete measurement.
Our study attempts to answer some issues regarding the optimum number of PEF readings needed to determine diurnal variability, and the objective index to be used for this assessment. It is important to study such issues in a healthy population before extrapolating these results for use in patients. For application in day-to-day practice, subject compliance can be better ensured if the number of daily measurements is reduced. In one study requiring performance of two-hourly PEF measurements for one week, less than two-third subjects made at least four recordings on all seven days. Twice daily PEF measurements at waking and at bedtime would cause least inconvenience to the patients and ensure best compliance. However, as is clear from this study, the diurnal variation may be significantly underestimated if only these two readings are taken each day. From an operational point of view, we considered that the subjects would be most inconvenienced with recordings between 9 A.M. and 5 P.M. when they are likely to be busy at their place of work or study. We therefore included waking and bedtime readings in all schemes of partial measurement, and additionally evaluated the inclusion of one or more of other readings.
The optimal number of daily measurements needed to analyse PEF variability has not been established. Clinical studies have employed from two to twelve measurements each day. It is also not clearly known how much the magnitude or accuracy of estimates of diurnal variation is influenced by the number of measurements obtained. It is, however, well recognized that reducing the number of daily PEF measurements reduces the recorded variability. In one study using two-hourly PEF measurements, it was found that only 60-80% of actual variability was picked up with four daily readings, and only 20-45% with twice daily readings. We chose to measure PEF variability using at most five daily measurements as this number can pick up more than 80% of actual variability. Measuring PEF five times each day would be less cumbersome than recording two-hourly values, and would ensure better subject compliance, while missing only a small fraction of actual variability.
We expressed the diurnal variation as both A%M and SD%M. Of the several indices available, these two provide the best means of expressing PEF variability for epidemiological purposes. A%M is much simpler to calculate and most previous studies have used it as the index of diurnal variation. It has previously been reported that the effect of reducing the number of daily measurements may be different on the two indices. SD%M may be a better index, especially when calculating diurnal variation from less than five PEF measurements each day, and our observations are broadly similar. We have shown that the measured diurnal variability, as measured by A%M, decreases significantly as the number of daily measurements is reduced. When SD%M was used as the index of diurnal variation, the differences were less marked and a significant proportion of PEF variability could still be explained after removing one or more measurements. The mean SD%M actually marginally increased in some partial schedules of measurement. These differences between the two indices can be easily explained. If the dropped measures are outside the range of those considered, their exclusion results in a greater decrease in numerators than in denominator in those indices where numerators represent differences between readings (A%M) and not where numerators are standard deviations (SD%M).
We have also determined the limits of agreement between various schedules of measurement. Calculation of these limits may help an investigator in assessing if two measures of the same continuous variable can be used interchangeably in the clinical setting. In comparison to five daily measurements, all schedules of partial measurement produced wide limits of agreement. Such wide limits may not be clinically acceptable, even when comparisons between two schedules of measurement show no statistically significant differences. For example, using W92B schedule produces SD%M estimates which may be 1.26 less or 1.10 more than the value obtained from a profile of all five measurements; this difference is glaring if one recognizes that the mean SD%M value obtained from a complete profile is only 2.98 ? 1.31. Such discrepancy is not immediately apparent from simple tests of statistical significance, which otherwise point to a good agreement [Table:1].
We therefore conclude that reducing the number of daily recordings of PEF to less than five per day may produce estimates of diurnal variability that are in poor agreement with results obtained from five daily measurements, irrespective of the index used to describe variability. On reducing the number of readings, one is likely to miss a significant amount of variability when A%M is used. If a lesser number of readings are desired each day, the use of SD%M may be a better alternative.
Hetzel MR. The pulmonary clock. Thorax 1981;36:481-486.|
|2||D’Alonzo GE, Steinijans VW, Keller A. Measurements of morning and evening airflow grossly underestimates the circadian variability of FEV1 and peak expiratory flow rate in asthma. Am J Respir Crit Care Med 1995;152:1097-1099. |
|3||Zureik M, Liard R, Segala C, Henry C, Korobaeff M, Neukirch F. Peak expiratory flow rate variability in population surveys. Does the number of assessments matter? Chest 1995;107:418-423. |
|4||Aggarwal AN, Gupta D, Chaganti S, Jindal SK. Diurnal variation in peak expiratory flow in healthy young adults. Indian J Chest Dis Allied Sci 2000:42:15-19. |
|5||Quackenboss JJ, Lebowitz MD, Krzyzanowski M. The normal range of diurnal changes in peak expiratory flow rates. Relationship to symptoms and respiratory disease. Am Rev Respir Dis 1991;143:323-330. |
|6||Higgins BG, Britton JR, Chinn S, Jones TD, Jenkinson D, Burney PG, et al. The distribution of peak expiratory flow variability in a population sample. Am Rev Respir Dis 1989;140:1368-1372. |
|7||Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307-310. |
|8||Britton J. Measurement of peak flow variability in community populations: methodology. Eur Respir J 1997;24:42S-44S.