- Email: [email protected]
CALIBRATION OF SOME WRIGHT PEAK FLOW METERS
Br.J. Anaesth. (1980), 52, 461
CALIBRATION OF SOME WRIGHT PEAK FLOW METERS J. FISHER AND A. SHAW SUMMARY
In both its original (Wright and McKcrrow, 1959) and restyled form the Wright peak flow meter (p.fjn.) is widely accepted for measurement of peak expiratory flow rate. Recently a smaller, lowermost device, the mini p.f.m., has been introduced (Wright, 1978). Two calibration methods have been used to check the various designs of p.f.m. Steady air flows have been blown through a flow meter connected in series with the meter. This method does not test the response of the p i j n . to non-steady airflows. Also, it depends on a flow meter which cannot readily be checked for its absolute precision. The other method compares readings from the meter with those obtained from a pneumotachograph for a single group of people. This presupposes the pneumotachograph to be properly calibrated. More important, it is imprecise because subjects do not always produce the same peak flow. In a recent report (Perks et al., 1979) the scatter of results was considerable, particularly at low flows. Because of these deficiencies, we developed a calibration procedure which can be cross-checked independently. We report our findings for the three types of Wright p.f.m.: the original p.f.m., the restyled p.f.m. and the mini p.f.m.
A technical description and appraisal of our calibration procedure has been reported elsewhere (Shaw, Fisher and Hughes, 1979). A heavy piston is allowed to fall under gravity in a cylinder. Air is expelled from the lower end of the cylinder to the instrument being
calibrated. Sharp-edged orifices manufactured to British Standard 1042 are installed in the cylinder above the piston. These orifices control the acceleration of the piston until a constant velocity and therefore a steady flow has been reached. Various sizes of orifice are used to generate particular peak flows. The flow is checked independently by a timing unit which measures the velocity of the piston during its descent. Three original p.f.m., 10 restyled p.f.m. and six mini p.f.m. were tested in the survey. Four of the mini p.f.m. had repositioned scales as recommended by Wright (1978) and Perks and others (1979). All meters had been in use in hospitals in the West of Scotland, except two mini p.f.m. (A and C) which were assessed immediately after purchase. There had been no apparent abuse to any meters. Two restyled p.f.m. had just been calibrated by the manufacturer. Ten different peak flows were produced and measured for each p.fjn. Each peak flow was reproduced three times. The dynamic characteristic of every meter was determined by progressively shortening the period of time at which the peak flow rate was maintained. This was done by stopping the piston of the calibrator before it had completed its full stroke. It has been suggested that maintaining a peak flow for 10 ms was sufficient for the original p.f.m. to respond accurately (Wright and McKerrow, 1959). The pressure loss across each meter was measured using a differential pressure transducer (Elema Schonander E.M.T. 32). Care was taken to avoid air leakage between the calibrator and meter.
J. FISHER, B J C ; A. SHAW, B ^ C , C.ENG.; West of Scotland
RESULTS
MATERIALS AND METHODS
Health Boards, Department of Clinical Physics and BioThe measured peak flows were compared with the Engineering, 11 West Graham Street, Glasgow G4 9LF. Correspondence to A. S. readings on each meter. Percentage errors rather 0007-0912/80/040461-O4 J01.00
© Macmillan Journals Ltd 1980
Downloaded from http://bja.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on July 3, 2015
The performance of Wright peak flow meters is described using an absolute and precise calibration method. This showed errors as great as 25% with the peak flow meter and 40% with the mini peak flow meter. At large flow rates errors were less. The expected variation between individual peak flow meters was about 12%. The expected variation between individual mini peak flow meters was between 16% and 22%, depending on flow rate.
BRITISH JOURNAL OF ANAESTHESIA
462
TABLE I. Percentag4 trrortfor the restyled p.fjn. Meter number
2
1 0 10 13 13 11 4.1 2.2 -1.0 -2.8 -8.0
4.0 18 17 13 9.0 4.0 2.0
-2.0 -6.0 -12
3
4
5
6
12 16 20 16 14 7.9 4.5 2.3
12 17 18 16 11 6.5 3.7 0.2
13 17 21 19 14 10 4.8 2.0
-1.7 -8.0
-2.7 -9.3
-0.9 -8.1
13 23 24 21 17 12 9.0 4.3 0.5 -12
than absolute values were used. The percentage error was defined as: % error
(meter reading)—(actual flow) xlOO (actual flow)
Table I shows the per cent error of the restyled pXm. Meters 4 and 6 had just been re-calibrated by the manufacturer. Table II shows per cent errors for the mini p.f.m. Meters A, C, D and E had repositioned scales (Perks et al., 1979). TABLE II. Percentage errors for the mini p.f.m.
Meter number Actual flow (litre min"1) * &WU4U
^ ^ ^ ^ ^T
103 155 174 215 254 278 295 354 430 525
A 7 17 22 22 20 19 18
15 9.2 4.5
B
C
D
E
F
21 25 24 24 19 18 17 12 7.0 3.0
25 36 35 30 26 25 22 19 13 7.0
25 38 37 34 35 33 33 24 17 12
27 36 33 32 28 28 26 23 17 10
35 39 34 32 27 25 24 20 13 8.0
Figures 1 and 2 show examples of calibration curves for a single restyled p.f.m. and a single mini p.f.m. There were substantial differences in the errors produced by each meter. From our data for the 10 restyled p.f.m. and six mini p.f.m., predicted 95% confidence limits were established. Figure 1 shows our predicted upper and lower limits for restyled p.fjn. and figure 2 similar data for the mini p.f.m. If our sample batch of each type of meter is typical of all meters presently in use, 95 out of 100 instruments would fall within these bands.
7 15 15 15 13 12 4.5 0.0
-3.8 -4.8 -10
8
9
10
16 19 20 19
23 23 23 18 13 7.7 4.4 13 -2.1 -10
25 27 26 20 17 14 11 9.1 5.5 -2.7
15 8.4 5.7 2.1 -0.5 -8.8
Reducing the period at which peak flow was maintained showed that at 30 ms the meter readings were reduced by about 5%. The pressure loss across the various meters at 400 litre min" 1 was about 15 cm H,O for the mini p.f.m. and 5 cm H a O for the restyled and original pXm. DISCUSSION
Our calibrator produces peak flows the values of which are known to an accuracy better than 1.0%. Repeatability of a particular peak flow based on 30 measurements is better than 0.28% (Shaw, Fisher and Hughes, 1979). Tables I and II show that there are substantial errors in the meter readings for the restyled and mini p.f.m., especially at the lower end of the flow range. Percentage errors for the three original p/.m. tested tended to be smaller than for the restyled p.f.m., but the pattern of errors was similar: per cent error was positive at slow flow rates and negative at greater flow rates. The tables show the wide variation in errors between individual restyled and mini p.f.m. Figure 1 shows that there is a specific calibration curve for the single meter illustrated (no. 3 in table I). For clarity, curves for the other meters are not shown, but they have similar profiles and are an equally good fit. The predicted confidence limits for all restyled meters shown in figure 1 depend upon our sample batch being typical of all instruments in use. Since our instruments were obtained from a wide variety of sources this assumption is reasonable. The calibration curves for two of the original p.f.m. were less than the 95% confidence limits for the restyled p.f.m. This suggests a change in the calibration of the p.f.m. when it was restyled. Curiously, two restyled
Downloaded from http://bja.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on July 3, 2015
/\ciuai now (litre min- 1 ) 120 175 230 305 334 395 445 510 575 730
463
CALIBRATION OF PEAK FLOW METERS +30-i
-20- 1
Fio. 1. Percentage errors for a restyled p.f.m. Dotted lines show 95% confidence limits for all the restyled p.f.m. tested.
40-
305 u
\
/
\
\
\
\ \
20* -
B
s
& 10-
s
\
/ 0"3
200
400
600
actual flow (litre min"1)
FIG. 2. Percentage errors for a mini p.f.m. Dotted lines show 95% confidence limits for all the mini p.f.m. tested.
p.f.m. with consecutive serial numbers (meters 4 and 5) had very similar characteristics. It is possible that the sample variation is reduced considerably if several instruments are bought at the same time. Figure 2 also shows a non-linear relationship between per cent error and flow for the mini p.f.m. The typical calibration curve shown is for meter C. Similar shaped curves are produced for the other meters from the data in table II. The 95% confidence limits show a big variation between the individual mini p.fjn. The characteristic variation is much greater than that found by Perks and others (1979). It is obvious from figure 2 that it is inappropriate to
use a straight-line regression analysis or to change the calibration simply by repositioning the scale (Wright, 1978; Perks et al., 1979). The original meter was expected to measure accurately peak flows of duration 10 ms or more. Our findings are that meter readings begin to be affected at a duration of about 30 ms. This dynamic response is sufficiently good to measure peak flow in most subjects. From our findings several conclusions can be reached. For the restyled p i j n . it is reasonable to compare results taken on different meters. Our predicted 95% confidence band is about 12% for flows greater than 300 litre min" 1 . At slower flows the characteristics of the individual p i j n . may have an increasing effect on the results. We would not recommend comparing peak flows measured on different mini pXm. at flow rates less than about 400 litre min" 1 . The absolute calibration of both designs of meter could be improved by alterations to the scale. Both designs of meter would require a new non-linear scale. This would not reduce the variation between individual meters. A unique calibration graph can be produced for every meter. In this way it is possible to correct the indicated reading, thus obtaining a highly accurate measurement of the actual peakflowgenerated. However, the T"'"' p i j n . is a greater resistance meter
Downloaded from http://bja.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on July 3, 2015
actual flow
464
BRITISH JOURNAL OF ANAESTHESIA
than is the restyled meter so, even with such corrections, the peak expiratory flows obtained should not be compared directly.
l'ordre de 12%. Les variations attendues des divers mini dibitmetres a maxima se situaient entre 16% et 22%, suivant le dibit.
REFERENCES
KALIBRffiRUNG EINIGER WRIGHT-SPITZENDURCHFLUSSMESSER
ETALONNAGE DE CERTAINS DEBITMETRES A MAXIMA DE WRIGHT
ZUSAMMENFASSUNG
Die Leistung von Wright-Spitzendurchflussmessern wird mittels einer absoluten und prazisen Kalibrierungsmethode beschrieben. Dabei zeigten sich Fehler bis zu 25% mit Spitzendurchflussmessem, und bis zu 40% mit dem MiniSpitzendurchflussmesser. Bei grossen Durchflussraten waren die Fehler geringer. Die erwarteten Variationen zwischen einzelnen Spitzendurchflussmessem betrugen etwa 12%. Die erwarteten Variationen zwischen einzelnen Mini-Spitzendurchflussmessern betrugen zwischen 16% und 22%, je nach der Durchflussrate. CALIBRACION DE ALGUNOS CONTOMETROS WRIGHT DE APICE SUMARIO
RESUME
On decrit dans cet article les performances de certains dibitmetres a nur'nm de Wright, mesurees a l'aide d'une mithode d'italonnage precise et absolue. Celle-ci a fait retsortir des erreurs aussi importantes que 25% dans le cas du dfbitmetre a maxima, et 40% dans le cas du mini dibirmetre a ma-rima A des dibits Clevis, les erreurs ont ttt moins importantes. Les variations auxquelles on s'attendait entre les divers dibitmetres a miTima itaient de
Se da una resena del fundonamiento de los cont6metros de apice Wright al usar un mithodo de calibraci6n preciso y absolute Este puso de manifiesto errores de hasta 25% del cont6metro de apice y hasta 40% del mini-contometro de apice. Con ritmos de flujo mayores, los errores eran menores. La vaxiaci6n anticipada entre los cont6metros de apice individuates era de alrededor del 12%. La variaci6n anticipada entre los mini-cont6metros de apice se ubicaba entre el 16 y el 22%, segun el ritmo del flujo.
Downloaded from http://bja.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on July 3, 2015
British Standard 1042. Methods for the Measurement of Fluid Flow in Pipes. Part 1. Orifice Plates, Nozzles and Venturi Tuba. Perks, W. H., Tarns, I. P., Thompson, D. A., and Prowse, K. (1979). An evaluation of the mini-Wright peak flow meter. Thorax, 34, 79. Shaw, A., Fisher, J., and Hughes, D. W. (1979). A flow and volume calibrator for respiratory measuring equipment J. Med. Eng. Technol., 3, 248. Wright, B. M. (1978). A miniature Wright peak flow meter. Br. Med. J., 2, 1627. McKerrow, C. B. (1959). Maximum forced expiratory flow rate as a measure of ventilatory capacity. Br. Med. J., 2, 1041.
CALIBRATION OF SOME WRIGHT PEAK FLOW METERS J. FISHER AND A. SHAW SUMMARY
In both its original (Wright and McKcrrow, 1959) and restyled form the Wright peak flow meter (p.fjn.) is widely accepted for measurement of peak expiratory flow rate. Recently a smaller, lowermost device, the mini p.f.m., has been introduced (Wright, 1978). Two calibration methods have been used to check the various designs of p.f.m. Steady air flows have been blown through a flow meter connected in series with the meter. This method does not test the response of the p i j n . to non-steady airflows. Also, it depends on a flow meter which cannot readily be checked for its absolute precision. The other method compares readings from the meter with those obtained from a pneumotachograph for a single group of people. This presupposes the pneumotachograph to be properly calibrated. More important, it is imprecise because subjects do not always produce the same peak flow. In a recent report (Perks et al., 1979) the scatter of results was considerable, particularly at low flows. Because of these deficiencies, we developed a calibration procedure which can be cross-checked independently. We report our findings for the three types of Wright p.f.m.: the original p.f.m., the restyled p.f.m. and the mini p.f.m.
A technical description and appraisal of our calibration procedure has been reported elsewhere (Shaw, Fisher and Hughes, 1979). A heavy piston is allowed to fall under gravity in a cylinder. Air is expelled from the lower end of the cylinder to the instrument being
calibrated. Sharp-edged orifices manufactured to British Standard 1042 are installed in the cylinder above the piston. These orifices control the acceleration of the piston until a constant velocity and therefore a steady flow has been reached. Various sizes of orifice are used to generate particular peak flows. The flow is checked independently by a timing unit which measures the velocity of the piston during its descent. Three original p.f.m., 10 restyled p.f.m. and six mini p.f.m. were tested in the survey. Four of the mini p.f.m. had repositioned scales as recommended by Wright (1978) and Perks and others (1979). All meters had been in use in hospitals in the West of Scotland, except two mini p.f.m. (A and C) which were assessed immediately after purchase. There had been no apparent abuse to any meters. Two restyled p.f.m. had just been calibrated by the manufacturer. Ten different peak flows were produced and measured for each p.fjn. Each peak flow was reproduced three times. The dynamic characteristic of every meter was determined by progressively shortening the period of time at which the peak flow rate was maintained. This was done by stopping the piston of the calibrator before it had completed its full stroke. It has been suggested that maintaining a peak flow for 10 ms was sufficient for the original p.f.m. to respond accurately (Wright and McKerrow, 1959). The pressure loss across each meter was measured using a differential pressure transducer (Elema Schonander E.M.T. 32). Care was taken to avoid air leakage between the calibrator and meter.
J. FISHER, B J C ; A. SHAW, B ^ C , C.ENG.; West of Scotland
RESULTS
MATERIALS AND METHODS
Health Boards, Department of Clinical Physics and BioThe measured peak flows were compared with the Engineering, 11 West Graham Street, Glasgow G4 9LF. Correspondence to A. S. readings on each meter. Percentage errors rather 0007-0912/80/040461-O4 J01.00
© Macmillan Journals Ltd 1980
Downloaded from http://bja.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on July 3, 2015
The performance of Wright peak flow meters is described using an absolute and precise calibration method. This showed errors as great as 25% with the peak flow meter and 40% with the mini peak flow meter. At large flow rates errors were less. The expected variation between individual peak flow meters was about 12%. The expected variation between individual mini peak flow meters was between 16% and 22%, depending on flow rate.
BRITISH JOURNAL OF ANAESTHESIA
462
TABLE I. Percentag4 trrortfor the restyled p.fjn. Meter number
2
1 0 10 13 13 11 4.1 2.2 -1.0 -2.8 -8.0
4.0 18 17 13 9.0 4.0 2.0
-2.0 -6.0 -12
3
4
5
6
12 16 20 16 14 7.9 4.5 2.3
12 17 18 16 11 6.5 3.7 0.2
13 17 21 19 14 10 4.8 2.0
-1.7 -8.0
-2.7 -9.3
-0.9 -8.1
13 23 24 21 17 12 9.0 4.3 0.5 -12
than absolute values were used. The percentage error was defined as: % error
(meter reading)—(actual flow) xlOO (actual flow)
Table I shows the per cent error of the restyled pXm. Meters 4 and 6 had just been re-calibrated by the manufacturer. Table II shows per cent errors for the mini p.f.m. Meters A, C, D and E had repositioned scales (Perks et al., 1979). TABLE II. Percentage errors for the mini p.f.m.
Meter number Actual flow (litre min"1) * &WU4U
^ ^ ^ ^ ^T
103 155 174 215 254 278 295 354 430 525
A 7 17 22 22 20 19 18
15 9.2 4.5
B
C
D
E
F
21 25 24 24 19 18 17 12 7.0 3.0
25 36 35 30 26 25 22 19 13 7.0
25 38 37 34 35 33 33 24 17 12
27 36 33 32 28 28 26 23 17 10
35 39 34 32 27 25 24 20 13 8.0
Figures 1 and 2 show examples of calibration curves for a single restyled p.f.m. and a single mini p.f.m. There were substantial differences in the errors produced by each meter. From our data for the 10 restyled p.f.m. and six mini p.f.m., predicted 95% confidence limits were established. Figure 1 shows our predicted upper and lower limits for restyled p.fjn. and figure 2 similar data for the mini p.f.m. If our sample batch of each type of meter is typical of all meters presently in use, 95 out of 100 instruments would fall within these bands.
7 15 15 15 13 12 4.5 0.0
-3.8 -4.8 -10
8
9
10
16 19 20 19
23 23 23 18 13 7.7 4.4 13 -2.1 -10
25 27 26 20 17 14 11 9.1 5.5 -2.7
15 8.4 5.7 2.1 -0.5 -8.8
Reducing the period at which peak flow was maintained showed that at 30 ms the meter readings were reduced by about 5%. The pressure loss across the various meters at 400 litre min" 1 was about 15 cm H,O for the mini p.f.m. and 5 cm H a O for the restyled and original pXm. DISCUSSION
Our calibrator produces peak flows the values of which are known to an accuracy better than 1.0%. Repeatability of a particular peak flow based on 30 measurements is better than 0.28% (Shaw, Fisher and Hughes, 1979). Tables I and II show that there are substantial errors in the meter readings for the restyled and mini p.f.m., especially at the lower end of the flow range. Percentage errors for the three original p/.m. tested tended to be smaller than for the restyled p.f.m., but the pattern of errors was similar: per cent error was positive at slow flow rates and negative at greater flow rates. The tables show the wide variation in errors between individual restyled and mini p.f.m. Figure 1 shows that there is a specific calibration curve for the single meter illustrated (no. 3 in table I). For clarity, curves for the other meters are not shown, but they have similar profiles and are an equally good fit. The predicted confidence limits for all restyled meters shown in figure 1 depend upon our sample batch being typical of all instruments in use. Since our instruments were obtained from a wide variety of sources this assumption is reasonable. The calibration curves for two of the original p.f.m. were less than the 95% confidence limits for the restyled p.f.m. This suggests a change in the calibration of the p.f.m. when it was restyled. Curiously, two restyled
Downloaded from http://bja.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on July 3, 2015
/\ciuai now (litre min- 1 ) 120 175 230 305 334 395 445 510 575 730
463
CALIBRATION OF PEAK FLOW METERS +30-i
-20- 1
Fio. 1. Percentage errors for a restyled p.f.m. Dotted lines show 95% confidence limits for all the restyled p.f.m. tested.
40-
305 u
\
/
\
\
\
\ \
20* -
B
s
& 10-
s
\
/ 0"3
200
400
600
actual flow (litre min"1)
FIG. 2. Percentage errors for a mini p.f.m. Dotted lines show 95% confidence limits for all the mini p.f.m. tested.
p.f.m. with consecutive serial numbers (meters 4 and 5) had very similar characteristics. It is possible that the sample variation is reduced considerably if several instruments are bought at the same time. Figure 2 also shows a non-linear relationship between per cent error and flow for the mini p.f.m. The typical calibration curve shown is for meter C. Similar shaped curves are produced for the other meters from the data in table II. The 95% confidence limits show a big variation between the individual mini p.fjn. The characteristic variation is much greater than that found by Perks and others (1979). It is obvious from figure 2 that it is inappropriate to
use a straight-line regression analysis or to change the calibration simply by repositioning the scale (Wright, 1978; Perks et al., 1979). The original meter was expected to measure accurately peak flows of duration 10 ms or more. Our findings are that meter readings begin to be affected at a duration of about 30 ms. This dynamic response is sufficiently good to measure peak flow in most subjects. From our findings several conclusions can be reached. For the restyled p i j n . it is reasonable to compare results taken on different meters. Our predicted 95% confidence band is about 12% for flows greater than 300 litre min" 1 . At slower flows the characteristics of the individual p i j n . may have an increasing effect on the results. We would not recommend comparing peak flows measured on different mini pXm. at flow rates less than about 400 litre min" 1 . The absolute calibration of both designs of meter could be improved by alterations to the scale. Both designs of meter would require a new non-linear scale. This would not reduce the variation between individual meters. A unique calibration graph can be produced for every meter. In this way it is possible to correct the indicated reading, thus obtaining a highly accurate measurement of the actual peakflowgenerated. However, the T"'"' p i j n . is a greater resistance meter
Downloaded from http://bja.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on July 3, 2015
actual flow
464
BRITISH JOURNAL OF ANAESTHESIA
than is the restyled meter so, even with such corrections, the peak expiratory flows obtained should not be compared directly.
l'ordre de 12%. Les variations attendues des divers mini dibitmetres a maxima se situaient entre 16% et 22%, suivant le dibit.
REFERENCES
KALIBRffiRUNG EINIGER WRIGHT-SPITZENDURCHFLUSSMESSER
ETALONNAGE DE CERTAINS DEBITMETRES A MAXIMA DE WRIGHT
ZUSAMMENFASSUNG
Die Leistung von Wright-Spitzendurchflussmessern wird mittels einer absoluten und prazisen Kalibrierungsmethode beschrieben. Dabei zeigten sich Fehler bis zu 25% mit Spitzendurchflussmessem, und bis zu 40% mit dem MiniSpitzendurchflussmesser. Bei grossen Durchflussraten waren die Fehler geringer. Die erwarteten Variationen zwischen einzelnen Spitzendurchflussmessem betrugen etwa 12%. Die erwarteten Variationen zwischen einzelnen Mini-Spitzendurchflussmessern betrugen zwischen 16% und 22%, je nach der Durchflussrate. CALIBRACION DE ALGUNOS CONTOMETROS WRIGHT DE APICE SUMARIO
RESUME
On decrit dans cet article les performances de certains dibitmetres a nur'nm de Wright, mesurees a l'aide d'une mithode d'italonnage precise et absolue. Celle-ci a fait retsortir des erreurs aussi importantes que 25% dans le cas du dfbitmetre a maxima, et 40% dans le cas du mini dibirmetre a ma-rima A des dibits Clevis, les erreurs ont ttt moins importantes. Les variations auxquelles on s'attendait entre les divers dibitmetres a miTima itaient de
Se da una resena del fundonamiento de los cont6metros de apice Wright al usar un mithodo de calibraci6n preciso y absolute Este puso de manifiesto errores de hasta 25% del cont6metro de apice y hasta 40% del mini-contometro de apice. Con ritmos de flujo mayores, los errores eran menores. La vaxiaci6n anticipada entre los cont6metros de apice individuates era de alrededor del 12%. La variaci6n anticipada entre los mini-cont6metros de apice se ubicaba entre el 16 y el 22%, segun el ritmo del flujo.
Downloaded from http://bja.oxfordjournals.org/ at University of Iowa Libraries/Serials Acquisitions on July 3, 2015
British Standard 1042. Methods for the Measurement of Fluid Flow in Pipes. Part 1. Orifice Plates, Nozzles and Venturi Tuba. Perks, W. H., Tarns, I. P., Thompson, D. A., and Prowse, K. (1979). An evaluation of the mini-Wright peak flow meter. Thorax, 34, 79. Shaw, A., Fisher, J., and Hughes, D. W. (1979). A flow and volume calibrator for respiratory measuring equipment J. Med. Eng. Technol., 3, 248. Wright, B. M. (1978). A miniature Wright peak flow meter. Br. Med. J., 2, 1627. McKerrow, C. B. (1959). Maximum forced expiratory flow rate as a measure of ventilatory capacity. Br. Med. J., 2, 1041.