Reproducibility of the airway response to an exercise protocol standardized for intensity, duration, and inspired air conditions, in subjects with symptoms suggestive of asthma

Background Exercise testing to aid diagnosis of exercise-induced bronchoconstriction (EIB) is commonly performed. Reproducibility of the airway response to a standardized exercise protocol has not been reported in subjects being evaluated with mild symptoms suggestive of asthma but without a definite diagnosis. This study examined reproducibility of % fall in FEV1 and area under the FEV1 time curve for 30 minutes in response to two exercise tests performed with the same intensity and duration of exercise, and inspired air conditions. Methods Subjects with mild symptoms of asthma exercised twice within approximately 4 days by running for 8 minutes on a motorized treadmill breathing dry air at an intensity to induce a heart rate between 80-90% predicted maximum; reproducibility of the airway response was expressed as the 95% probability interval. Results Of 373 subjects challenged twice 161 were positive (≥10% fall FEV1 on at least one challenge). The EIB was mild and 77% of subjects had <15% fall on both challenges. Agreement between results was 76.1% with 56.8% (212) negative (< 10% fall FEV1) and 19.3% (72) positive on both challenges. The remaining 23.9% of subjects had only one positive test. The 95% probability interval for reproducibility of the % fall in FEV1 and AUC0-30 min was ± 9.7% and ± 251% for all 278 adults and ± 13.4% and ± 279% for all 95 children. The 95% probability interval for reproducibility of % fall in FEV1 and AUC0-30 min for the 72 subjects with two tests ≥10% fall FEV1 was ± 14.6% and ± 373% and for the 34 subjects with two tests ≥15% fall FEV1 it was ± 12.2% and ± 411%. Heart rate and estimated ventilation achieved were not significantly different either on the two test days or when one test result was positive and one was negative. Conclusions Under standardized, well controlled conditions for exercise challenge, the majority of subjects with mild symptoms of asthma demonstrated agreement in test results. Performing two tests may need to be considered when using exercise to exclude or diagnose EIB, when prescribing prophylactic treatment to prevent EIB and when designing protocols for clinical trials.


Background
Exercise is a widely recognised stimulus for provoking transient airway narrowing. Exercise-induced bronchoconstriction (EIB) is the term used to describe this phenomenon. The most commonly used measure to express severity of EIB is the post-exercise fall in forced expiratory volume in one second (FEV 1 ), as a percentage of the pre-exercise value [1]. A ≥10% fall in FEV 1 is reported to provide the best discrimination between asthmatic and normal responses in laboratory based running tests [2]. It is also the value suggested as the cut off for a positive test in the ATS and ERS guidelines for testing for EIB [3,4]. A second index of EIB severity is the area under the % fall in FEV 1 time curve (AUC 0-30 min ), which summarizes the extent and duration of bronchoconstriction. This second index is used to assess the benefit of medications that enhance recovery to a greater extent than their benefit on the immediate post exercise fall in FEV 1 [5]. The AUC 0-30 min reflects the contribution of the numerous mediators involved in EIB [6,7].
EIB commonly occurs in people with clinically recognized asthma [8] and has been reported in school children, elite athletes, and military recruits without other clinical signs and symptoms of asthma [9][10][11]. EIB is often the first indication of asthma [12] so it is important to diagnose and then treat underlying asthma recognized by exercise intolerance. We recently studied and reported a large number of adults and children with signs and symptoms suggestive of asthma but without a definitive diagnosis [13]. The study investigated sensitivity and specificity of airway responsiveness to methacholine and mannitol to identify EIB and a physician diagnosis of asthma [13]. The study examined duplicate controlled exercise challenges in 373 subjects and the data provided an opportunity to examine reproducibility of the airway response to exercise in the type of individual most likely to be referred for exercise testing for EIB.
Exercise testing to identify EIB in the laboratory is affected by the type of exercise, intensity and duration of exercise, inspired air conditions, baseline lung function and time since last medication or exercise. This paper reports the reproducibility of the % fall in FEV 1 and AUC 0-30 min in response to an exercise protocol that carefully controlled these variables.

Subjects: Inclusion/Exclusion Criteria
Subjects were enrolled if they were aged 6-50 years with a BMI of <35, and reported signs and symptoms suggestive of asthma according to the National Institute of Health (NIH) Questionnaire [14]. They were required to have an FEV 1 ≥70% of the predicted value at the Screening Visit [15,16]. Subjects were required to have a National Asthma Education and Prevention Program (NAEPPII) asthma severity score of Step 1 with neither a firm diagnosis of asthma nor an exclusion of the diagnosis of asthma.
Step 1 of NAEPPII is the mildest and is defined as symptoms ≤2 times per week, asymptomatic and normal peak expiratory flow measurements between exacerbations, exacerbations from only a few hours to a few days, night time symptom frequency of ≤ 2 times per month, FEV 1 or PEF ≥80% predicted and PEF variability ≤20%.
Subjects were excluded from participation if they: had any known other pulmonary disease; had smoked more than 1 cigarette per week within the past year or had a ≥10 pack year smoking history; had a respiratory tract infection within the previous 4 weeks; had been skin test positive to aeroallergens that were present in the environment during the time of enrolment and reported worsening of symptoms when exposed to these aeroallergens during the study; had been diagnosed at the Screening Visit as definitively (95 to 100% likelihood) having or not having asthma; had clinically significantly abnormal chest x-ray or ECG; or had failed to observe washout time of medications that would interfere with exercise (including, but not limited to, no use of corticosteroids within 4 weeks of the Screening Visit).
The disposition of the study population is given in Figure 1. The data presented are from the 375 subjects in the per protocol population that included all subjects with no major protocol violations previously reported [13]. Of the 375 subjects, two completed only one exercise challenge leaving 373 who completed two exercise tests; there were 95 children and 278 adults.

Procedures
The protocol was approved by institutional review boards and performed at 25 sites in the USA. Each subject or parent gave written informed consent or assent for minors <18 years of age. At screening the following were assessed: eligibility; demography; medical history; medications; spirometry with reversibility (following 360 mcg of albuterol/salbutamol from a pressurised metered dose inhaler); and allergy skin-prick testing to 10 common allergens (positive test taken as a wheal size ≥3 mm of the control). The NIH NAEPPII Questionnaire was administered and a score was assigned.
Exercise was performed on two separate occasions beginning 1 -4 days after the screening visits and within 2 hrs of the same time of day. Medication withholding was confirmed (Table 1), and spirometry was measured to determine consistency with values obtained at screening as previously described [13]. The exercise was performed on consecutive visits (2 and 3) with the second challenge being in 1 -4 days after the first. FEV 1 needed to be >70% predicted and within 15% of FEV 1 at screening in order for an exercise challenge to be performed.

Exercise protocol
Exercise was performed by running on a motorized treadmill while breathing medical grade dry air (20-25°C) from a reservoir (Douglas Bag) via a two-way non-rebreathing valve [17]. Subjects began by walking then running with the treadmill speed at 2.5 mph with 2.5% incline. Speed and incline were increased over 2 minutes so that heart rate (HR) reached 80-90% of predicted maximum (220age) and then was maintained for 6 minutes for a total duration of 8 minutes. This intensity aimed to achieve a ventilation rate between 14 and 21 times FEV1 L values that represent between 40 and 60% of maximum predicted ventilation (35 × FEV 1 ) [18]. The challenge could be stopped at any time. HR was monitored during and for 30 min after exercise. FEV 1 and FVC were measured before and FEV 1 (not FVC) was measured 5, 10, 15, and 30 minutes after exercise. The % fall in FEV 1 was calculated by subtracting the lowest value recorded after exercise taking the best of two acceptable attempts at each time point, from the value measured immediately before exercise, expressed as a percentage of the pre-exercise value. Values were not rounded; a 9.99% fall was considered negative. A subject was deemed positive if there was a fall of ≥10% in FEV 1 at one time point on at least one of the two exercise challenges [3,4]. Values are reported as mean and standard deviation (SD). Values for FEV 1 post-exercise that remained higher than the pre-exercise value were censored as 0% falls. The AUC 0-30 min was calculated by the trapezoidal method [19] and expressed as % fall in FEV 1 min -1 .
Spirometry data were captured by using ClinDataLink® (CDL) (CompleWare Corporation, North Liberty, IA) and met or exceeded the requirements proposed by American Thoracic Society/European Respiratory  Society Joint Statement [20]. Calibration was verified each day at three flow rates before use. WebCDL® software displayed an electronic record of the volume-time curves, flow-volume displays, and flow-time displays.
An estimate was made of ventilation in the 2 nd and 6 th minutes of exercise based on the relationship between speed and incline of treadmill and oxygen consumption in ml [21]. The ventilatory equivalent was estimated as 27 L per L of VO 2 [22], and ventilation was expressed as % of maximum voluntary ventilation (MVV). The estimate of oxygen consumption in mls was: 1.262*weight*(3.5 + (5.36*speed) + (0.24*speed*incline)) for running 1.262*weight*(3.5 + (2.68*speed) + (0.48*speed*incline)) for walking.
Weight is expressed in kilograms and speed is expressed in miles per hour. Three miles per hour was taken to be running.

Statistical Analysis
Reproducibility of the exercise test response was illustrated using a Bland-Altman-type plot [23] and calculated using the method of Chinn [24]. In brief, the standard deviation of a single measurement was calculated by dividing the standard deviation of the differences in % fall in FEV 1 values between the two tests (i.e. 7.6 for the whole group) by the square root of 2 giving a 5.4% fall, from which we calculated a 95% probability interval of ± 10.8%. This interval defines a 95% probability that the difference between any single measurement and the true value for the subject is within that range. This gives information about variability of the response that can be expected in an individual with repeated testing.

Demography
For the per protocol population (n = 375): females comprised 51.5%; subjects were 76.3% Caucasian, 8.3% Hispanic and 8.5% Black; subjects had near-normal baseline spirometry (Table 2); and 7.2% responded positively to a bronchodilator with ≥12% and ≥200 ml increase in FEV 1 above baseline. The characteristics of the 95 children and 278 adults are summarised in Table 2. The mean NAEP-PII asthma score was 1.22 (SD 0.52) for the adults and 1.21 (0.48) for the children. Positive skin tests to at least one allergen were seen in 78% of the adults and children.

Reproducibility of the Response
The 373 subjects who completed two exercise challenges did so within 2.6 ± 3.2 (median 2) days. The agreement The reproducibility (95% probability value) of the % fall in FEV 1 and the AUC % fall in FEV 1 min -1 for the whole group and for adults and children separately are given in Table 3, together with mean and highest falls in FEV 1  The reproducibility of the exercise response in relation to the different NAEPPII scores is given in Table 3. There was no relationship between the NAEPII score and the severity of the response to exercise expressed as the % fall in FEV 1 after exercise ( Figure 4).

Exercise Response
Post-exercise, 163 of the 375 subjects had ≥10% fall in FEV 1 (mean % fall ± SD was 19.1% ± 9.25 or 610 ± 330 ml) after at least one exercise challenge with 86 having ≥ 15% and 56 ≥ 20% fall in FEV 1 . Those 77 with very mild EIB i.e. 10 to 15% fall in FEV 1 had a mean fall of 12.3% ± 1.5 or 395 ± 116 ml. The distribution of the values for the maximum % fall in FEV 1 is given in The mean values for % fall in FEV 1 for adults and children and for those with two negative (< 10% fall), two positive (≥10% fall) and one positive and one negative test on each occasion are illustrated Figure 6a. AUC 0-30 min associated with these % falls in FEV 1 is given in Figure 6b. There was no significant difference in the response to exercise between adults and children. There was a significant correlation between the maximum % fall in FEV 1 and the corresponding 'maximum' AUC 0-30 min (r = 0.87, p < 0.001).

Work Load
The exercise load was similar on both tests days. Exercise resulted in a HR, % predicted maximum at 2 and 6 minutes of 82.1% ± 5.6 and 86.6% ± 8.9 on Day 1 and of 81.5% ± 6.7 and 89.9% ± 6.5 on Day 2 in adults (p = NS) and 81.9% ± 5.7 and 85.9% ± 10.3 on Day 1 and 81.8% ± 6.3 and 86.7% ± 4.9 on Day 2 in children (p = NS). There was no significant difference in the estimated ventilation expressed as a % of maximum voluntary ventilation between Days 1 and 2 for either the adults (Day 1 at 2 min 56.8% ± 15.3 and Day 2 58.0% ± 15.2) and children (Day 1 at 2 min 54.7% ± 13.1 and Day 2 56.3% ± 11.9). There was no significant difference in the HR, % of predicted maximum at 2 and 6 minutes on the day of the highest percent fall in FEV 1 of 82.0% ± 5.0 and 87.4% ± 5.0 in adults, and 82.4% ± 5.1 and 86.9% ± 5.1 in children.
The distribution of the estimated ventilation as % of MVV during the exercise is shown in Figure 7. The mean estimated ventilation calculated as a percent of maximum voluntary ventilation during the 2 nd and 6 th minute of the exercise with the highest fall in FEV 1 was 57.3% ± 14.5 and 53.1% ± 12.9 for adults and 54.6% ± 12.9 and 51.1% ± 11.0 for the children. The estimated ventilation as % of MVV on the 2 nd exercise test showed a small (+1.21% MVV) though significantly (< 0.009) higher value compared with the 1 st test for adults and a small (+1.35% MVV) but not significantly (P < 0.052) different value for children.
There was no significant difference between the HR % predicted and estimated ventilation % MVV between the test on the day the highest % fall in FEV 1 was documented, and on the test on the day the lowest % fall in FEV 1 was recorded for the different groups of subjects (data not shown). There was also no significant difference in baseline FEV 1 % predicted for the two days in the group where the % falls in FEV 1 ≥10% with both tests. The FEV 1 % predicted was higher on the day of the highest % fall in FEV 1 for all the other groups; however, the baseline values for FEV 1 % predicted were always above 90% and all the differences were less than 2.4% predicted.

Discussion
One problem in using an exercise challenge to identify EIB in the laboratory is ensuring that intensity of  The interval defines the 95% probability that the difference between a single measurement and the true value for the subject is within that range. exercise, exercise duration, and condition of the inspired air are controlled and are adequate for eliciting the EIB response. In this multicentre study exercise duration was 8 minutes, inspired air was dry, and intensity of exercise was sufficient for HR to reach the value required by the protocol, i.e. 80-90% predicted maximum by the 2 nd minute of exercise and HR was not significantly different on the two test days. Appropriate times for withdrawal of medications were verified and pre-exercise FEV 1 was >70% predicted in all but 2 subjects (both children) and it was similar on both occasions (and was actually greater than a mean of 90%). No subject had taken inhaled corticosteroids within the last 4 weeks, or long or short-acting beta 2 agonist for 48 hours or 8 hours, respectively. Minimising the difference in these variables between tests allowed us to examine the natural variation of the airway response within a few days. We used one time point ≥10% fall to identify a positive test because this has been common practice. However we allowed a period of 5 minutes for recovery before the first FEV 1 was measured. We excluded those who were symptomatic to the allergens to which they tested positive to a skin test at the time to reduce variability due to environmental factors. We are unaware of any other study that has given this level of attention to variables when performing two exercise challenges to identify EIB. Knowledge about normal variation in the exercise response is critically important when interpreting a negative test or when evaluating an exercise response to a therapeutic agent. The ventilation reached and sustained during exercise is a primary determinant of the % fall in FEV 1 [4]. However equipment for measuring ventilation during exercise is expensive and heart rate has been preferred to confirm the intensity of exercise in the United States of America. To ensure that subjects reached the minimum ventilation (40% of MVV recommended by other protocols [4]) we made an estimate of oxygen consumption from the speed and slope of the treadmill and the weight of the subject protocols and assumed a ventilatory equivalent of 27L of ventilation per L of VO 2 using published equations [4]. This target ventilation was achieved between by the 2 nd minute of exercise and MVV exceeded 50% in the majority of adults and children. While a direct measurement of ventilation would have been preferable the estimated values, based on the work load and expressed as a % MVV, at 2 min and  6 min were the same as the values measured in adults during 8 minutes of bicycle exercise [25].
As may have been expected from a group of patients without a definitive diagnosis of asthma, the response to exercise, when positive, was mild and 77% of the subjects had a fall in FEV 1 < 15% on both exercise challenges. In only 34 of 161 subjects did a ≥15% fall occur on both exercise challenges, a frequency probably consistent with their mild symptoms and indefinite diagnosis of asthma. A fall in FEV 1 after exercise of ≥20% is the value suggested for inclusion in clinical trials to evaluate a drug for EIB (FDA Guidance for Industry, http://www.fda.gov./cder/guidance). This value occurred on two exercise challenges in only 19 of the 161 subjects (11.8%) with EIB in this study or only in 5.1% of the subjects who were exercised twice.
For those who had two exercise challenges with falls greater than 10%, the mean maximum fall after exercise was 24.7% ± 9.7, leaving little doubt about a diagnosis of EIB. The reproducibility of the response in this group was ±14.6% and compares well with the value of ±15.8% calculated in adults with an established diagnosis of asthma performing repeated exercise on a cycle ergometer [25].
We assigned a value of 0% fall for those demonstrating only a rise in FEV 1 in response to exercise; a post-exercise fall is characteristic of asthma while a post-exercise rise in FEV 1 is not and occurs in many non-asthmatic subjects [26]. The mean maximum fall in FEV 1 plus 2SDs (4.9% ± SD 2.9) for the group with two negative challenges (e.g. those who had <10% fall in FEV 1 on both challenges) was 10.7% and similar to that reported for groups of normal adults or children, without a history of symptoms of asthma, exercising in ambient air in a laboratory [2,10,27]. Thus, subjects with an NAEPPII asthma severity score of ≥1 can have a reproducible response to exercise similar to that of a healthy subject with no history of asthma The study results confirm that there is little difference between adults and children for the indices used to express EIB and we used a value of 10% in both groups. However higher cut-off values have been recommended to identify EIB in children [28,29]. Using the 15% cut point recommended by Haby [28], the prevalence of EIB in the children was reduced from 51.5% (49/95) to 28.4% (27/95). We consider that the 5 times difference in the degree of EIB in those with ≥10% fall in FEV 1 (24.7% ± 9.7) on both occasions and those with ≤10% fall on both occasions (4.9% ± 2.9) supports the use of a 10% cut-off to include or exclude a definitive diagnosis of EIB when challenges are repeated over a short period.
We used a cut off point of ≥10% fall in FEV 1 to analyse the AUC 0-30 min and its reproducibility. There was also >5 times difference in the AUC 0-30 min between those with two challenges with ≥10% fall in FEV 1 (-525 ± 245% FEV 1 min -1 ) compared with those with two challenges with <10% fall in FEV 1 (-89 ± 75% FEV 1 min -1 ). Based on the mean plus 2SDs in those with two challenges with <10% fall in FEV 1 , we suggest an upper cut-off value for AUC 0-30 min of 240% fall in FEV 1 min -1 for a negative test. The utility of having values for the reproducibility of AUC 0-30 min is that there are drugs such as montelukast that have limited effect on the maximum % fall in FEV 1 but have a profound benefit in enhancing recovery of FEV 1 to baseline [5]. In keeping with others [30] who reported a smaller group of known asthmatic subjects over a longer period, the values for reproducibility of the % fall in FEV 1 were superior to the AUC 0-30 min .
In the 89 subjects positive on only one challenge ( Figure 6) we considered that this variation may have been due to a change in the intensity of exercise on the two test days or perhaps other characteristics of this group. However the variation in the % fall in FEV 1 on the two test days was not explained by differences in the ventilation % MVV, HR % predicted maximum. The FEV 1 % predicted was significantly higher (p < 0.02) on the day of the positive challenge (92.1% ± 11.3) compared with the day of the negative test (90.2% ± 11.1) although the difference was small. The variability between a positive and negative test result may be due to other factors, perhaps environmental or dietary, or simply the intrinsic reproducibility of the test itself.
The study group had mild symptoms and signs suggestive of asthma but the NAEPPII grading could not be relied upon either to identify EIB or to predict its severity or reproducibility of the response. However, the NAEPPII is a score of asthma severity [14] and does not necessarily include symptoms provoked by exercise. This may not be important in that other investigators who have questioned subjects specifically about exercise Anderson et al. Respiratory Research 2010, 11:120 http://respiratory-research.com/content/11/1/120 symptoms have found symptoms alone to be unreliable predictors of either presence or severity of EIB [10,31,32].
The data presented here are a secondary analysis of a previously reported study (NCT00252291) [13]. The protocol required two exercise challenge tests to be performed under the same controlled conditions on consecutive visits prior to a mannitol and a methacholine challenge. All but two subjects of the 375 in the previously reported study performed two challenge tests. For these reasons this study offered an ideal opportunity to determine reproducibility of the response to exercise in a large group in an unbiased manner.
The usefulness of these data are not only in understanding that more than one test may be required to exclude a diagnosis of EIB but also in determining the benefit of treatment or how severe EIB should be for inclusion in a drug trial. For example the variability in the % fall in FEV 1 as expressed by the 95% probability value for subjects with two tests ≥20% was 14.3% and the mean % fall in FEV 1 was 30.3%. That means that on a second test a subject with a fall of 30.1% on initial testing would fall 30.1% ± 14.3% (range 44.4-15.8%) on a second occasion exercising under identical conditions within a few days. Thus for a drug to be regarded as beneficial the % fall would need to be less than 15.8% on repeated challenge.
In our subjects with mild symptoms of asthma, good lung function, and a low response rate to bronchodilator, a single exercise test did not rule out mild EIB and a second exercise test under the same conditions identified an extra 44 subjects, 27% of the total positive, with ≥10% fall in FEV 1 . It is unlikely that repeat exercise challenge is useful in those recording a rise in FEV 1 on the initial challenge, as the chance of being positive on the second test was low and, even when the exercise challenge was positive, the falls in FEV 1 were very mild.

Conclusions
The majority of subjects with signs and symptoms suggestive of asthma without a definitive diagnosis will have the same outcome i.e. positive or negative test result following rechallenge when exercise is standardized for intensity, duration, and condition of the inspired air. However a minority will have a positive test result on only one exercise test. These data also show that for most subjects the EIB will be mild (< 15% fall in FEV 1 ) and particularly so for those positive on a second challenge after the first exercise challenge was negative. This study provides evidence for the degree of variability in response to duplicate exercise challenges and suggests that for some subjects with mild symptoms more than one test may be required before either a diagnosis of EIB is excluded or prophylactic treatment is prescribed. Finally, these data in a large number of adults indicate that the reproducibility of the response in adults is similar to that observed in children. Competing interests SDA is the inventor of the mannitol test however the intellectual property is owned by her employer, the Sydney South West Area Health Service (SSWAHS). SDA receives a 10% share of the royalties paid to SSWAHS. SDA has undertaken research studies that were funded by Pharmaxis. She is a shareholder in Pharmaxis but holds no options. She acts as a consultant to Pharmaxis for which she has received fees since April 2009. DSP, KWR, HB, CAS participated in the study through their respective centers (see below) that received a research grant for study participation from Pharmaxis Ltd. CPP owns shares in Pharmaxis Ltd which she herself has purchased. She has also acted as a paid consultant to Pharmaxis SN is the statistician employed by CompleWare and carried out the statistical analysis. JW is the President of, and is a shareholder in, CompleWare Corporation. CompleWare received a fee from Pharmaxis Ltd. for services in carrying out the clinical trial. There are no other competing interests or conflicts of interest.