In contrast, both the Expert Panel [4] and the British Thoracic Society [5] recommend the routine use of oral or intravenous theophylline in adults admitted to the hospital for acute exacerbation of asthma, despite the lack of controlled studies. Because theophylline has a narrow therapeutic index and its use can result in serious adverse effects when dosage is not carefully adjusted on the basis of serum concentration measurements [6], its use can be justified only if a benefit is clear, especially since the advent of inhaled ß₂-selective agonists and the early use of high-dose corticosteroids [7]. Because the observation periods in two of the studies questioning aminophylline's efficacy were 3 hours or less [2, 3], we postulated that the drug's benefit may occur later in the course of therapy. Thus, the purpose of our study was to determine whether the addition of intravenous aminophylline (during a 48-hour period) improves pulmonary function faster than frequent nebulizations of albuterol and intravenous methylprednisolone alone.

Methods

Top Methods Results Discussion Author & Article Info References

Study Design and Patient Selection

We did a randomized, placebo-controlled, double-blind study in patients admitted to the medical service of a tertiary-care university teaching hospital. The study protocol was approved by the Institutional Review Board of the University of Florida, and each patient gave written informed consent.

Eligible persons included adults 18 to 50 years old who were admitted to the hospital with an acute exacerbation of asthma by the emergency department physician. Each patient met the American Thoracic Society criteria for the diagnosis of asthma [8] and had failed to respond to three or more doses of inhaled albuterol (with or without the addition of subcutaneous terbutaline) and a loading dose of intravenous corticosteroids. Chest radiographs, arterial blood gas values, and serum theophylline concentrations measured by fluorescence polarization immunoassay (Abbott TD_x; Abbott Diagnostics, Irving, Texas) [9] were obtained from each patient. Entry into the study (designated as time 0) was identified by the first spirometric measurement done by the investigators (after treatment in the emergency department and at hospital admission). Patients were excluded if they were unable to perform spirometry correctly or consistently; were intubated; were pregnant; had a lower respiratory tract infection (lobar consolidation or new pulmonary infiltrate on chest radiograph); had a PCO₂ greater than 50 mm Hg after three consecutive nebulized albuterol treatments; had chronic cardiopulmonary disease, chronic bronchitis, or emphysema; or if the patient had a forced expiratory volume in 1 second (FEV₁) greater than 80% of the predicted value at time zero.

Interventions

All patients received routine intravenous fluids; supplemental oxygen to keep oxygen saturation greater than 92% by pulse oximetry; intravenous methylprednisolone, 125 mg, in the emergency department, followed by 60 mg every 6 hours; nebulized albuterol treatments; and either intravenous aminophylline or placebo.

The dose and frequency of administration of as-needed albuterol nebulizations were titrated to clinical response, FEV₁, and adverse effects. The dose (2.5 or 5.0 mg) was selected by the internal medicine housestaff based on their judgment of the adequacy of the clinical response, the occurrence of adverse effects on heart rate or blood pressure, or if tremor or headache was intolerable. The frequency of nebulizations was based on the last measured FEV₁. The following criteria for as-needed albuterol administration were used. If the FEV₁ was less than 30% (of the predicted value), albuterol treatments were given every half hour. If FEV₁ was 30% to 50%, treatment was given every 1 to 2 hours (the choice between 1 and 2 hours was based on the physician's assessment of the severity of symptoms and presence of adverse effects). If FEV₁ was more than 50%, albuterol was given every 4 hours. This individualization of albuterol administration is consistent with recommendations in the NHLBI Expert Panel Report [4]. Albuterol (Proventil 5% solution; Schering Corporation, Kenilworth, New Jersey) was diluted to a final volume of 4 mL with normal saline and delivered by a Whisper Jet hand-held nebulizer (Marquest, Englewood, Colorado) driven by 100% oxygen at a flow rate of 8 L/min. Under these conditions, 85% of the dose of albuterol is aerosolized [10].

Patients were selected randomly to receive an intravenous loading dose of either aminophylline or placebo, which was prepared by an unblinded clinical pharmacist. Aminophylline was added to a commercial plastic intravenous bag containing 5% dextrose in water, whereas the intravenous solution alone was used for placebo. The intravenous bags were identical in appearance and labeling. In the patients treated with aminophylline, the loading dose was based on the serum theophylline concentration measured in the emergency department and was infused for 30 minutes. One milligram per kilogram of theophylline was administered for each 2 µg/mL desired increase in serum theophylline concentration to achieve a target concentration of 15 µg/mL [11]. Thereafter, a continuous infusion of aminophylline (starting at 0.6 mg/kg hour) or placebo was begun. The infusion rate of aminophylline was adjusted by the clinical pharmacist to maintain theophylline concentrations between 10 and 20 µg/mL. To assure double-blinding, patients receiving placebo infusions had "sham" theophylline concentrations measured, followed by adjustment of the placebo infusion rate. The randomization code was not revealed until the last patient completed the study.

Measurements and End Points

Spirometry was done using a water-sealed Collins spirometer attached to an Eagle 1 microprocessor and printer (W.E. Collins, Braintree, Massachusetts). Each spirometric maneuver (the better of two efforts) was done approximately 1 hour after a nebulized albuterol treatment. The spirometer was calibrated each day with a 3-L syringe.

Forced expiratory volume in 1 second was measured at x 0 (entry point; baseline), 1, 3, 6, 12, 24, 36, and 48 hours. For each patient at each time point, FEV₁ was converted to a percentage of the predicted value based on age, height, sex, and race [12] (Appendix). Our primary outcome measure was relative improvement in FEV₁ from hour 0. The change in FEV₁ from baseline to time t was standardized relative to the difference between the baseline value and the 100% predicted value for that patient [12] (Appendix). This measurement, when plotted against the log of time, provides a good linear fit for assessing rate of improvement in FEV (₁). It measures the improvement in FEV₁ in a way that adjusts for intersubject differences in baseline severity of airway obstruction and it is also independent of body size. Some investigators have suggested that this is the best clinical method for assessing bronchodilator response [13, 14].

Adverse effects, vital signs, and an asthma symptom score were measured at the same times as FEV₁. The symptom score was the sum of the subjective rating of the patient's sensation of dyspnea plus the study physician's assessment of the severity of wheezing based on auscultation. The following was the basis for the symptom score: absent = 0, mild = 1, moderate = 2, and severe = 3. The maximum possible score at each time point was 6 points. Also recorded was the number of as-needed nebulizations of albuterol and the total dose of albuterol given during the 48-hour study period.

Whole-blood theophylline concentration was measured from a fingerstick at the bedside by an unblinded clinical pharmacist using an AccuLevel kit (Syntex Medical Diagnostics, Palo Alto, California) [15] at 0.5 hours and 8 hours after completion of the intravenous bolus of aminophylline or placebo and then approximately every 24 hours thereafter. The coefficient of variation for measurement of a 10-µgrams/mL control solution by the pharmacists participating in the study was 11% (n = 31). This method of measuring theophylline concentration was used to avoid unblinding by the hospital laboratory computer system.

The study drug infusion and intravenous methylprednisolone were discontinued at 48 hours or before if the patients became relatively asymptomatic and were ready for discharge. At that time, both aminophylline- and placebo-treated patients were converted to oral prednisone, 40 to 60 mg daily, and to oral slow-release theophylline. Patient discharge decisions were made blinded by housestaff physicians. Because patients were charged for all hospital costs, we considered it to be unethical to keep them in the study if they became asymptomatic before 48 hours.

Data Analysis

Mean differences between aminophylline- and placebo-treated patients on quantitative demographic and outcome variables were assessed by ordinary two-sample Student t-tests and Wilcoxon rank-sum tests. Because the two methods produced similar conclusions, only the results of the two-sample Student t-tests are reported, along with their associated CIs. The Fisher exact test was used to analyze categorical variables such as sex, race, smoking status, and incidence of adverse effects. For computing CIs for differences in proportions, we used the exact method of Conlon and Thomas [16]. The robust tests of O'Brien [17] and Brown and Forsythe [18] were used to compare the group's intersubject variabilities.

Our primary hypotheses were tested using a random coefficient growth curve model [19] to characterize the percentage improvement in FEV₁ across time and between groups. This technique generalizes the ordinary repeated-measures model by not requiring data at all times, which was important because four patients were discharged early and thus had missing data at 36 and 48 hours. It also provides a method to summarize and test specific components of the time trends in percentage improvement that does not rely on multiple t-tests (see Appendix). Analyses of covariance were used to determine how percentage improvement in FEV₁ was related to baseline measures, including heart rate and symptom score.

Results

Top Methods Results Discussion Author & Article Info References

Between February 1990 and February 1991, 28 adults, ages 22 to 48 years, were evaluated for participation in the study. Seven patients were excluded. One could not do spirometry adequately, 3 refused to participate, 1 had a PCO₂ value greater than 50, and two patients had FEV₁ values of 80% or more of the predicted value after nebulized albuterol treatments in the emergency department. Of the 21 patients enrolled, 10 were randomly assigned to receive aminophylline and 11 to receive placebo. The groups were similar with respect to age (mean ±SD: aminophylline, 32.8 ± 10.3 compared with placebo, 33.9 ± 6.7 years), sex (female to male ratio: 6:4 compared with 5:6), race (black to white ratio: 5:5 compared with 7:4), smoking status (yes to no ratio: 2:8 compared with 3:8), baseline arterial blood gas measurements [arterial pH: 7.43 ± 0.12 compared with 7.42 ± 0.07; arterial CO₂ tension: 37.8 ± 10.4 mm Hg compared with 37.1 ± 7.2 mm Hg]; arterial O₂ tension: 65.8 ± 6.9 mm Hg compared with 70.9 ± 14.4 mm Hg), serum theophylline concentration measured in the emergency department (1.9 ± 2.5 µg/mL compared with 3.8 ± 5.3 µg/mL), total nebulized albuterol dose received in the emergency department (20 ± 5 mg compared with 21 ± 11 mg) and, importantly, baseline FEV₁ (49% ± 19% compared with 43% ± 13% of the predicted value). Patients randomly assigned to receive aminophylline had a lower mean heart rate at baseline (107.8 ± 13.5 beats/min compared with 124.7 ± 13.0 beats/min) and slightly lower asthma symptom scores (3.0 ± 0.47 compared with 3.5 ± 0.69 of a possible maximum of 6 points).

Improvement in lung function was slow and variable in both groups during the 48-hour study period (Figure 1). At 3 hours, clinically important bronchodilation ( 20% improvement in FEV₁) was achieved in 6 of 10 (60%) patients treated with aminophylline compared with 2 of 11 [18%] patients receiving placebo. The apparent tendency for greater variability in the patients treated with aminophylline was not supported by appropriate tests of this hypothesis. The fit of the repeated-measures linear regression model was excellent Figure 2, top). Comparison of the mean intercept value for each group indicates that patients treated with aminophylline had relatively greater improvement in FEV₁ at hour 3 (29% ± 23% compared with 10% ± 10%; mean difference, 19 percentage points; CI, 3 to 35 percentage points; P = 0.023). However, the means for the slope values for each group were not statistically different (P = 0.32), indicating that the rate of improvement remained relatively constant after 3 hours (8.0% ± 4.8%/log hour in the aminophylline group compared with 5.1% ± 7.9%/log hour in the placebo group; mean difference, 2.9%; CI, –3.1% to 9.0%). A "model free" evaluation of the primary hypothesis was done by computing the CIs for the differences in percentage improvement at each hour. None of the intervals for hours 3 to 48 include a 0% difference, indicating that the associated t-tests were all significant at P < 0.05 [two-tailed test] Figure 2, bottom). Thus, a greater percentage improvement in FEV₁ was observed in the aminophylline group by hour 3, and the benefit was sustained thereafter. At 48 hours FEV₁ was 75.4% ± 19.1% of predicted in patients treated with aminophylline and 57.9% ± 14.5% of predicted in patients given placebo (mean difference, 17 percentage points; CI, 0.2 to 34.8 percentage points; P = 0.048).

View larger version (18K): [in this window] [in a new window]   Figure 1. Time course of improvement in FEV1 in each patient. Broken lines represent the percentage improvement in FEV1 over baseline in each patient treated with aminophylline (top) or placebo (bottom) expressed as a percentage of achievable improvement assuming that the maximum possible FEV1 is 100% predicted for age, height, sex, and race. Solid lines indicate the means for each group.

View larger version (31K): [in this window] [in a new window]   Figure 2. Mean improvement in FEV1.Top. Log-linear regression fit of mean and 95% CIs of percentage improvement in FEV1 for each group. The circles represent the mean values for each treatment (\#9679; = aminophylline, = placebo). By definition, all patients had 0% improvement at baseline. The linear fit for hours 3 to 48 (log-scaled) was excellent. Mean ±SD percentage improvement in FEV1 at hour 3 was 29% ± 23% in the patients treated with aminophylline and 10% ± 10% in the patients given placebo (P = 0.023, two-tailed). After 3 hours, mean rates of improvement in the two groups (slopes of the fitted lines) were not statistically different. Bottom. Mean difference between treatments and associated 95% CIs at each measured time point. Intervals excluding 0% indicate that t-tests are significant at P > 0.05, two-tailed.

Concordant results were obtained in various analyses of intercepts and slopes of percentage improvement in FEV₁ versus time plot and in the individual percentage improvement values when measures taken at baseline were used as covariates in assessing differences between patients treated with aminophylline and those given placebo. In particular, patients randomly assigned to receive aminophylline had slightly lower heart rates and slightly lower symptom scores at baseline. With no covariates, the estimated treatment difference is 18.8 percentage points, with a standard error of 7.5 percentage points (P = 0.023). Using baseline heart rate and baseline symptom score as covariates, the estimated treatment effect is 18.1 percentage points, with a standard error of 8.9 percentage points [P = 0.057]. The fact that this latter two-tailed probability value slightly exceeds 0.05 is not critical in that 1) the change from 0.023 to 0.057 is small, 2) the hypothesis was designed to be one-tailed [thus the probability values would be halved], and 3) the covariates were collectively (and individually) not significant in the model (F [2, 17] = 2.65, P = 0.10).

Aminophylline-treated patients required less as-needed nebulized albuterol throughout the study's 48 hours (Hotelling's T², P = 0.022), as measured by fewer nebulizations (10.3 ± 3.8 compared with 16.4 ± 5.3; mean difference, –6.1;CI, –10.3 to –1.8)and less total dosage (34 ± 16 mg compared with 70 ± 34 mg; mean difference, –36 mg; CI, –60.6 to –11.3 mg) (Figure 3). Asthma symptom scores, based on the patient's sensation of dyspnea and severity of wheezing at 3 and 48 hours, were not different for the two groups (P > 0.2).

View larger version (19K): [in this window] [in a new window]   Figure 3. Requirements for as-needed nebulized albuterol. Each data point represents the number of as-needed nebulizations and total dose of albuterol required during the 48-hour study period for each patient (\#9679; = aminophylline, = placebo). The triangles represent mean values for aminophylline () and placebo () recipients. Aminophylline-treated patients received less albuterol, averaging 10.3 nebulizations and a total dose of 34 mg, compared with 16.4 nebulizations and a total dose of 70 mg for patients given placebo (Hotelling T2, P = 0.022).

Deviation from the study protocol occurred in two patients. One patient in the aminophylline group received two puffs of ipratropium bromide on three occasions and one patient in the placebo group received 0.25 mg of terbutaline subcutaneously on two occasions. Exclusion of these patients from the data analysis does not change the results. Three patients in the aminophylline group and one patient in the placebo group were discharged from the hospital before 48 hours by housestaff physicians because they were symptom free. Patients treated with aminophylline were admitted to the hospital for 2.2 ± 1.1 days compared with 3.1 ± 1.3 days for the placebo group (CI for difference,-2.0 to 0.3 days; P = 0.13), but this end point was confounded in that oral theophylline was added to the regimens of the patients given placebo at 48 hours when the study was discontinued.

Whole-blood theophylline concentration 30 minutes after completion of the aminophylline loading dose was 13.2 ± 3.4 µg/mL for patients treated with aminophylline and 4.4 ± 2.0 µg/mL for patients given placebo. Thereafter, the mean theophylline concentration ranged from 12.3 to 14.5 µg/mL for patients treated with aminophylline. By 8 hours, values were less than 4.0 µg/mL in all patients given placebo.

The incidence of adverse effects was not statistically different for the two groups. Three of ten (30%) patients treated with aminophylline but only 1 of 11 (9%) patients given placebo reported nausea, a difference of 21 percentage points (P = 0.31; CI for the difference, –9 to 51 percentage points). Both groups of patients had a similarly high incidence of tremor and headache, presumably adverse effects of the albuterol, but these symptoms were not sufficiently severe to require discontinuation of the study. No clinically important adverse drug effects such as vomiting, arrhythmias, or seizures occurred in either group. Heart rate decreased at the same rate in both groups.

Discussion

Top Methods Results Discussion Author & Article Info References

This is the first prospective, controlled trial to document the benefit of aminophylline combined with aggressive use of nebulized albuterol and intravenous methylprednisolone in adults admitted to the hospital for acute exacerbation of asthma. The mean rate of improvement in FEV₁ during the first 3 hours was faster in patients treated with aminophylline than in those given placebo. Aminophylline seemed to give patients a head start early in the course of therapy [see Figure 2, top], and the difference persisted during the 48 hours of the study Figure 2, bottom). Patients treated with aminophylline required fewer as-needed nebulizations of albuterol and about one half the total dose (see Figure 3). Also, aminophylline was well tolerated, probably because theophylline concentrations were maintained at a level of 10 to 20 µg/mL.

In contrast with our findings, Self and colleagues [7] found no objective or subjective benefit from the addition of aminophylline to inhaled albuterol and oral prednisone in adults admitted to the hospital with acute asthma. They reported no statistical difference in FEV₁ (percentage of predicted value) between patients given aminophylline and those given placebo during a 32-hour observation period. However, they did not use baseline FEV₁ as a covariant in their analysis, whereas this measure was included in our calculation of percentage improvement (our primary outcome measure) and in the analysis of covariance. Their patients treated with aminophylline started at a lower baseline FEV₁ than did patients given placebo, and the improvement in FEV₁ between 0 and 8 hours appeared to be greater in the aminophylline group. This simply may be due to regression-to-the-mean effects or it might indicate true aminophylline efficacy. Further, they did not measure FEV₁ at 3 and 6 hours. Thus, their data are not inconsistent with our conclusion that aminophylline gives patients a head start early in the course of therapy that is more easily seen when the rate of improvement in FEV₁ over time is measured. Also, we titrated the dose and frequency of albuterol on the basis of FEV₁ and severity of symptoms consistent with the NHLBI guidelines [4], whereas Self and colleagues [7] administered albuterol at a fixed dose and interval (every 4 hours).

Interestingly, another group of investigators in our pediatric department, using a nearly identical protocol, found no statistical difference in FEV₁ change in children treated with intravenous aminophylline or given placebo [20]. These patients, however, received a much higher total dose of albuterol both in absolute milligrams and especially in terms of milligrams per kilogram dosage. Maximal bronchodilation was probably achieved in these patients and therefore theophylline provided no added benefit.

Theophylline enhanced ß-agonist relaxation of tracheal smooth muscle in a concentration-dependent manner in a canine model [21]. Although there are species-related differences in responses to these drugs, an additive effect might account for the greater improvement in FEV₁ or FEV₂ in patients treated with aminophylline.

The results of our study support the routine administration of aminophylline to adults who are ill enough to be admitted to the hospital for acute exacerbation of asthma, especially for those who are not adequately responding to conventional therapy with nebulized albuterol and systemic corticosteroids. However, aminophylline can cause serious adverse effects when the dosage is not carefully adjusted on the basis of serum concentration measurements [6, 22]. It should be prescribed only by physicians who understand the pharmacokinetics of this drug and are willing to adjust dosages on the basis of serum concentration measurements.

Appendix

This appendix contains information pertaining to the methods used to calculate FEV₁ values and details about statistical methods.

Measurements

FEV₁ Percentage of Predicted Value

For patient s and time t hours, FEV₁(s,t) was converted to a percentage of predicted value as follows:

% predicted FEV₁ (s,t) = (FEV₁ [s,t]/FEV₁

[s, 100% predicted]) x 100

where FEV₁(s, 100% of predicted value) is based on the age, height, sex, and race for patient s [12].

FEV₁ Percentage Improvement

Our primary outcome measure for patient s was the relative improvement in FEV₁ from baseline to time t, expressed as a percentage of achievable improvement [13], as follows:

% improvement FEV₁ (s,t) = (FEV₁ [s,t]-FEV₁(s,0)/

FEV₁(s, 100% predicted value)-FEV₁[s,0]) x 100

where FEV₁(s,t) is the measured value at time t, FEV₁(s,0) is the measured value at baseline, and FEV₁(s, 100% predicted value) is the value predicted for patient s based on age, height, sex, and race. This calculation assumes that the maximum achievable FEV₁ in patient s is 100% of the predicted value.

Data Analysis

The repeated-measures data for patient s were used to fit the parameters of the linear regression model by an ordinary least-squares method as follows:

% improvement FEV₁(s,t) =b₀(s)+b₁(s)(ln[t]-ln[3])

The model is defined so that the intercept b₀(s) estimates the percentage improvement of patient s at hour 3, and the slope b₁(s) is the estimated linear change in percentage improvement over log hours after hour 3. We planned to fit the model only for hours 3 to 48 and to base the intercept at hour 3 because we hypothesized that the treatment effects would probably not be evident before the third hour based on previous reports [2, 3]. Thus, the repeated-measures data from the 21 patients were summarized by 21 independent pairs of intercepts (at hour 3) and slopes (after hour 3), which were analyzed by ordinary methods, such as t-tests, to assess group differences in percentage improvement in FEV₁. Analyses of covariance were used to determine how percentage improvement was related to baseline measures, including heart rate and symptom score.