Predicting Response to Omalizumab, an Anti-IgE Antibody, in Patients With Allergic Asthma

Predicting Response to Omalizumab, an Anti-IgE Antibody, in Patients With Allergic Asthma* Jean Bousquet, MD; Sally Wenzel, MD, FCCP; Stephen Holgate, MD; William Lumry, MD; Peter Freeman, PhD; and Howard Fox, MD

Study objective: To determine baseline characteristics predictive of response to omalizumab, an anti-IgE antibody, in patients with allergic asthma. Design: Pooled analysis of two multicenter, double-blind, randomized, placebo-controlled phase III studies with omalizumab. Patients: One thousand seventy allergic asthma patients symptomatic despite moderate-to-high doses (mean, 725 ␮g/d) of inhaled beclomethasone dipropionate (BDP). Interventions: Omalizumab (n ⴝ 542) or placebo (n ⴝ 528) were administered at a 4-weekly subcutaneous dose of at least 0.016 mg/kg/IgE (IU/mL) for 16 weeks in addition to stable BDP therapy. Measurements and results: Univariate logistic regression was performed to explore baseline variables predictive of best response. Various aspects of response (reduced symptom scores, reduced usage of rescue medication, improved lung function, improved quality of life [QoL]) were explored as well as a composite definition of response (response in at least one of these four aspects with no asthma exacerbation during 16 weeks of treatment). Time to onset of response as well as the ability to predict eventual response were also determined for the composite definition of response. A consistent pattern of predictive covariates was seen over all definitions of response (except for QoL). For the composite definition, a history of emergency asthma treatment in the past year was the factor most predictive (p ⴝ 0.015) of best response on active treatment (response rate for those with such history was 67% for omalizumab and 42% for placebo; for those without a history the response rates were 63% and 54%, respectively). Another factor predictive of best response on active treatment was high BDP dose (p ⴝ 0.037; response rate for those treated with > 800 ␮g/d was 65% for omalizumab and 40% for placebo; for those treated with < 800 ␮g/d, the response rates were 63% and 55%, respectively). A low FEV1 was also predictive (p ⴝ 0.072; response rates for those with FEV1 < 65% predicted were 60% for omalizumab and 40% for placebo; for those with FEV1 > 65% predicted, the response rates were 67% and 53%, respectively). Seventy-six percent of patients had at least one of these factors. This subgroup showed odds of being a responder (composite definition) 2.25 times higher (95% confidence interval, 1.68 to 3.01) than placebo. Some 38% of patients treated with omalizumab showed a response (composite definition) at the first evaluation time point at 4 weeks, increasing to 64% at week 16 (vs 48% for placebo; p < 0.001). Among omalizumab responders at 16 weeks, only 61% had responded at 4 weeks whereas 87% had responded at 12 weeks. Conclusions: Patients who benefit most when omalizumab is administered as add-on therapy are those receiving high doses of BDP, those with a history of frequent emergency asthma treatment, and those with poor lung function. Patients should be treated with omalizumab for a minimum duration of 12 weeks. (CHEST 2004; 125:1378 –1386) Key words: allergic asthma; anti-IgE antibody; omalizumab Abbreviations: BDP ⫽ beclomethasone dipropionate; CI ⫽ confidence interval; ICS ⫽ inhaled corticosteroids; QoL ⫽ quality of life; PEF ⫽ peak expiratory flow

recognized that IgE plays a central role in I ttheis well inflammatory response to allergen exposure in the atopic patient.1 Specific targeting of IgE with an anti-IgE antibody therefore represents a promising new approach to the treatment of allergic asthma 1378

and other IgE-mediated inflammatory disorders.2 Recently, a recombinant humanized monoclonal anti-IgE antibody, omalizumab, was developed. This agent binds to the same region of the IgE molecule that interacts with IgE receptors, ie, the C⑀3 doClinical Investigations

main,3 thereby inhibiting the allergic cascade by markedly reducing the serum concentration of free IgE.4 The validity of this approach is supported by the findings of two large placebo-controlled studies5,6 in patients with allergic asthma who were symptomatic despite moderate-to-high doses of inhaled corticosteroids (ICS). When added to existing therapy, treatment with omalizumab was effective in improving symptom control with parallel reductions in the frequency of exacerbations and requirement for ICS, changes that were clinically meaningful to patients in view of significant improvement in their asthma-related quality of life (QoL).7,8 With any medical therapy, some patients will experience greater clinical benefit than others, and knowing which patients are likely to achieve the best response aids clinical decision making. The aim of the present exploratory analysis was to determine baseline patient characteristics predictive of the best response to omalizumab therapy for allergic asthma. In addition, we determined the time to onset of response and how long patients need to be treated with omalizumab before a response could be accurately predicted.

Materials and Methods The present analysis used data from adults and adolescents (aged ⱖ 12 years) with allergic asthma, symptomatic despite moderate-to-high daily doses of ICS (beclomethasone dipropionate [BDP]), who were randomized to treatment in two phase III, double-blind, placebo-controlled, multicenter studies. The identical study design and similarities in patient populations and clinical findings justified pooling of the two studies, the protocol *From the Service de Pneumologie (Dr. Bousquet), Hoˆpital Arnaud de Villeneuve, Montpellier, France; the National Jewish Medical and Research Center (Dr. Wenzel), Denver, CO; Southampton General Hospital (Dr. Holgate), Southampton, UK; the Asthma and Allergy Research Associates (Dr. Lumry), Dallas, TX; and Novartis Horsham Research Centre (Drs. Freeman and Fox), Horsham, UK. This study was supported by Novartis Pharma AG, Basel, Switzerland, and Genentech Inc., South San Francisco, CA. According to their statement, the authors and all study investigators have not made any financial arrangement whereby the value of the compensation could be influenced by the outcomes of the study, have not received significant payments of other sorts from the sponsors (excluding the costs of conducting the study), do not have a proprietary or financial interests in the test product such as patent, trademark, or licensing agreements, and do not hold a significant equity interest in the sponsor of the study (exceeding $50,000). Drs. Freeman and Fox hold permanent positions with Novartis Pharma AG. Manuscript received March 15, 2002; revision accepted July 30, 2003. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (e-mail: [email protected]). Correspondence to: Jean Bousquet, MD, Service de Pneumologie, Hoˆpital Arnaud de Villeneuve, 371 avenue du Doyen G Giraud, Montpellier 34295, France; e-mail: [email protected] www.chestjournal.org

and main efficacy findings of which are reported in full elsewhere.5,6 Briefly, the studies enrolled patients with a diagnosis of asthma of at least 1 year in duration, serum total IgE level of 30 to 700 IU/mL, and a positive, immediate skin-prick test result to at least one common allergen (dust mite, dog or cat [both studies] or cockroach [Busse et al5 only]). Further inclusion criteria included baseline FEV1 (without bronchodilators) ⱖ 40% and ⱕ 80% of predicted, increasing by ⱖ 12% within 30 min of receiving an inhaled, short-acting ␤2-agonist. Each study was performed in accordance with the latest revisions to the Declaration of Helsinki, with patients (or guardians, where appropriate) providing written informed consent before enrollment. Ethical approval was obtained from the institutional review board of each participating study center. Both studies commenced with a 4- to 6-week run-in and baseline period, during which patients were switched to inhaled BDP and maintained on the lowest dose required to maintain asthma symptoms and lung function at levels acceptable to the patient and investigator. Thereafter, patients were randomized (using a computer-generated scheme) to receive subcutaneous omalizumab or matching placebo every 2 weeks or 4 weeks (at least 0.016 mg/kg/IgE [IU/mL] every 4 weeks) for 16 weeks in addition to stable BDP therapy. Blinding was maintained by the preparation of study medication by nonstudy-related personnel at each center. Concomitant asthma medications (other than BDP and short-acting ␤2-agonist [albuterol] as rescue) were not permitted during the study, throughout which patients recorded symptoms of asthma, morning peak expiratory flow (PEF), and usage of rescue bronchodilators in daily diary cards. Asthma symptoms were recorded twice daily; nocturnal symptom assessment included the use of rescue medication, and daytime symptom assessment incorporated limitations in physical activity. Both were assessed using a 4-point scale. Total symptom score was therefore computed as the sum of the daytime and nocturnal scores, plus morning score of 0 ⫽ no and 1 ⫽ yes for asthma symptoms on awakening, with a maximum score of 9. No imputation was used for missing diary card (symptom scores) or spirometry data. Asthma-related QoL was evaluated at baseline and on completion of 16 weeks of treatment using the 32-item Juniper Asthma Quality of Life Questionnaire.9 Each question was answered by the patient on a 7-point scale with 1 representing the greatest impairment and 7 representing no impairment during the past 2 weeks. Lower scores on this questionnaire therefore reflect increased impairment. Items were equally weighted and reported as an overall score. The last observation-carried-forward approach was used for missing asthma-related QoL data. Time to First Asthma Exacerbation Asthma exacerbations were protocol defined as those requiring a doubling of baseline BDP dose or use of systemic steroids. For the 16-week treatment period, we assessed the distribution of time to first exacerbation episode using the Kaplan-Meier (product-limit) method. For the purposes of this analysis, patients who discontinued (prior to their first exacerbation) were assumed to have done so because of an exacerbation. For other patients who were prematurely withdrawn, the number of exacerbations was imputed using the following formula: imputed No. ⫽ observed No. ⫹ (No. of days from withdrawal to planned end of phase)/14. An imputation method was applied given that withdrawn patients had different lengths of exposure to risk of exacerbation. Between-group differences for time to first exacerbation episode were analyzed using the nonparametric log-rank test. The use of this test was deemed preferable to an assumption-dependent one, such as Cox analysis, given that nearly 80% of patients were CHEST / 125 / 4 / APRIL, 2004

1379

exacerbation free, so all that was known about their time to first exacerbation was that it was ⬎ 16 weeks. Probability of Response Univariate logistic regression10 was performed to determine whether baseline covariates (Table 1) were predictive of a greater probability of response on omalizumab compared with placebo. Three terms were included in each regression equation: treatment, baseline covariate, and interaction, which indicated differential prediction of response for the two treatments. Interaction coefficients with a p value ⱕ 0.2 were considered predictive (covariates significantly predictive of response, ie, p ⬍ 0.05, were highlighted). This level of probability was chosen in order to encompass covariates that, while being only slightly predictive, could be of potential clinical interest and therefore practically useful. Various a priori-defined aspects of response were explored: (1) reduced symptoms (reduction of ⱖ 1 in mean total asthma symptom score [range 0 to 9] between baseline and study end [mean over week 12 to week 16] with no increase in mean use of rescue albuterol); (2) reduced usage of rescue medication (reduction of ⱖ 1 in mean number of puffs of rescue medication per day between baseline and study end with no increase in total symptom score); (3) improved lung function (increase in mean morning PEF of ⱖ 15% between baseline and study end); and (4) improved QoL (increase of ⱖ 1.0 in overall score on the Juniper Asthma Quality of Life Questionnaire9 between baseline and study end). A composite definition of response, defined as response in at least one of these four aspects with no asthma

exacerbation during 16 weeks of treatment, was also investigated. Between-group comparisons for equality of percent responders were evaluated using the Cochran-Mantel-Haenszel test. Three severity factors were subsequently identified from the main analysis as important determinants of response: BDP dose, history of emergency asthma treatment in the past year (a composite variable based on history in the last year of at least one of the following events: overnight hospital admission or treatment in an ICUs for asthma; emergency department or doctor’s office visit for urgent asthma treatment), and FEV1. In order to explore the interrelationships between these three factors, BDP was classified as high if ⱖ 800 ␮g/d and FEV1 as low if ⱕ 65% predicted. Each patient was then classified according to how many of these factors he or she possessed, giving four subgroups corresponding to 0, 1, 2, or 3 factors. Within each subgroup, the proportion of patients who responded with either omalizumab or placebo was calculated (p1 and p2, respectively). The odds of being a responder was subsequently determined by O1 ⫽ p1/ (1 ⫺ p1) and O2 ⫽ p2/(1 ⫺ p2), giving the odds ratio of O1/O2.11 An odds ratio of 1 indicated equality of incidence of responders on the two treatments, while a value ⬎ 1 indicated more responders with active treatment than with placebo. Statistical significance (5% level) was assumed if the lower limit of the 95% confidence interval (CI) for the odds ratio was ⬎ 1. Specific statistical validation of the logistic regression model used in the present study was not performed. Rather, we relied on the generalized likelihood ratio statistic for testing the goodness of fit of the model, with its associated p value based on the asymptotic ␹2 distribution. We believed that the sample sizes

Table 1—Baseline Characteristics* Characteristics Age, yr (range) Female/male gender, No. BDP dose, ␮g/d† Rescue albuterol use, puffs/d Serum total IgE, IU/mL Active dermatitis, No. (%) No. of positive allergens Duration of asthma, years Total symptom score (0 to 9) Nocturnal symptom score (0 to 4) Overall QoL score (1 to 7)‡ In the past year: No. of visits to a doctor’s office for urgent asthma treatment Overnight hospital admission for asthma, No. (%) No. of emergency department visits for asthma treatment Asthma treatment in an ICU, No. (%) History of any emergency asthma treatment, No. (%)§ FEV1 Absolute value, mL % predicted ⱕ 65% predicted, No. (%) Mean FVC, mL Mean FEF25–75%, mL/s Mean morning PEF, L/min

Omalizumab (n ⫽ 542)

Placebo (n ⫽ 528)

39.7 (12–76) 297/245 724 (500–1,600) 4.7 (0–14.6) 198 (20–860) 36 (6.6) 2.72 (0–5) 20.5 (1–68) 4.11 (0–8.36) 1.2 (0–3.7) 4.31 (1.68–6.52)

39.0 (12–74) 290/238 726 (200–2,000) 4.7 (0–18.1) 196 (21–814) 55 (10.5) 2.67 (0–5) 20.8 (1–63) 4.14 (0–8.64) 1.2 (0–4) 4.31 (1.15–6.59)

0.94 (0–15) 17 (3.3) 0.20 (0–12) 6 (1.1) 216 (41.5)

1.01 (0–24) 31 (6.3) 0.25 (0–20) 5 (0.9) 201 (41.4)

2,426 (650–5,120) 69 (30–112) 118 (21.8) 3,489 (1,370–7,760) 1,832 (250–6,860) 347 (114–618)

2,441 (640–5,050) 69 (22–111) 112 (21.2) 3,481 (380–7,480) 1,825 (220–4,980) 354 (145–656)

*Data are presented as mean (range) unless otherwise indicated. FEF25–75% ⫽ forced expiratory flow, midexpiratory phase. †Ex-valve. ‡Determined by the Juniper Asthma Quality of Life Questionnaire.9 §A composite variable based on history in the last year of at least one of the following events: overnight hospital admission or treatment in an ICU for asthma, or at least one emergency department or doctor’s office visit for urgent asthma treatment. 1380

Clinical Investigations

were sufficient to justify the asymptotics and relied on this statistic for validation of the logistic regression. Colinearity was not a problem. Predicting Eventual Response Substantially the same a priori responder definitions were adopted in order to determine, from a composite perspective, the time to onset of response and how response at 4, 8, and 12 weeks related to eventual response for those completing 16 weeks of treatment and with no asthma exacerbation. The QoL improvement category was omitted, in view of the fact that QoL scores were only determined after 16 weeks of treatment and not at earlier time points.

Results The pooled patient population comprised 1,070 adults and adolescents with allergic asthma who were randomized to treatment with either omalizumab (n ⫽ 542) or placebo (n ⫽ 528) in addition to stable BDP therapy. Overall, the two treatment groups were comparable in terms of baseline characteristics (Table 1). Completer rates for the 16-week, steroidstable phase were 95.2% (516 of 542 patients) for those treated with omalizumab and 90.7% (479 of 528 patients) for placebo recipients. Time to First Asthma Exacerbation Omalizumab-treated patients had significantly longer times to their first asthma exacerbation than placebo-treated patients (p ⬍ 0.001). By week 16, the probability of having had an exacerbation episode was 30% for placebo compared with 16% for omalizumab. Probability of Response Overall, response rates with omalizumab were significantly superior to those with placebo (Table 2). In terms of predictive covariates, a consistent pattern was seen over all definitions of response, except for QoL (Table 3; covariates that were significantly predictive [p ⬍ 0.05] are indicated). For the composite definition, factors predictive of a greater probability of response on active treatment included high BDP dose, history of emergency asthma treatment in the past year, and low FEV1 (Fig 1). Overall, 742 of 979 patients (76%) had at least one of these three factors. With regard to the composite definition, this subgroup of 742 patients showed odds of being a responder 2.25 times higher (95% CI, 1.68 to 3.01) with omalizumab than with placebo. The odds ratio for the composite definition of response increased with increasing number of concomitant high-severity covariates (Fig 2), a finding that was also apparent across all four individual definitions of response (Table 4). www.chestjournal.org

Table 2—Response Rates (After 16 Weeks of Treatment), According to Definition of Response* Response

Omalizumab

Placebo

p Value

Reduced symptoms† Reduced usage of rescue medication‡ Improved lung function§ Improved QoL㛳 Composite definition¶

276/517 (53) 275/517 (53)

203/481 (42) 189/481 (39)

⬍ 0.001 ⬍ 0.001

107/519 (21)

49/486 (10)

⬍ 0.001

184/447 (41) 334/521 (64)

129/414 (31) 235/488 (48)

0.002 ⬍ 0.001

*Data are presented as No./No. of patients (%) with sufficient data to enable response or nonresponse to be determined. Patients with missing data were excluded. †Reduction of ⱖ 1 in mean total asthma symptom score with no increase in mean use of rescue albuterol. ‡Reduction of ⱖ 1 in mean number of puffs of rescue medication per day with no increase in total symptom score. §Increase in mean morning PEF rate of ⱖ 15%. 㛳Increase in overall score of ⱖ 1.0 on the Juniper Asthma Quality of Life Questionnaire.9 ¶Response in at least one of the four response categories with no asthma exacerbation during 16 weeks of treatment.

Predicting Eventual Response Further analysis of the composite response rates was completed to determine time to onset of response and how response at 4, 8, and 12 weeks related to eventual response at study end. Many patients receiving omalizumab could be classified as responders at the first analysis time point at 4 weeks, and the proportion of responders continued to increase throughout the 16-week period. In terms of predicting response to omalizumab, 61% of eventual responders at 16 weeks had responded at 4 weeks, increasing to 87% at 12 weeks (Table 5). Discussion Identification of those patients most likely to achieve the greatest benefit from omalizumab therapy has obvious implications for clinical decision making on the utility of this new treatment option for allergic asthma. In the present study, based on the findings of two large placebo-controlled phase III clinical trials,5,6 a pooled exploratory analysis was therefore performed in an attempt to characterize such patients. Various aspects of response to omalizumab (as add-on therapy to standard care with ICS) were explored relative to placebo, given that asthma affects different patients in different ways. The corresponding definitions of response also took into account the large, and comparable, placebo response observed in the two original studies.5,6 This placebo effect was particularly apparent in terms of asthmarelated QoL,7,8 and can probably be explained by CHEST / 125 / 4 / APRIL, 2004

1381

Table 3—Summary of Predictive Covariates, According to Definition of Response*

Covariates Age Gender BDP dose Total IgE level Active dermatitis No. positive allergens Duration of asthma Total symptom score Overall QoL score In the past year: No. of visits to a doctor’s office for urgent asthma treatment Overnight hospital admission for asthma Number of emergency department visits for asthma treatment Treatment for asthma in an ICU History of emergency asthma treatment¶ FEV1 Absolute value % predicted ⱕ 65% predicted or ⬎ 65% predicted FVC FEF25–75% Morning PEF

Reduced Symptoms

Reduced Usage of Rescue Medication

0.025†

0.062

Improved Lung Function

Improved QoL

0.16

Older

0.037† 0.20

Higher Higher

0.066

Higher

0.045†

0.14

0.18

0.085

0.082

0.055

Composite Definition

Direction of Greatest Benefit with Omalizumab‡

0.12

Higher§ Lower㛳

0.027†

Higher

0.12

Yes

0.092

0.083

0.057

0.015†

Yes

0.086 0.13

0.011† 0.057

0.071 0.14

0.072

0.043†

0.015†

0.096 0.005† 0.074

Lower Lower ⱕ 65% Lower Lower Lower

0.019† 0.14 0.092 0.17

*Values shown are p values for the significance of the treatment-by-predictive covariate interaction. A small p value means that the graph of probability of response against the value of the baseline covariate gives lines for omalizumab and placebo groups that are very nonparallel. See Table 1 for expansion of abbreviation. †Covariates that were significantly predictive of response (p ⬍ 0.05); absence of stated value indicates p ⬎ 0.2. For definitions of response, see Table 2. ‡Indicates the trend associated with the greatest benefit from omalizumab relative to placebo, in terms of the composite definition of response. §A high total symptom score indicates a patient with severe symptoms. 㛳Lower score indicates greater impairment of QoL. ¶A composite variable based on history in the last year of at least one of the following events: overnight hospital admission or treatment in an ICU for asthma, or at least one emergency department or doctor’s office visit for urgent asthma treatment.

improved compliance with BDP treatment and/or intensive medical input. It is worth noting that steroid dosing in most of these patients was at a level above which a further dose response would be unlikely. Overall, on the basis of extensive clinical information available, it was apparent that the greatest benefit from add-on omalizumab therapy (relative to placebo as add-on to standard care with ICS) was observed for the subgroup of more difficult-to-treat patients, ie, those receiving higher doses of ICS, patients with poor lung function, and those with a history of emergency asthma treatment in the last year. Indeed, patients with any of these factors had more than twice the odds of responding with omalizumab than with placebo. In this respect, it is notable that Holgate et al12 reported that in patients with allergic asthma at high risk of serious asthmarelated morbidity and mortality, omalizumab offers 1382

the potential to halve the rate of asthma exacerbations and improve disease control. Patients were defined as being high risk if they had prior intubation or had visited an emergency department, experienced an overnight hospitalization, or had undergone treatment in an ICUs within the last year. Interestingly, baseline total IgE level was not predictive of response, which can probably be explained by the fact that the reduction in serum free IgE is so pronounced with omalizumab. Indeed, the customized omalizumab dosing strategy that was employed in these studies, a 4-weekly subcutaneous dose of at least 0.016 mg/kg/IgE (IU/mL), was specifically designed to achieve serum free IgE levels ⬍ 10 IU/mL irrespective of the baseline value. Reference to the findings of both Busse et al5 and Sole`r et al6 shows that a reduction in serum free IgE levels to below this threshold was successfully achieved in both studies. Clinical Investigations

Figure 1. Probability of response (composite definition; predicted from logistic regression model) in relation to baseline dose of BDP (top) and baseline FEV1 (bottom).

Treatment with omalizumab allowed a significant reduction in concomitant ICS dosage in the two large placebo-controlled phase III clinical trials5,6 that formed the basis of the present analysis. The steroid-reduction phase was, of course, reported fully as part of the efficacy analyses of the two studies, but this end point was not investigated as a possible response criterion in the present responder analysis because it only introduced confounding features. Until recently it has not been possible to determine the precise contribution of IgE mechanisms with increasing severity of allergic asthma. In such patients, the involvement of T cells of the T-helper type 2 subtype, which are considered to orchestrate www.chestjournal.org

the basophil, mast cell, and eosinophil responses through the release of a variety of cytokines, has been clearly established.13 What the current analysis clearly shows is that IgE continues to contribute to disease pathogenesis in atopic subjects who have persistent asthma despite treatment with ICS. IgE has been shown to be produced locally in the mucosa of subjects with allergic airways disease,14 and in vitro studies15 show that corticosteroids enhance, rather than diminish, IgE production by a direct action on B cells. Thus, while having the ability to suppress T-helper type 2 cytokine production, a downside of repeated high-dose ICS treatment could be the maintenance of an ongoing local IgE response that could be amenable to omalizumab therapy. CHEST / 125 / 4 / APRIL, 2004

1383

Figure 2. Venn diagram of odds ratios (corresponding 95% CIs) for response with omalizumab relative to placebo for the composite definition of response, according to three high-severity baseline covariates.

As highlighted above, the present study aimed to determine the population of omalizumab-treated patients with best response relative to placebo when administered as add-on therapy to current standard care with ICS. A large placebo response was apparent, in line with the original study findings of Busse et al5 and Sole`r et al.6 However, Figure 1 shows that this placebo effect occurred predominantly in the less severe patients. In contrast, the probability of response with omalizumab was relatively independent of disease severity, as shown by the near horizontal nature of the plots of probability response (Fig 1). Thus, divergence of the omalizumab and placebo plots of the probability of response was greatest for those with more severe disease. In the case of FEV1, for example, this meant a higher likelihood of response with active treatment than with placebo for those with the poorest lung function.

A potential criticism of our study, which gives rise to the exploratory nature of its findings, is that we had to assign a priori cut-off points in order to dichotomize outcomes into responders and nonresponders for purposes of predictive analysis. In the case of “reduced symptoms,” for example, response was assumed if mean total asthma symptom score decreased by ⱖ 1 point with no additional usage of rescue bronchodilators. This cut-off was chosen because it was considered a meaningful improvement relative to a mean baseline score of 4 to 5. Similarly, a reduction in mean usage of rescue medication of ⱖ 1 puffs per day, with no increase in total symptom score, was considered clinically meaningful vs mean baseline usage of 4 to 5 puffs per day. Another response category was an improvement in mean morning PEF of ⱖ 15%. This cut-off was assigned given that only modest improvements in PEF were seen in the original studies.5,6 Thus, the PEF re-

Table 4 —Odds Ratios for Response With Omalizumab vs Placebo After 16 Weeks of Treatment, According to Number of High-Severity Covariates and Definition of Response* HighSeverity Covariates, No.†

No.

OR (95% CI)

No.

OR (95% CI)

No.

OR (95% CI)

No.

OR (95% CI)

No.

OR (95% CI)

1 2 3

351 303 82

1.58 (1.04–2.42) 2.01 (1.26–3.19) 3.38 (1.33–8.60)

351 303 82

1.35 (0.89–2.06) 2.82 (1.76–4.52) 3.11 (1.25–7.74)

352 304 84

1.75 (0.94–3.24) 3.63 (1.85–7.13) 5.10 (1.34–19.41)

308 247 73

1.58 (0.99–2.52) 1.84 (1.08–3.11) 2.48 (0.87–7.06)

353 304 85

1.50 (0.97–2.31) 3.02 (1.90–4.82) 4.20 (1.69–10.45)

Reduced Symptoms

Reduced Usage of Rescue Medication

Improved Lung Function

Improved QoL

Composite Definition

*OR ⫽ odds ratio. For definitions of response, see Table 2. †BDP dose ⱖ 800 ␮g/d and/or history of emergency asthma treatment in the past year and/or FEV1 ⱕ 65% predicted. 1384

Clinical Investigations

Table 5—Prediction of Eventual Response (Composite Definition)* False-Positive†

False-Negative‡

Sensitivity§

Specificity㛳

Time Points

Omalizumab

Placebo

Omalizumab

Placebo

Omalizumab

Placebo

Omalizumab

Placebo

Week 4 Week 8 Week 12

8 (16/206) 11 (29/272) 11 (32/303)

18 (23/130) 15 (27/176) 13 (27/206)

38 (120/314) 27 (67/248) 18 (39/217)

29 (104/356) 20 (63/310) 12 (33/280)

61 (190/310) 78 (243/310) 87 (271/310)

51 (107/211) 70 (149/212) 84 (179/212)

92 (194/210) 86 (181/210) 85 (178/210)

92 (252/275) 90 (247/274) 90 (247/274)

*Data are presented as % (No./total patients). †Percentage of patients responding who did not respond at 16 weeks. ‡Percentage of patients not responding who became responders at 16 weeks. §Percentage of patients who were responders at 16 weeks and had achieved response at the respective time point. 㛳Percentage of patients who were not responders at 16 weeks who were not responders at 4 weeks.

sponse threshold was set at a level that was low enough to identify patients who did show some response, but was sufficiently high enough to be considered a meaningful improvement for those who had at least moderate impairment of their baseline lung function. Finally, we considered response in terms of asthma-related QoL improvement, identifying responders as those patients whose overall score on the Juniper Asthma Quality of Life Questionnaire increased by ⱖ 1.0 point. This threshold is higher than the minimal important change of ⱖ 0.5 points identified by Juniper and colleagues.16 However, because of the large placebo effect in both studies,7,8 the proportion of responders in both groups was believed to be too high to be informative if the latter definition of QoL improvement was used. The differentiation between groups was likely to become more apparent when we required a larger improvement in QoL. We therefore adopted a cutoff for response of an improvement in score of ⱖ 1.0 point, which is consistent with a “moderate” improvement of QoL according to Juniper and colleagues.16 Overall, however, while the present analysis was exploratory in nature and therefore hypothesis generating, the fact that the results are so consistent and hang together clinically as well as statistically encourages us to believe that they are reliable and replicable. The (composite) response to omalizumab was clearly characterized by a progressive onset, with 38% of patients responding within 4 weeks of starting treatment, increasing to 64% at week 16. Overall, a low percentage of patients responding at early time points failed to respond at 16 weeks, indicating that in most cases response to omalizumab is durable. An important finding was that only 61% of patients who were responders at 16 weeks had responded at 4 weeks, increasing to 87% at 12 weeks. Such findings indicate that it would be inadvisable to treat patients for a minimum of 4 weeks and then discontinue treatment if response was not observed, as at least one third of all eventual responders would be potenwww.chestjournal.org

tially missed using this strategy. In contrast, only 18% would be missed at 12 weeks. These findings support a minimum treatment duration of 12 weeks of add-on therapy with omalizumab before deciding whether to continue therapy. Conclusion Treatment with omalizumab significantly reduces the risk of asthma exacerbations in patients with allergic asthma. Asthma patients who benefit most from add-on treatment with omalizumab are those receiving high doses of ICS, those with a history of frequent emergency asthma treatment, and patients with poor lung function. A minimum treatment duration of 12 weeks appears necessary before deciding whether a satisfactory response will be achieved with omalizumab. Taken together, these findings will aid the clinical decision of whether a patient is likely to benefit from receiving omalizumab in addition to standard therapy for allergic asthma.

References 1 Spector SL. Allergic inflammation in upper and lower airways. Ann Allergy Asthma Immunol 1999; 83:435– 444 2 Jardieu PM, Fick RB Jr. IgE inhibition as a therapy for allergic disease. Int Arch Allergy Immunol 1999; 118:112–115 3 Presta L, Shields R, O’Connell L, et al. The binding site on human immunoglobulin E for its high affinity receptor. J Biol Chem 1994; 269:26368 –26373 4 Milgrom H, Fick RB, Su JQ, et al. Treatment of allergic asthma with monoclonal anti-IgE antibody. N Engl J Med 1999; 341:1966 –1973 5 Busse W, Corren J, Lanier BQ, et al. Omalizumab, anti-IgE recombinant humanized monoclonal antibody, for the treatment of severe allergic asthma. J Allergy Clin Immunol 2001; 108:184 –190 6 Sole`r M, Matz J, Townley R, et al. The anti-IgE antibody omalizumab reduces exacerbations and steroid requirement in allergic asthmatics. Eur Respir J 2001; 18:254 –261 7 Buhl R, Hanf G, Sole`r M, et al. The anti-IgE antibody omalizumab improves asthma-related quality of life in paCHEST / 125 / 4 / APRIL, 2004

1385

tients with allergic asthma. Eur Respir J 2002; 20:1088 –1094 8 Kishiyama JL, Lainer BQ, Corren J, et al. rhuMAb-25 (E25) improves asthma-specific quality of life in patients with allergic asthma [abstract]. Allergy Clin Immunol Int 2000; 105 (Suppl 2):115 9 Juniper EF, Guyatt GH, Epstein RS, et al. Evaluation of impairment of health-related quality of life in asthma: development of a questionnaire for use in clinical trials. Thorax 1992; 47:76 – 83 10 Agresti A. Analysis of ordinal categorical data. New York, NY: John Wiley and Sons, 1984 11 Agresti A. Categorical data analysis. New York, NY: John Wiley and Sons, 1990 12 Holgate S, Bousquet J, Wenzel S, et al. Efficacy of omalizumab, an anti-immunoglobulin E antibody, in patients at

13 14

15 16

high risk of serious asthma-related morbidity and mortality. Curr Med Res Opin 2001; 17:233–240 Cameron L, Hamid Q. Regulation of allergic airways inflammation by cytokines and glucocorticoids. Curr Allergy Asthma Rep 2001; 1:153–163 Durham SR, Gould HJ, Thienes CP, et al. Expression of epsilon germ-line gene transcripts and mRNA for the epsilon heavy chain of IgE in nasal B cells and the effects of topical corticosteroid. Eur J Immunol 1997; 27:2899 –2906 Jabara HH, Ahern DJ, Vercelli D, et al. Hydrocortisone and IL-4 induce IgE isotype switching in human B cells. J Immunol 1991; 147:1557–1560 Juniper EF, Guyatt GH, Willan A, et al. Determining a minimal important change in a disease-specific quality of life questionnaire. J Clin Epidemiol 1994; 47:81– 87

3CIENTIFIC!BSTRACT 2ELATED!WARDSFOR#(%34 2ECOGNIZINGEXCELLENCEINRESEARCH THESEAWARDSAREGRANTEDTOAUTHORSOF OUTSTANDINGORIGINALINVESTIGATIVEABSTRACTSPRESENTEDATTHEANNUAL#(%34 MEETING HELDTHISYEAR/CTOBERnIN3EATTLE!LLPRIMARYAUTHORSOF ABSTRACTSAREELIGIBLEFORABSTRACT RELATEDAWARDS !CCESSMOREDETAILSATWWWCHESTNETORG

!BSTRACTSUBMISSIONDEADLINE-AY  (WWSPJH[PVUZMVYHSSH^HYKZHYLILPUNHJJLW[LKUV^(WWS`;VKH`

(JJLZZTVYLKL[HPSZMVYTVZ[H^HYKZH[^^^JOLZ[MV\UKH[PVUVYNVYJVU[HJ[ :\L*PLaHKSVH[ZJPLaHKSV'JOLZ[UL[VYNVY  -VY[OLZJPLU[PÄJHIZ[YHJ[YLSH[LKH^HYKZVUS`HJJLZZKL[HPSZH[^^^JOLZ[UL[VYN

1386

Clinical Investigations