Randomised controlled trials can provide strong evidence of the efficacy (or lack) of new drug treatments as well as the occurrence of common side effects.1 However, clinical safety issues can arise after new medications receive regulatory approval, particularly in relation to rare serious adverse events.2 Investigation of rare adverse events is often fraught with difficulty, leading to uncertainty, particularly when there is conflicting evidence from research utilising different methodologies. One recent example is the observed association between leukotriene receptor antagonist (LTRA) therapy, used in the treatment of asthma, and Churg–Strauss syndrome (CSS), a vasculitis of uncertain aetiology (also known as allergic granulomatous angiitis). CSS is certainly both rare and serious, with a background incidence of 3 per million per year in the general population, and a 1 year mortality rate of 7%.3^–5

Soon after the introduction of LTRAs (zafirlukast, pranlukast and montelukast), numerous case reports and case series were published of patients who developed CSS after starting this therapy.6^–13 The temporal relationship between the introduction of LTRA therapy and the development of CSS suggested a possible causal relationship. An underlying mechanism was proposed whereby LTRA therapy may lead to an imbalance in leukotriene receptor stimulation, resulting in unopposed activity of LTB₄, a potent chemoattractant for eosinophils as well as neutrophils, which could potentially lead to eosinophilic tissue infiltration and the initiation of systemic vasculitis.14 15

A feature of a number of the case reports was that the introduction of the LTRA allowed significant oral steroid reduction, suggesting that this therapy may have unmasked previously existing CSS that had been suppressed by the steroids prescribed for asthma. Another observation was that some cases had severe or unstable asthma at the time LTRA therapy was introduced, which arguably may have represented the earliest phase of CSS, characterised by asthma and rhinosinusitis. It has been proposed that in these cases, LTRA therapy may have been prescribed in response to the initial phase of CSS, which then progressed to eosinophilia with pulmonary infiltration and subsequently systemic vasculitis during LTRA therapy.16 It has been difficult to assess the validity of these hypotheses because of the paucity of data that has been available, limited primarily to either case reports or case series in which the method of case selection was not specified, giving potential for selection and publication bias.

These hypotheses were not supported by the observation that there were very few published reports of CSS following the introduction of other asthma treatments, including inhaled corticosteroids (ICS), theophylline, cromolyns and long acting β agonists (LABAs), all of which have shown some steroid sparing effects.13 Similarly, of 63 cases of CSS reported to the UK Committee on Safety of Medicines (CSM) through the Yellow Card Scheme since 1963, 59 were documented in 1998 and 1999; of these, 90% were associated with drugs used to treat asthma, mainly LTRAs.17 Likewise, in the US FDA Adverse Event Report System database, reporting of CSS was strongly associated with LTRA use and not other commonly used asthma therapies, including ICS and LABA therapy.18 ICS are more likely to have oral steroid sparing effects,19 20 and LABAs greater ICS sparing effects21 22 than LTRA therapy.23 Thus if CSS does indeed appear in response to steroid withdrawal, these commonly prescribed therapies are more likely than LTRAs to result in unmasking of CSS. Furthermore, the recommended use of ICS and LABA therapy in severe asthma24 might be expected to result in their greater coincident use in patients with pre-existing CSS manifest by severe unstable asthma.

To balance this interpretation there is recent evidence that the predominant association of CSS with LTRA and not with other commonly prescribed therapies may be due in part to selection bias. In a recent US based case control study, only 6/47 (13%) patients with CSS were exposed to LTRA therapy,25 markedly lower than the proportion of CSS cases on LTRA reported to the FDA or CSM, or published in the medical literature. This study reported a strong association between LTRA use and CSS, but in multivariate analyses controlling for asthma drug use, no significant association was observed. It was acknowledged that because of the low power of the study, it was not possible to rule out an up to fourfold increased risk of CSS with LTRA therapy. There are also a number of recent case series reporting the occurrence of CSS unrelated to LTRA therapy, including cases in which ICS therapy allowed oral corticosteroid withdrawal.26 27 However, others have reported an increase in the incidence of CSS driven primarily by cases using LTRA therapy.28

As a result, there is conflicting evidence on whether the association between LTRA and CSS is primarily a result of confounding by indication, or represents a genuine causal association. Two papers, one published in this issue of Thorax,29 and the other published recently,30 utilising different methodologies, have addressed this issue and come to different conclusions (see page 883). Nathani et al present a systematic review of cases and cases series that report CSS in association with LTRA use.29 There were 62 patients from 40 publications which were identified from Medline included in the review. The main finding was that the majority of patients exhibited a clear temporal relationship between initiation of LTRA therapy and development of CSS, with no evidence of pre-existing disease or steroid withdrawal. Indeed, seven (11%) cases were entirely steroid naïve, not on ICS or oral corticosteroids when LTRA therapy was initiated. Cases were identified in which there was a relapse of CSS following reintroduction of LTRA therapy, and other cases in which there was remission of signs and symptoms of CSS on withdrawal of LTRA therapy, which could not be attributed to systemic steroid or immunosuppressive therapy. The authors concluded that the available evidence suggests an association between LTRA and CSS that may be causal.

In the other publication, Hauser et al used an extension of case control methodology to investigate exposure to the LTRA montelukast (and other asthma medications) and the development of CSS.30 The case control study identified 78 patients from two databases, one a French cohort who were participating in a clinical trial of therapy for CSS, and the other a German cohort which was part of a prospective registry for a tertiary referral centre for vasculitis management. Both databases were accumulated over a 10 year period and the study focused on montelukast as it was the only LTRA approved in France and Germany. The main findings were that the use of montelukast in the 3 months before the development of CSS was more likely than use in the period more than 3 months before its development, and that positive estimates were also obtained for other long term asthma control medications with effects that were independent of montelukast use. The authors concluded that because the association with CSS onset was not specific to montelukast but also occurred with other asthma medications, it was more likely to be caused by a general escalation of treatment in response to worsening asthma. They also noted that there was increasing use of montelukast during the study period, adding weight to a confounding rather than causative role of montelukast in CSS development.

Both papers highlight the fact that CSS is rare, and that as a result, statistical testing is likely to lack power to detect important associations (highlighted by the wide confidence intervals in the case control study). Both studies are also likely to be influenced by selection bias, which may limit the generalisability of the findings. The evidence of causation provided by case reports and case series is usually prescribed lesser weight than that of case control studies. However, the systematic review by Nathani and colleagues29 does provide important information that CSS may occur in association with LTRA therapy in the absence of pre-existing disease, may occur on rechallenge with LTRA therapy and may remit on withdrawal of LTRA therapy without the requirement to start or increase the dose of systemic steroid or immunosuppressive therapy.

The evidence against causation provided by the case control study of Hauser and colleagues30 is open to debate. The authors did in fact identify a strong association between CSS and LTRA therapy and were unable to convincingly demonstrate that it was confounding by indication that led to this association. In the multivariate analyses, in which all four asthma medication classes were included as covariates, the risk of developing CSS within 3 months was 6.7 (95% confidence interval (CI) 1.3 to 34.1) for montelukast, 4.2 (95% CI 1.2 to 14.6) for oral steroids, 2.9 (95% CI 0.6 to 13.3) for LABAs and 1.0 (95% CI 0.2 to 4.8) for ICS. Thus there was a significant risk associated with LTRA, but not with ICS and LABA therapy, although it must be noted that all the confidence intervals were wide, resulting in considerable uncertainty in interpretation. The increase in risk with oral steroids is likely to relate to their use in severe exacerbations of asthma, resulting from the progression of CSS. Notably, Hauser and colleagues30 have also provided substantive evidence (consistent with the previous case control study25) that there is likely to have been major selection bias in the reporting of the association between asthma therapy and CSS, in which the vast majority of reported cases relate to LTRA therapy, whereas such cases represented only a quarter of their database.

The study of rare serious adverse events associated with medication use is inherently difficult and the investigation of the association between LTRA therapy and CSS is no exception. However, despite the apparent conflicting nature of the data, it can be brought together in a consistent manner to conclude that there are a number of different clinical circumstances in which CSS may occur in association with LTRA therapy. These include its use for worsening asthma which may or may not be coincidentally a symptom of an early phase of a progressive pre-existing CSS; its use resulting in clinical improvement leading to a reduction in steroid use potentially allowing unmasking of underlying CSS; and its use in patients with no evidence of pre-existing CSS or change in steroid use. In addition, its use may also lead to a recurrence of CSS and its withdrawal may be associated with a remission independent of systemic steroid or immunosuppressive therapy. There is now sufficient evidence to suggest that all of these distinct clinical scenarios may occur, with the priorities now to determine the proportion of cases in which the association is causative, and the pathophysiological processes by which this occurs.31 Doctors need to be mindful of the risk of CSS associated with LTRA therapy, particularly when prescribed to patients with moderate to severe asthma or in the setting of steroid withdrawal, and be aware of the clinical manifestations of CSS and the requirement to withdraw LTRA therapy should CSS develop.