Results

The search strategies yielded a total of 3188 references which were manually searched. One additional clinical trial protocol was identified at the time of peer review of this paper; however, although the trial protocol for this proposed randomised study has been published,10 the results were not obtained. Thirteen clinical trials on antistaphylococcal treatment in cystic fibrosis were identified. Medline Express was most efficient (8/13, 62%), and the manual bibliography search identified two of the 13 studies (including one randomised controlled trial) not picked up by searching any of the three data bases. The Cystic Fibrosis Trust (UK) mailing list was used to inform all cystic fibrosis physicians in the UK of this study; four replies were received and no additional clinical data were identified. The principal characteristics of the clinical trials are listed in table 1.

Five studies on antistaphylococcal treatment were adequately randomised and controlled (table 1). Two examined different end points in the same population,3 ,11 and another1 used a crossover design with patients acting as their own controls over time and thus was very different in design from the other randomised controlled trials (it was therefore omitted from table 2 and is included in table 3). In the three studies potentially suitable for inclusion in a meta-analysis there was a large variation in clinical outcomes assessed, with weight and clearance of S aureus from the sputum being the only outcomes common to all three studies (table 2). As there are potential dangers of interpretation associated with the use of meta-analysis in the context of such interstudy variability, it was felt to be more appropriate to present a qualitative analysis (in tabular form) of the outcome measures assessed by randomised controlled trials (table 2). Table 3presents a qualitative analysis of the results of non-randomised clinical trials; for ease of interpretation trials assessing continuous (long term) antistaphylococcal antibiotic administration and those assessing intermittent antistaphylococcal antibiotic administration have been listed separately.

Discussion

Although quantitative synthesis of the currently available data on the clinical response to antistaphylococcal treatment in cystic fibrosis is precluded by trial heterogeneity, our systematic review supports the hypothesis that antistaphylococcal treatment in cystic fibrosis achieves increased sputum clearance of S aureus. The randomised controlled trials reported by Weaveret al 3 and Beardsmoreet al 11 compared the clinical benefit of prophylactic flucloxacillin with the intermittent use of a heterogeneous group of antibiotics (with a varying effect onS aureus). Despite this shortcoming, these two trials (which have different outcomes for the same population) effectively addressed the question of possible benefit from long term oral flucloxacillin. The two studies demonstrated increased clearance of S aureus from respiratory secretions with “prophylactic” flucloxacillin (compared with intermittent use of heterogeneous antibiotics) and correlated this increase with improved clinical (reduced cough frequency and hospital admissions) and pharmacoeconomic (reduced antibiotic courses) outcomes, but failed to show an associated improvement in lung function.3 ,11 The results of the trial by Schlesinger et al,15 using multiple antibiotics with broad spectrum antibacterial activity beyond that on S aureus for both limbs of the trial, are more difficult to interpret, while the randomised trial of Harrisonet al 17 had a short follow up period and consequently showed no clinical benefit (table 2).

Of the three randomised controlled trials listed in table 2, only that by Weaver et al 3 showed a significant difference in sputum clearance of S aureus between continuous and intermittent antistaphylococcal therapy (17/20 versus 6/18 subjects, respectively, p<0.01). Schlesinger15 addressed the same question but her numbers were small (sputum clearance of S aureus in both subjects receiving continuous treatment compared with two of seven with intermittent treatment). Although the third trial by Harrison17 again supported the hypothesis that antistaphylococcal therapy clears sputum S aureus in cystic fibrosis (table 2), the paper addressed a different question in this context, comparing clearance ofS aureus with cephalexin (7/7 subjects) and dicloxacillin (4/6 subjects) when administered continuously orally for two months; both drugs appeared to be effective. Sputum clearance of S aureus was again consistently improved with staphylococcus directed treatment in all nine non-randomised clinical trials that studied this outcome.

Beyond the above, however, there is limited evidence of clinical or pulmonary function improvement with antistaphylococcal therapy. Cough frequency, weight, frequency of respiratory exacerbations, hospital admissions, and antibiotic courses were the most common clinical outcome measures used. Colonisation of the sputum withS aureus is of particular importance in young children with cystic fibrosis (11 of the 13 trials were on children aged five years or under) where measurement of pulmonary function is difficult and complicated. This is likely to account for the infrequent use of pulmonary function testing as an outcome measure (two of 13 trials), in contrast to the evaluation of nebulised antipseudomonal antibiotic therapy where pulmonary function testing represents an important outcome measure for most randomised controlled trials.7

While all nine studies that addressed the clearance ofS aureus from sputum showed evidence of benefit, only in two studies (both randomised controlled trials with follow up of one year or more) was there a beneficial effect on weight, and one study (a randomised controlled trial with crossover design and follow up of two years) demonstrated a beneficial effect on pulmonary function. It follows that, when designing future studies in this field, eradication of sputum S aureus could act as an early and sensitive end point for evaluation, while the possibility of changes in weight, height, and pulmonary function would need to be studied in large populations of randomised subjects over a much longer time scale.

Finally, there is considerable diversity in clinical trial design, choice of outcome measures, antibiotics, and routes of administration that have been used to address the question of antistaphylococcal antibiotic efficacy in cystic fibrosis. Even if flucloxacillin, cloxacillin and dicloxacillin are considered as a single drug, as many as 19 different pharmaceutical approaches have used 11 clinical and six laboratory outcome measures in 13 clinical trials. This heterogeneity of clinical trials makes it more difficult to compare or synthesise clinical information from different studies. This systematic review has, however, identified our current state of knowledge and provides a framework for the design of future randomised controlled trials in this area. It is important to take this review forward as a multicentre randomised controlled trial to determine whether the use of “prophylactic” versus “intermittent” antistaphylococcal treatment in cystic fibrosis is associated with (a) improved lung function and/or chest radiographic scores, (b) an increase in bacterial resistance, or (c) earlier acquisition ofPseudomonas aeruginosa.