Twelve studies were included in the meta-analysis, of which 7 were developmental trials conducted by Roussel Uclaf and 5 were published postmarketing studies. There were 384 participants identified as having an infection caused by Haemophilus influenzae; of these, 331 received roxithromycin and 53 received a comparator antibiotic.
Overall clinical response.
On a per protocol basis, the overall satisfactory clinical response to roxithromycin was 87% (128 ot of 147 cases; 95% CI: 82, 92) where Haemophilus influenzae was the only pathogen identified. On an intention to treat basis, the overall satisfactory clinical response for roxithromycin was 78% (197 out of 253 cases; 95% CI: 73, 83) where Haemophilus influenzae was the only pathogen isolated.
Pure versus mixed infections.
In 24% of roxithromycin-treated cases, Haemophilus influenzae was not the only pathogen identified at baseline. A satisfactory clinical response level was found which was slightly less in mixed infections, except for acute bronchitis.
Hospital versus ambulatory patients.
Thirty-eight out of 331 cases were hospitalised, and these constituted 18% (14 out of 78) of the failures. The overall response rate for pure infections was 65% in these studies (22 out of 34 cases; 95% CI: 52, 82). In the 7 studies which enrolled only community-treated patients, the overall response rates were 79% (88 out of 112 cases; 95% CI: 71, 86) for pure infections and 76% (121 out of 160 cases; 95% CI: 69, 83) for pure or mixed infections.
Influence of roxithromycin regimen.
The overall response rate in all studies was 76% (163 out of 214 cases; 95% CI: 71, 82) for the 150 mg twice daily regimen and 87% (34 out of 39 cases) for the 300 mg once daily regimen; a non significant difference was found.
Clinical response of comparator antibiotics.
Overall response rate was 70% (33 out of 47 cases; 95% CI: 57, 83). The response rate was 72% (28 out of 39 cases) on a per protocol basis.
Bacteriological response:
The overall bacteriological response rate for the roxithromycin-treated cases with pure isolates of Haemophilus influenzae was 83% (120 out of 145 cases). The response rate for comparators ranged from 29% (cefaclor) to 100% (amoxicillin combined with clavulanic acid); however, the numbers that could be analysed for each antibiotic were small.
Influence of study quality on clinical response.
Of the 12 studies included in the meta-analysis, 7 were rated as 'very good', 3 as 'good' and 2 as 'borderline'. In studies of very good quality, a 75% satisfactory clinical response (intention to treat) was observed, compared with 86 and 83% for good and borderline studies, respectively. The difference between very good and borderline studies was non significant.
Relationship of in vitro susceptibility to outcome: 30% of Haemophilus influenzae isolated were fully susceptible, 29% were intermediate and 41% resistant.
In addition to the above, results for different subgroups analyses are presented.