Using external data to assess the external validity of a randomised controlled trial

Background Few studies have addressed external validity of randomized controlled trials in infectious diseases. We aimed to assess the external validity of an investigator-initiated trial on treatment for uncomplicated urinary tract infection. Methods In the original study, women (n = 513) with urinary tract infection were randomized to nitrofurantoin or fosfomycin treatment in three countries between 2013 and 2017. In the present study we compared women who were screened for enrolment but excluded to women who participated in the trial, both groups in Israel. The primary outcome was the rate of emergency department index visits resulting in hospitalization within 28 days. Results We compared 127 included to 110 excluded patients. The most common reasons for exclusion were logistic difficulties in recruitment and antibiotic use in the preceding month. Included patients tended to be older [39 (IQR 29-59) vs. 35.5 (IQR 24-56.25 years)], more likely to have history of recurrent infection and had more urinary symptoms. Among excluded patients, 13.6% (15/110) had initial visits resulting in hospitalization compared to 3.1% (4/127) of included participants (p = .003). The rate of emergency department visits within 28 days was similar in both groups. Clinical and microbiological failures were significantly more common in included patients [26% (33/127) vs. 1.8% (2/110), p < .001; 7.9% (10/127) vs. 0% (0/110), p = .003; respectively]. Conclusions While differences were observed between included and excluded patients, the excluded group did not represent a more 'complicated' population. The present study shows the importance of collecting data on patients excluded from randomized controlled trials.

Automated summary

This study aimed to assess the external validity of a randomized controlled trial on treatment for uncomplicated urinary tract infection. The results showed that excluded patients were more likely to require hospitalization after initial visits and had higher rates of clinical and microbiological failures compared to included participants.