We thank our colleagues Michaeli, Nassar and Zentner for their comment on the risk of rheumatic fever when prescribing penicillin for a period shorter than 10 days. Indeed , in areas with poorer living conditions than in The Netherlands, a 10-day course of penicillin is needed. Most of the rheumatogenic subtypes Dr Nassar mentioned were isolated from some of our group A streptococci-positive patients. The distribution of the T/M types among our patients and among healthy controls will be published shortly.
We were surprised by the difficulties experienced by Fleetcroft, by Cantekin, and also by Del Mar in his editorial comment, in view of the methodological question how to present the resolution of symptoms when they recurred during the first week of observation. A survival analysis gave us the most realistic picture of any difference of symptom duration between the treatment groups.
Unfortunately, the results of the 3-day penicillin group (with 40% recurrences in the first week) had to be forced into the Kaplan-Meier straitjacket, by using the definition of permanent resolution of symptoms. We definitely do not agree with Cantekins interpretation of our findings. Apparently he did not understand that we chose the endpoint of permanent resolution of symptoms. We thank Dr Burke for his additional comments supporting our earlier explanation.
Little and colleagues focus on the generalisability, being the most relevant issue of any clinical trial. Indeed the 21% proportion of patients who did not meet the required number of clinical criteria was lower than the 50-60% rate we expected by interpreting the results of a previous study in The Netherlands. Probably the participating GPs underreported the sore throat patients who were not eligible 'at first sight'. Naturally, the prevalence of positive throat swabs would have been much lower in this group.
We agree with Little et al that we studied a selected population. However, this was exactly our intention. We tried to mimick the GP's daily practice of patient selection. In this way, we were not surprised to find penicillin to be more effective than in the population of Little et al, who probably included many patients with a viral infection. Interestingly, their open trial also included a subgroup of 94 ill patients with a symptom-sign complex similar to our complete patient group. Their table 3 indicates that antibiotics beneficially influenced the duration of sore throat in these patients.
Contrary to Little et al, we are not worried that the artificial use of throat-swabs and medication-tray would bias the recording of symptoms in the diary. How can bias play a role if randomised patient-groups are compared? In case the compliance was artificially improved by the use of a medication-tray and a second consultation, the outcome of our study should stimulate doctors even more to convince patients to complete the 7-day course.
Our trial was not designed to study whether penicillin would prevent quinsy or peritonsillar cellulitis. Nevertheless, the reason that the protective effect of penicillin in our patient-group appeared to be at the same level as in some 1950-studies may again be explained by the strict inclusion criteria we used, selecting probably the most ill patients, most of them having a streptococci-positive throat swab.
We agree with Little et al that the population studied should be taken into account when generalising the findings. However, because we found significant and relevant effects of penicillin, we feel free to reverse the association between result and population. To incite further discussion on the management of sore throat in adults, we therefore recommend the selection criteria we used in our study and that are also used in Dutch daily practice to other routine settings, such as UK primary care.
The fear of Fleetcroft and Little is that penicillin will be used as a routine drug in a routine setting, thus leading to medicalized patients and resistant strains. We share their fear, but believe that doctors and patients can only make a well-balanced decision when they are informed not only about the costs and adverse effects of antimicrobial therapy, but also about the benefits. 
Professor of clinical epidemiology
Ruut de Melker
Professor of general practice
1 Zwart S, Ruijs GJHM, Sachs APE, van Leeuwen WJ, Gubbels JW, de Melker RA. Beta-hemolytic streptococci isolated from acute sore throat patients: cause or coincidence? A case-control study in general practice. Scand J Infect Dis 2000 (accepted for publication).
2 Del Mar C. Sore throats and antibiotics. [editorial] BMJ 2000;320:130-1.
3 Ioannides JPA, Lau J. Can quality of clinical trials and meta-analyses be quantified? [commentary]. Lancet 1998;352:590.
4 Dagnelie CF, Touw-Otten FWMM, Kuyvenhoven MM, Rozenberg-Arska M, de Melker RA. Bacterial flora in patients presenting with sore throat in Dutch general practice. Fam Pract 1993;10:371-7.
5 Little PS, Gould C, Williamson I, Warner G, Gantley M, Kinmonth AL. An open randomised trial of prescribing strategies for sore throat. BMJ 1997;314:722-7.
6 Chamovitz R, Rammelkamp CH, Wannamaker LW, Denny FW. The effect of tonsillectomy on the incidence of streptococcal respiratory disease and its complications. Pediatrics 1960;26:355-67.
7 Bradley CP. Factors which influence the decision whether or not to prescribe: the dilemma facing general practitioners. Br J Gen Pract 1992;42:454-8.
8 Graham A, Fahey T. Sore throat: diagnostic and therapeutic dilemmas. BMJ 1999;319:173-4.
Competing interests: No competing interests