Mathematical error and other issues
The first thing that stood out in this trial was the small number of
participants. Only 121 patients were randomised. Considering this was a
multi-centre trial involving nine hospitals over three years, this works
out at an average of four patients per hospital per year. Any small public
hospital would have that many eligible patients with pneumonia every week.
The study aimed to “exclude inferiority of early discontinuation of
amoxicillin treatment, defined with a 10% margin in the clinical success
rate.” This wordy definition presumably translates to: see whether three
days of antibiotics is as effective as eight, with a 10% difference in
outcome being acceptable.
The primary outcome in this study was “clinical cure”. A 10% “margin”
implying that 1 in 10 people who were treated with 3 days of antibiotics
would have a worse outcome than those treated with 8. Is this really an
acceptable difference for such a common condition?
There also appears to be a mathematical error in the results. “In the
per protocol analysis the cure rates were 93% (50/54) in the three day
treatment group compared with 93% (56/60) in the eight day treatment group
(difference 0.1%, -9% to 10%).”
50/54 gives 92.6%. 56/60 gives 93.3%. The difference is therefore 0.7% not
Have the remaining statistical analysis been adequately checked?
It was also vaguely embarrassing to read an acknowledgement to
GlaxoSmithKline for their financial contribution in donating all the
drugs, but then not listing them under the funders or as a competing
interest. So it’s ethically OK to get the drugs for free as long as they
don’t give cash or offer conferences overseas? And if the study actually
favoured longer treatment durations, would GlaxoSmithKline then be
considered a competing interest if they donated the pills?
Finally, the discussion of antibiotic resistance is simplistic. The
study cites three papers that have looked at overall antibiotic
consumption on a population level and levels of resistance. The first
study (reference 6 in the original paper) from Finland found a reduction
in macrolide resistant group A strep with reduced numbers of prescriptions
in the community. The other studies (references 7 and 8) simply showed
higher rates of antibiotic resistant bacteria in European countries with a
higher total antibiotic consumption. These trials are all related to
reducing the number of prescriptions, not with reducing the duration of
antibiotic treatment, and hence do not provide additional support for
shorter dosage regimes affecting resistance.
At medical school we learn that failure to complete a course of
antibiotics, even when symptoms have resolved, can actually encourage
resistant strains of bacteria – could this happen here?
In conclusion, the small sample size, confusing terminology,
mathematical error, lack of identifying financial contributors and
simplistic discussion of antimicrobial resistance makes this study seem
much less polished than a multicentre trial with 15 authors.
Competing interests: No competing interests