Some think of medicine, or at least antibiotic treatment duration, as a religion. In the Bible “seven” stands for completeness (I got that from Wikipedia), and therefore infections should be treated with antibiotics for seven, fourteen, or twenty-one days.

The impressive BALANCE study published in this week’s NEJM offers a randomized comparison of 7 versus 14 days of antibiotic treatment of 3,608 patients with bloodstream infections. Ninety days after randomization, “261 patients (14.5%) receiving antibiotics for 7 days had died and 286 patients (16.1%) receiving antibiotics for 14 days had died (difference, −1.6 percentage points [95.7% confidence interval {CI}, −4.0 to 0.8]), which showed the noninferiority of the shorter treatment duration.”

A remarkable study for which patient enrolment took 9 years in 74 hospitals in 7 (!) countries, mainly Canada and Australia.

Should textbooks and guidelines be rewritten immediately?

Let’s put the study against the three holy criteria of evaluation: validity, precision, and external validation. No questions about validity and precision: successful randomization, straightforward statistical analysis, and an effect estimate with a very narrow confidence interval overwhelmingly supporting non-inferiority.

Then comes the question “what patients were in the study”? Almost 60% had BSI caused by either E. coli or Klebsiella species, 75% of infections were community-acquired and 55% was admitted to ICUs. Yet apparently less than half of those admitted in ICU required mechanical ventilation (22% of all patients). Moreover, during the first 28 days the median number of ICU-free days was – in the total population – 25 days (IQR 20-28) and the median number of ventilator-free days was 28 (IQR 26-28). The median of vasopressor-free days during the first 14 days was 14 (IQR 12-14), all with no difference between groups. Fair to say that this was a mostly clinically stable patient population.

Importantly, the study tested two strategies rather than fixed durations. As a result, the median duration of antibiotic use was 8 days in the 7-day group with an IQR from 7 to 11 days and with 23.9% of the patients treated longer than the pursued 7 +2 days. So, 7 days is not exactly 7 days for all patients.

Antibiotic resistance appeared to play a minor role – at most – in this study: on the day of randomization, 2.9% of all patients received inappropriate initial treatment.

Antibiotic exposure, though, is interesting. Exposure data are provided at the time of and after randomization, as the percentage of patients exposed. For both the top six of all antibiotics used was similar, with no relevant differences between both study arms. On the day of randomization, proportions were 31% ceftriaxone, 21% piperacillin-tazobactam, 13% meropenem, 12% ciprofloxacin and 8% vancomycin. Yet within 28 days after randomization, proportions of patients exposed to these antibiotics were: 60% ceftriaxone, 52% piperacillin-tazobactam, 24% meropenem, 26% ciprofloxacin and 29% vancomycin. I presume most of these exposures were during the initial 7 and 14 days of treatment, and thus many received at least two antibiotics. For a Dutch physician that seems a lot, and aminoglycosides were hardly prescribed.

So, the results stem from a population of mostly clinically stable patients receiving appropriate initial antibiotic treatment and then treated with several antibiotics.

As I mentioned, this study provides strong support for reducing the duration of antibiotic exposure in patients with BSI: shorter was safe for those who participated in the study. Medicine is not a religion, I dare to say, and – ultimately – treatment duration should be based on biology and scientific evidence. It is highly unlikely that all patients with BSI will have the same optimal duration of antibiotic treatment for every bacterial cause. Next station: biomarker-guided discontinuation of antibiotic treatment compared to a strategy of seven days.