I wasn't planning to blog about the study, but I changed my mind after a reader emailed me a link to a news article reporting that the antibiotic cream was 100% effective. The lead investigator even claimed, "None of the test subjects went on to develop Lyme borreliosis." As described by the news sources, seven subjects in the control group developed Lyme disease. But the abstract of the paper states clearly that the antibiotic cream (azithromycin being the antibiotic) was not any better than the control cream; the investigators were even told to stop recruiting additional patients because the results was so clear with the patients who had already completed the study:
The trial was stopped early because an improvement in the primary endpoint in the group receiving azithromycin was not reached. At 8 weeks, 11 (2%) of 505 in the azithromycin group and 11 (2%) of 490 in the placebo group had treatment failure.
So how is it possible for the lead author to claim that "none" of the subjects treated with azithromycin came down with Lyme disease? The answer lies with the very last sentence of the abstract:
A subgroup analysis in this study suggested that topical azithromycin reduces erythema migrans after bites of infected ticks.
The subgroup analysis was done post-hoc (after looking at the data). I won't dwell on why we shouldn't make definitive conclusions from any post-hoc analysis since the investigators themselves emphasized its exploratory nature in the Discussion of their paper. However, even if you set that aside, you'll find another problem with the post-hoc analysis if you dig into the numbers.
Before I tell you what the problem is, let me first describe the study in greater detail so that you understand the issues that led to the post-hoc analysis.
The subjects were adults who had been bitten by a tick within the previous 72 hours and were able to save the tick. The subjects were randomized to receive a topical azithromycin cream or placebo cream. The cream was applied over the tick bite twice a day for three straight days. The patients were followed for 8 weeks. They were monitored for erythema migrans (EM), the characteristic rash of Lyme disease. Blood was drawn for serological testing at the beginning and at the end of the 8 week study period. "Treatment failure" was defined as the appearance of EM, seroconversion, or both by the end of 8 weeks. The ticks were tested for the bacteria that cause Lyme disease (Borrelia garinii, B. afzelii, and B. burgdorferi) by PCR.
As I alluded to earlier, the independent committee monitoring the trial recommended that the investigators stop recruiting new subjects. Among the patients who already completed the study, the group receiving azithromycin did not fare any better than the placebo group, and recruiting more patients to the study was unlikely to change the conclusion. I provided the numbers above, but you can also find them in the table below ("ITT population," first row of data).
The researchers also did a pre-planned subanalysis with the per-protocol group, an idealized situation to directly test the question, "Does topical azithromycin prevent Lyme disease in those who are bitten by an infected tick?". Patients bitten by a PCR-negative tick were excluded from the subanalysis. The small number of patients who failed to follow or complete the study protocol were also excluded.
Again, azithromycin was not any better than placebo in preventing EM or seroconversion (see table, "Per-protocol population"). Treatment failure was observed in 5% (3/62) of the azithromycin group and 7% (5/72) of the placebo group (P = 0.34).
The researchers could have stopped the analysis there and write up the study, but the monitoring committee pointed out that none of the patients in the azithromycin group had erythema migrans by day 30 whereas five in the placebo group did. The committee suggested that the investigators do a post-hoc subgroup analysis using a modified definition of treatment failure as EM by 30 days. Seroconversion was removed from the modified definition.
Looking at the numbers in the table ("Reanalyzed ITT population"), we now see where the news media got their numbers. No one in the azithromycin group (0/87, 0%) had EM by day 30, but seven in the placebo group (7/87, 8%) did. The difference was statiscially significant (absolute risk reduction in those receiving azithromycin: 8.05%, 95% CI 1.18-14.91). So, it's true that azithromycin prevented Lyme disease in all who were bitten by an infected tick - but only if you ignored the two patients who came down with EM after day 30 and a third patient who seroconverted.
This is why I'm so baffled by the lead author's quote, which I will repeat: "None of the test subjects went on to develop Lyme borreiosis." I'm guessing that the two patients with delayed EM would disagree.
Schwameis M, Kündig T, Huber G, von Bidder L, Meinel L, Weisser R, Aberer E, Härter G, Weinke T, Jelinek T, Fätkenheuer G, Wollina U, Burchard GD, Aschoff R, Nischik R, Sattler G, Popp G, Lotte W, Wiechert D, Eder G, Maus O, Staubach-Renz P, Gräfe A, Geigenberger V, Naudts I, Sebastian M, Reider N, Weber R, Heckmann M, Reisinger EC, Klein G, Wantzen J, & Jilma B (2016). Topical azithromycin for the prevention of Lyme borreliosis: a randomised, placebo-controlled, phase 3 efficacy trial. The Lancet. Infectious Diseases PMID: 28007428