Possible Reporting Bias Leads to Downgrading of a Recent Adjunct Metronidazole Trial

Preus 2013

Concerns about increasing antibiotic resistance (e.g., methicillin-resistant Staphylococcus aureus, multidrug-resistant tuberculosis, antibiotic resistance of bacteria causing common infections of the urinary tract, pneumonia, or bloodstream infections), which jeopardizes effective prevention and treatment of life-threatening infections should be taken seriously when considering adjunct antibiotic therapy of periodontal diseases. After all, periodontal infections are not life-threatening diseases and can usually be controlled without adjunctive antibiotics. Apart from generalized severe cases, chronic periodontitis should not be treated in the first place with adjunct systemic antibiotics. In cases of aggressive or refractory periodontitis, microbiological diagnosis may allow targeting specific pathogens such as Aggregatibacter actinomycetemcomitans or Porphyromonas gingivalis. Responsible use of antibiotics takes into account the possible development of bacterial resistance, antibiotic toxicity and the risk of sensitizing.

Moreover, reducing the need for periodontal surgery by adjunct antibiotics may be short-sighted (note that I had written about this on several occasions here on this blog). Anatomical defects such as furcation involvement and infrabony lesions, which are the main indications for periodontal surgery, won’t resolve after subgingival scaling and adjunct antibiotic treatment. In light of the global problem of antibiotic resistance, any recommendation for repeat courses of antibiotic therapy to reduce the need for minor surgical intervention in a not life-threatening disease should be considered inappropriate.

In a recent large-scale randomized clinical trial published in the Journal of Periodontology, that very purpose had obviously been in mind of investigators. In addition, the concept was applied that specific bacterial targets in complex dentogingival biofilm of patients with mostly advanced chronic periodontitis may require different treatment and adjunctive antibiotic strategies. The authors attempted to exclude patients colonized with A. actinomycetemcomitans and a number of non-oral, possibly metronidazole-resistant, gramnegative bacteria such as enterococci, pseudomonads, Escherichia coli, Serratia, Shigella or Acinetobacter. So, the oral microbiomes of remaining 184 patients, who were randomized into four different treatment arms, were considered to be metronidazole-sensitive (but that was not tested in any case). Two groups received adjunct metronidazole (1200 mg per day for ten days) just before and after mechanical nonsurgical periodontal treatment conducted by one experienced dental hygienist while placebo tablets were provided to patients in the other two groups. Either a “traditional” approach of nonsurgical periodontal treatment (two sessions of 65 minutes, 2 weeks apart), or a “full mouth disinfection” during two sessions within 24 hours including massive chlorhexidine usage were randomly employed. Groups consisted of about 45 patients each, and the numbers were determined beforehand by expecting a 1 mm difference in mean clinical attachment gain being significant at a type I error alpha of 5% and 80% power assuming a pooled standard deviation of 1.5 mm. What may be regarded new was that, before randomization, patients undervent a 3-month pretrial period  of supragingival scaling and oral hygiene enforcement.

The results during the first 12 months of the trial (authors expect that they can finally report 5-yr data) are reported in some detail in the paper. Given the above mentioned caveats, I took notice, but in view of the numerous RCTs on adjunct antibiotic treatment of periodontal disease in the last decade, the conclusions already highlighted in the abstract, did not really surprise.

“Metronidazole had a significant, adjunctive effect in patients with a metronidazole-sensitive subgingival microbiota on the clinical parameters of CAL [clinical attachment level], PD [periodontal probing depth], and absence of pockets >=5 mm.”

As has been reported in numerous trials, adjunct antibiotic treatment tends to yield transiently improved results as regards additional probing depth reduction and clinical attachment gain, in particular when metronidazole is employed since the subgingival microbiota has been shown in an almost indefinite number of studies to consist mainly of gramnegative anaerobes. When reading the paper more carefully, it should immediately be clear that the authors’ interpretation of their data may be misleading, though. While the last part of the final conclusion may be right, it was hard to find the first part. As a matter of fact, no significant (let alone clinically relevant) differences of average clinical attachment loss (the authors’ primary outcome) or average periodontal probing depths after 1 year were shown, and cumulative distribution plots convincingly visualized this main result (see figure above). In Table 3 of the paper, gains in mean clinical attachment (the main outcome) averaged between 0.61 (full mouth disinfection, placebo tablets) and 0.81 mm (traditional scaling and root planing and metronidazole). Probing depth reductions of between 0.81 mm and 1.03 mm were observed in different groups. So, about 0.2 mm margins, which might have been “significant” if actually more than 800 patients in each group had been randomized, a hopeless undertaking given the fact that a 0.2 mm margin is not clinically relevant.

It becomes quickly clear that authors did secondary analyses to indicate that adjuncive metronidazole had an effect. They looked at “absence of periodontal pockets” of >=5 mm and they combined the two groups receiving metronidazole and those receiving placebo tablets. And they struck. Absence of “residual” pockets was found in 32/44 and 28/45 patients in metronidazole groups but only 19/45 and 18/46 patients in the placebo groups. Authors report,

“[L]ogistic regression analysis showed that the odds of success were statistically significantly greater among those treated with metronidazole than among those treated with placebo (odds ratio = 2.56; 95% confidence interval [CI] = [1.01 to 5.88]).”


Well, if the numbers in Table 3 are correct, a logistic regression with absence of pockets of 5 mm or more as dependent variable would yield an odds ratio for adjunct metronidazole of 1.901 and a 95% CI of 1.000 to 3.161, see above.

When summarizing these observations, it has to be stated that sample size calculation was based on an unrealistic expectation of 1 mm difference of mean clinical attachment gain (while 1.5 mm standard deviation of baseline attachment level might have been realistic). The (underpowered) trial did not yield a statistically significant result, i.e., mean attachment gain (and pocket depth reduction) were very similar in the four experimental groups. A secondary analysis was done to show that metronidazole had a small overall effect on another outcome (no residual pockets of 5 mm or greater) but the reported odds ratio and 95% CI seem not to be based on the numbers in respective groups. All of that should have been addressed during peer review, of course. In particular, as the study is more or less confirmative and does not provide fundamentally new information which might guide clinicians in daily decision making. The main message, i.e. adjunct systemic metronidazole treatment is not indicated in patients with chronic periodontitis, as additional benefits might be too small to compensate for adverse effects (Table 4 of the article lists the number of patients reporting side effects which was more than twice as high in the groups receiving metronidazole), risk of resistance development and higher costs, the interested reader won’t find in the paper. But the above sentence in the abstract which is repeated in the discussion’s conclusion. This is, of course, highly misleading.

Although all of us are aware of the current poblems with proper peer review, postpublication peer review is rare. Letters to the editor are frequently unwelcome and if launched, authors usually address some, but not all of the raised problems and the scientific community will mercifully forget. PubPeer comments regarding dental papers are extremely rare, those in Pubmed Commons rather uncommon. Postpublication peer review is frequently done here on my blog, but mainly for educational purposes. It was amazing to learn that Marcello Faveri and co-workers from Guarulhos University in Brazil took actually the task to point to the apparent problems with the paper by Preus et al. (2013) in a recent commentary for the Journal of Evidence-Based Dental Practice. In that commentary, the trial was graded as “Level 2: Limited-quality, patient-oriented evidence.” The main reason for this classification seems in fact to be reporting bias. It is worthwhile reading the carefully outlined arguments for their decision which includes (i) unrealistic expectations for overall attachment gain, (ii) unjustified search for some “significant differences” among secondary outcomes and (iii) unsusubstantiated claims about metronidazole-sensitive and -insensitive parts of the oral microbiome. As had to be expected, Preus et al. provided a determined rebuttal of Faveri’s commentary, see here (see their rebuttal of the rebuttal here) raising numerous distracting issues such as how to report data from clinical trials, problems with regression to the mean, the “ongoing tyranny” of p-values and ethical issues, which are misleading since Faveri et al. did not demand design alterations but rather pointed to lack of data for some of the authors’ claims.

The study by Preus et al. (2013) is beyond any dispute a major effort. Randomizing 184 patients with chronic periodontitis and treat them by employing just one dental hygienist, collect the wealth of clinical (and microbiological) data, analyze them during and after one year and expect to be able to conclude the study after five years is an admirable task. The study was certainly designed to be definitive. On the other hand, if results turn out not to fit the authors’ bias, it would have been most appropriate to change preconceived opinions and report what had been found, even after 12 months: adjunct metronidazole has no part in the routine administration of proper periodontal treatment of chronic periodontitis, that is nonsurgical therapy and minor surgical interventions.

7 June 2015 @ 11:10 am.

Last modified June 10, 2015.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s