Amoxicillin: A Review of Two Articles
The purpose of this paper is two examine two studies on the topic of the efficacy of amoxicillin or amoxicillin clavulanate as treatments for otitis media. The paper will describe the two studies, point out the major differences, and concludes with speculation as to why one study recommends the broad-spectrum penicillin antibiotic whereas the other recommends a wait and see approach.
Design of the Studies
Both the Dutch study, Damoiseaux et al. (2000), and the Finnish study, Tahtinen et al. (2011), used a randomized, double blind, placebo-controlled design to test the efficacy of amoxicillin or amoxicillin clavulanate. The inclusionary criteria in the Damoiseaux et al. (2000, p. 350) were children ranging in age from 6 months to 24 months whereas the Tähtinen et al. (2011, p. 116) included children who were slightly older at 6 months to 35 months old. Otherwise the exclusionary and inclusionary criteria were similar.
The intervention drug and data collection intervals differed. In the Damoiseaux, et al. study, the intervention was 40 mg of amoxicillin per kilogram of weight in three doses per day for 10 days. The control group received a placebo suspension. Both groups received one drop of oxymetazoline 0.025% in each nostril three times a day for seven days. If pain control was necessary, both groups were permitted paracetamol (2000, p. 351). The intervention group in the Tähtinen et al. study received 40 mg of amoxicillin per kilogram of body weight per day and 5.7 mg of clavulanate per kilogram of weight per day twice daily. The control group received a placebo that was similar in taste and appearance to the intervention treatment for 7 days. Both groups were encouraged to use pain and fever control medication, analgesic ear drops, and decongestant sprays or nose drops as necessary (2000, p. 117). Data was gathered on 4 occasions in the Damoiseaux, et al. study: baseline, day 4, day 11, and a 6 week follow up home visit by one of the study authors (2000, p. 351). In the Tähtinen et al. study, data was gathered during office visits of study physicians at baseline, day 3 and day 8. In both of the studies, parents were requested to maintain daily diaries of medication administrated, symptoms, and any other related events, for the duration of the study (2011, p. 117-118).
The data collection methods of the two studies differed. In the Damoiseaux, et al. study, data was collected by 53 physicians who for the purposes of the study had been trained in ear drum classification and in the distinction between otitis media with effusion and otitis media (2000, p. 350). No information on interrater reliability was provided. In the Tähtinen et al. study, five physicians collected the data and over 90% of the data was collected by three physicians. Interrater reliability was good (kappa values = 0.80 to 0.92). In addition, images and videos of 150 patients taken at the time of diagnosis were rated by an ENT specialist not associated with the study and a 95% agreement with the diagnoses was achieved (2011, Supplementary appendix, p. 5).
Although training was provided, the fact that a large number of physicians in the Damoiseaux, et al. study were involved in the data collection on days 4 and 11 at day with no interrater analysis provided was a limitation in the study that was not noted by the authors. In the Tähtinen et al. study, the patients were either referred by primary care physicians or self-selecting, which could have introduced a bias in the study (2011, Supplementary Appendix, p. 2).
Outcome Variables of the Studies
The primary outcome variables for Damoiseaux, et al. (2000, p. 351) were the continuation of symptoms, crying and irritability, and a prescription for symptoms prior to day 4. Secondary outcome variables were persistent symptoms at day 11, including no improvement noted in the appearance of the tympanum, continued crying, presence of fluid in the middle ear as determined by otoscopy and tympanometry. Information taken from the parents’ diary included the use of analgesics and adverse events. The length of time was defined as the first day that the symptoms or signs were not present. For the Tähtinen et al. study, the primary outcome was the length of time to treatment failure, which was determined by the day that the physician first noticed the following events: no improvement in symptoms by day 3, no improvement as defined by otoscopy signs on day 8, and perforation of the ear drum or severe infection at any time (2011, p. 118).
Statistical Analysis of the Studies
There were no major differences in the choice of analytic methods or statistical tests. The data from the both studies was analyzed on an intention to treat basis and used KaplanMeier curves and log rank test to compare the control and intervention groups for risk (95% CI) of continued symptoms at the set intervals (Damoiseaux, et al. 2000, p. 351; Tähtinen et al. 2011, p. 118). Damoiseaux, et al. (2000, p. 351) calculated the between-group difference for amount of analgesic taken with a Mann Whitney U test. For categorical variables, Tähtinen et al. used chi-square tests and students t-tests to determine differences in means (2011, p. 118).
Damoiseaux, et al. applied a best and worst case scenario when participants were lost to follow up (2000, p. 351, 352). The best case scenario test with the intervention group as “cured” and control group as “not cured” did not show a significant difference. The finding that all the participants lost to follow up were in the control group might have made a difference, but was not included in the discussion.
Results and Conclusions of the Studies
The Damoiseaux, et al. study showed statistically significant between-group differences in symptoms at day 4, in the median number of days with fever, and mean amounts of analgesics consumed. At day 11, there was no statistically significant between-group differences in clinical treatment failure (2000, p. 353). The results were more robust in the Tähtinen et al. (2011) as the rate of treatment failure was statistically significantly higher in the placebo group (2011, p. 122). Based on the results of the day 11, Damoiseaux, et al. (2000) were justified in stating that the differences between the two groups were not clinically important enough to warrant prescribing at the first visit and recommended a “watchful waiting” approach (2000, p. 353). The Tähtinen et al. results were more clear and the authors were justified in their conclusion that amoxicillin clavulanate was beneficial (2011, p. 125).
Given the results of day 4, Damoiseaux, et al. (2000) could have made a cautious recommendation of amoxicillin, but they did not. In the introduction, Damoiseaux, et al. state that antibiotics seem ineffectual for acute otitis media and comment on the overuse and misuse of antibiotics in the face of multi-drug resistant microbes. Also, the authors state that the Netherlands remains one of the few countries in which a minority of episodes of acute otitis media are treated with antibiotics (2000, p. 350).
In summary, this paper compared two studies that examined a similar drug in the treatment of otitis media. Some potential problems in data collection and sampling were pointed out. For Tähtinen et al., (2011) the results were straight forward. As for Damoiseaux, et al. (2000), the recommendations could have reflected the social context of the study.
References
Damoiseaux, R, van Balen, F, Hoes, A 2000, ‘Primary care based randomised, double blind trial amoxicillin versus placebo for acute otitis media in children aged under two years’, British Medical Journal, vol. 320, pp. 350-354.
Tähtinen, P, Laine, M, Huovinen, P 2011, ‘A Placebo-Controlled Trial of Antimicrobial Treatment for Acute Otitis Media’, New England Journal of Medicine. Vol. 364, pp. 116-126.
Tähtinen, P, Laine, M, Huovinen, P 2011, ‘A Placebo-Controlled Trial of Antimicrobial Treatment for Acute Otitis Media, Supplementary Appendix’, New England Journal of Medicine. Vol. 364, pp. 1-18.
