Antibiotics for sore throat

Types of studies

Randomised controlled trials (RCTs) or quasi‐RCTs.

Types of participants

Patients presenting to primary care facilities with symptoms of sore throat.

Types of interventions

Antibiotics or placebo control.

Types of outcome measures

Electronic searches

For this update we searched the Cochrane Central Register of Controlled Trials (CENTRAL) 2013, Issue 6, part of The Cochrane Library,www.thecochranelibrary.com (accessed 11 July 2013), which contains the Cochrane Acute Respiratory Infections Group's Specialised Register, MEDLINE (May 2011 to July week 1, 2013) and EMBASE (May 2011 to July 2013). See Appendix 1 for details of previous searches.

MEDLINE and CENTRAL were searched using the search strategy shown below. We combined the MEDLINE search string with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity and precision‐maximising version (2008 revision) (Lefebvre 2011). We adapted the search string for EMBASE (Appendix 2). There were no language or publication restrictions.

Searching other resources

We searched ClinicalTrials.gov and WHO ICTRP (11 July 2013) for completed and ongoing trials. We hand‐checked references of selected studies and relevant reviews to find additional studies.

Selection of studies

Two review authors (AS, CD) independently screened abstracts of potential studies and retrieved full articles for those that were trials. Two review authors (AS, CD) examined the full articles and either selected for inclusion or rejected to the excluded studies list. We resolved differences in opinion by discussion.

Data extraction and management

Two review authors (AS, CDM) independently extracted data from the included studies based on patient‐relevant outcomes: namely the complications and symptoms listed above. Data extraction involved reading from tables, graphs and, in some cases, contacting trial authors for raw data (Dagnelie 1996; Little 1997; Zwart 2000; Zwart 2003).

Assessment of risk of bias in included studies

We assessed risk of bias according to the approach indicated in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We used the following six criteria: adequate sequence generation, allocation concealment, blinding, incomplete outcome data, selective reporting and other bias.

Measures of treatment effect

All treatment effect outcomes were dichotomous data, reported as risk ratios (RR). We reported occurrence of complications during the study period for suppurative and non‐suppurative complications. We assessed the presence of symptoms (sore throat, fever, headache) when possible at day three and week one (days six to eight). We also calculated numbers needed to treat to benefit (NNTB) for the primary outcomes.

Dealing with missing data

We performed an intention‐to treat (ITT) analysis for all outcomes.

Assessment of heterogeneity

We assessed heterogeneity by using the Chi² test with the significance level set at 0.1. We determined the effect of heterogeneity by the I² statistic which indicates the proportion of total variability which can be explained by heterogeneity. We interpreted values of the I² statistic greater than 50% as indicating substantial heterogeneity, in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Data synthesis

We combined data where possible in order to perform meta‐analyses to report RR for all relevant outcomes. We used a random‐effects meta‐analytical method (Mantel‐Haenszel) in order to account for heterogeneity that was detected using the methods described above. Not all studies were able to contribute data to each of the meta‐analyses performed.

Subgroup analysis and investigation of heterogeneity

We performed a series of subgroup analyses to assess the differences in outcomes across various subgroups within the participant population:

1. treatment with penicillin (omitting other antibiotics);

2. children compared with adults;

3. positive throat swab versus negative throat swab versus untested and/or inseparable data for group A beta‐haemolytic Streptococcus (GABHS).

Sensitivity analysis

We performed sensitivity analyses to assess the degree to which results were influenced by the following criteria:

1. early (pre‐1975) versus later (post‐1975) studies;

2. blinded versus unblinded studies;

3. antipyretics administered versus no antipyretics administered.

Results of the search

A total of 61 studies were considered for the review. Of these, there were 27 controlled studies that met the inclusion criteria and were included in the review. There were no new trials included in this 2013 update. However, three new trials were considered and subsequently excluded.

Included studies

The included studies investigated a total of 12,835 cases of sore throat. The majority of studies were conducted in the 1950s, during which time the rates of serious complications (especially acute rheumatic fever) were much higher than today. Seven studies published in the last 15 years (between 1996 to 2003) were included. However, no new studies have been published since 2003.

The age of participants ranged from less than one year to older than 50 years. The participants of eight early studies were young male recruits from the United States Air Force (Brink 1951; Brumfitt 1957; Catanzaro 1954; Chamovitz 1954; Denny 1950; Denny 1953; MacDonald 1951; Wannamaker 1951). Seven of the remaining studies recruited children up to 18 years of age only (El‐Daher 1991; Krober 1985; Nelson 1984; Pichichero 1987; Siegel 1961; Taylor 1977; Zwart 2000), three recruited only adults or adolescents aged 15 years or over (Howe 1997; Petersen 1997; Zwart 2003) and nine studies recruited both adults and children (Bennike 1951; Chapple 1956; Dagnelie 1996; De Meyere 1992; Landsman 1951; Leelarasamee 2000; Little 1997; Middleton 1988; Whitfield 1981).

All studies recruited patients presenting with symptoms of sore throat. The majority of studies did not distinguish between bacterial and viral aetiology. However, seven studies included or analyzed results for group A beta haemolytic Streptococcus (GABHS) positive patients only (Catanzaro 1954; De Meyere 1992; El‐Daher 1991; Krober 1985; Middleton 1988; Nelson 1984; Pichichero 1987), one study distinguished differences in outcomes between GABHS‐positive and negative patients (Dagnelie 1996) and two studies specifically excluded patients who were GABHS‐positive (Petersen 1997; Taylor 1977).

Excluded studies

The most common reason for exclusion was lack of appropriate control group (n = 13). Other reasons for exclusion were: irrelevant or non‐patient centred outcomes (n = 6), main complaint other than acute sore throat (n = 6), inappropriate or no randomisation to treatment (n = 5), an intervention other than antibiotics was being tested (n = 2), the study tracked natural course of illness only (n = 1) or that the study reported previously published data already included (n = 1).

Allocation

In most early studies, participants were randomised to treatment and control groups by methods that could potentially introduce bias (for example, Air Force serial number, drawing a card from a deck, hospital bed number) or not randomised at all. Allocation methods were generally appropriate in the later studies.

Blinding

Eighteen of the studies were double‐blinded and three were single‐blinded.

Incomplete outcome data

Outcome data were complete for nearly all studies. For one study it was not clear how many participants maintained pain score diaries and some participants who were initially randomised were excluded due to being GABHS‐positive (Petersen 1997).

Other potential sources of bias

The use of antipyretic analgesics was not stated in nine studies, administered routinely in five studies and prohibited in four studies. The prohibition of analgesics might exaggerate any small symptomatic benefit of antibiotics over control if antipyretic analgesics are usually recommended in normal practice.

Primary outcomes

Secondary outcomes

Subgroup analysis of symptom reduction

1. Symptoms

Throat soreness and fever were reduced by about half when using antibiotics. The greatest difference was seen at day three. The number needed to treat to benefit (NNTB) to prevent one sore throat at day three was less than six; at week one it was 21. Antibiotics were more effective against symptoms at day three and one week if throat swabs were positive for Streptococcus compared to negative throat swabs.

2. Non‐suppurative complications

Antibiotics showed a trend for protecting against acute glomerulonephritis but there were too few cases for the results to reach statistical significance. Antibiotics reduced acute rheumatic fever by more than two‐thirds.

3. Suppurative complications

Antibiotics significantly reduced the incidence of acute otitis media by two‐thirds, acute sinusitis by a half and quinsy by 85% compared to those taking placebo.

Authors' conclusions  

Antibiotics confer relative benefits in the treatment of sore throat. However, the absolute benefits are modest. Protecting sore throat sufferers against suppurative and non‐suppurative complications in high‐income countries requires treating many with antibiotics for one to benefit. This NNTB may be lower in low‐income countries. Antibiotics shorten the duration of symptoms by about 16 hours overall.

Summary

1. The objectives as they are stated in the abstract include an assessment of the harms associated with the use of antibiotics in the management of sore throat, but the objectives as stated in the text of the review no longer refer to any assessment of harm. Indeed, the review does not address any adverse effects of antibiotics [which are not unimportant] and does not provide a reasonable explanation as to why this is not done other than to state in the discussion that this was not possible because of inconsistencies in the way these data were recorded. In the absence of RCT data on harmful effects the authors might have considered whether usable information could be provided by other study designs.

2. Reviews on this subject should treat adults and children separately, but this review does not attempt to do this.

3. All clinically important outcomes have not been addressed by the review and others such as resource use, re‐attendance and time off school or work are probably at least as important as those that were selected. It may have been more helpful to have collected data on all available outcomes provided that they are free from detection bias.

4. The question addressed by the review is not sufficiently well defined to allow the review to be executed systematically. Clear definitions are not given for the key elements of the question.

Most importantly, clear definitions of what is meant by primary care and sore throat are not given, leading to confusion around inclusion and exclusion decisions. Many of the control groups of the included studies do not involve a placebo but instead simply compare treatment with antibiotics to no treatment, so that some excluded studies would be eligible for inclusion, such as Catanzaro 1958 which was excluded because it compared antibiotics with sulfadiazine.

Apparent errors in inclusion and exclusion decisions have arisen probably as a result of the general lack of clarity discussed above. Specifically, the lack of a clear definition of what is meant by primary care appears to have led to the inclusion of an odd assortment of studies. For example, a couple of the included trials studied only people with sore throat who were admitted to hospital (Siegal 1961 and Bennike 1951). In addition, there appears to be an issue around the definition of a sore throat particularly in relation to positive or negative Streptococcus throat swabs. Streptococcal sore throats are a small sub‐set of the total population of sore throats and the failure of the reviewers to address this in the inclusion criteria means that the results of pragmatic trials of sore throat are mixed in with those of
 streptococcal sore throat.

There is a failure to always faithfully report the detailed results of the included studies, and there are several numerical errors in the data abstracted. For example, in Bennike 1951 the baseline numbers include patients in the "ulcerative tonsillitis" group even though most outcomes are not reported for this group.

5. The search strategy is restricted to a Medline search, a search of the Cochrane Library and citation checking. No attempt appears to have been made to search other databases. The reviewers are not explicit about the details of their searching activities nor about how they used the work of the Cochrane Acute Respiratory Infections Group.

6. References to the included and excluded studies were incomplete. Specifically they were not provided for Dagnelie 1996, Howie 1997, Little 1997 and Peterson 1997 (included) and Herx 1988, Howie 1970, Marlow 1989, McDonald 1985, Schalen 1993 and Todd 1984 (excluded).

7. Given the nature of the data presented, it is possible that a formal meta‐analysis was inappropriate. A descriptive analysis may have been more appropriate and more informative.

8. There is considerable uncertainty around the effectiveness of antibiotics on sore throat on the basis of the existing research examined by this review and this is not emphasised by the authors. Particular problems exist around the relevance of the trials to the present day with regard to the outcomes examined (rheumatic fever and glomerulonephritis), the poor quality of the majority of the included trials and the generalisability of the trials with regard to the study populations (e.g. United States air force recruits).

Jackie Young (on behalf of an interdepartmental critical appraisal workshop based in the Department of Public Health and Epidemiology, The University of Birmingham, UK) Email: j.m.young.20@bham.ac.uk

Reply

1. This is valid criticism: we need to describe the inadequacies of the information in the trials (after checking again) in the text.

2. A subgroup analysis on the basis of age is a good idea, and we will attempt this at the next major review.

3. This is a good idea, and we will attempt this at the next major review.

4. Certainly the issue of definitions is particularly difficult in this group of illnesses. One of us has written a paper on these difficulties (Del Mar C. Managing sore throat: a literature review. I. Making the diagnosis. Med J Aust 1992;156:572‐5.). There is a particular difficulty in the fact that primary care doctors use the terms 'sore throat' tonsillitis and pharyngitis in slightly different ways, including interchangeably. Moreover the notion that patients with positive swabs for Streptococcus have a different illness can be challenged. Nevertheless a subgroup analysis for this with swab‐positive and swab‐negative is a good idea which we will incorporate with our next review.

Thank for pointing numerical errors out to us, and we will check on this. Please could you detail other numerical errors for us?

5. We are explicit about our search method. At the time we undertook the search the Cochrane Acute Respiratory Infections Group had no material to assist us. This will be reviewed at the next major update.

6. Thank you for drawing our attention to this.

7. As is often the case, there is considerable variation in the population groups, treatments, outcomes measures, etc in these trials. This does not make a synthesis inappropriate, but rather allows us to examine whether these factors appear to make a difference. We also felt it important to specifically attempt to calculate the SIZE of the benefits, as this is what clinicians are interested in, and what will persuade them to modify their practice. It is then important to recognise that the size of the effect will vary in different populations: as we point out, in groups at high risk of rheumatic fever ‐ such as Australian aboriginals ‐ the prevention of RF is important; we are also interested in trying to better predict which sub‐groups will experience the most or least symptom relief, and plan to detail this in the next update.

8. We think we have discussed this in the Review. However we will reconsider what we have written in the overhaul.

Contributors

The review team.

Summary

I noticed that trials with no events in either groups are not (cannot) be part of the pooled estimates. Although I see there is a statistical/technical problem here it does not seem right. It appears to imply that no events is no evidence. I wonder whether it is defensible to add one event in both groups and add the evidence as one would normally do?

Gerben ter Riet

I certify that I have no affiliations with or involvement in any organisation or entity with a direct financial interest in the subject matter of my criticisms.

Reply

Many thanks for this. We have gone back and checked with statisticians about your point. The issue seems to be:

1. Whether empty cells are a problem. The concern is that because one cannot divide anything by zero, this might represent a problem. We think not, because in no forest plots are there totals with zero‐‐except for acute glomerulonephritis (there were no cases in the intervention arms of any trials, and only two in the control arms).

2. Whether the empty cells represent no evidence or evidence of no effect. We only recoded a zero where the study declared the outcome. Thus we assume that "no events" implies no events, rather than no reporting of events that might have occurred.

We have reported in Peto Odds ratios, the best measure for rare events.

Contributors

Chris Del Mar

Summary

Feedback: There seems to be a printing error in the abstract: the total number of cases according to the full text is 12835, but the number given in the abstract is 2835.

Martti Teikari (Feedback comment submitted 27 August 2008)

Reply

Many thanks. We will correct the typing error.

Contributors

Chris Del Mar

Summary

Comment: This work is important and useful. I have 2 concerns. First is a value judgment about the size of the treatment effect, especially concerning quinsy. Second, is the exclusion of other causes of adolescent and young adult pharyngitis ‐ group C (see Zwart 2000) strep and Fusobacterium necrophorum. Adolescents and young adults have a significant risk of suppurative complications, and most are not due to group A strep. A complete review in 2014 should acknowledge that sore throat in those age groups include other bacterial causes.

I agree with the conflict of interest statement below:
 I certify that I have no affiliations with or involvement in any organization or entity with a financial interest in the subject matter of my feedback.

Robert Centor
 Professor Internal Medicine
 University of Alabama at Birmingham

Reply

We thank Dr Centor for commenting on the review with his thoughtful points.

1. Our comment on the size of the reduction of the complication of quinsy

The comments we made in the review are these:

“Antibiotics are effective at reducing the relative complication rate of people suffering sore throat. However, the relative benefit exaggerates the absolute benefit because complication rates are low and the illness is short‐lived. Interpretation of these data is aided by estimating the absolute benefit, which we attempt below.

In these trials, conducted mostly in the 1950s, for every 100 participants treated with antibiotics rather than placebo, there was one fewer case of acute rheumatic fever, two fewer cases of acute otitis media and three fewer cases of quinsy. These figures need to be adapted to current circumstances and individuals. For example, the complication rate of acute otitis media among those with sore throats before 1975 was 3%. A NNTB of about 50 to prevent one case of acute otitis media can be estimated from the data. After 1975, this complication rate fell to 0.7% and applying the odds of reducing the complication with antibiotics from the data table yields a NNTB of nearly 200 to prevent one case of acute otitis media. Clinicians will have to exercise judgement in applying these data to their patients….”

In other words we think that it is important to keep in mind the incidence of complications (and the absolute risk reduction we can expect from antibiotics) rather than simply focus on the relative risk reduction. In clinical settings (such as low‐income countries, and in Australia for example among indigenous communities) where complications are much more common, then clinicians will interpret the finding of this review by increasing the threshold for using antibiotics.

We also, incidentally, mention under “Agreements and disagreements with other studies or reviews” that “A recent review analysing the risk‐benefit profile of antimicrobial prescribing for children concluded that antibiotics show little benefit in preventing quinsy following sore throat (Keith 2010).”

2. Exclusion of the other aetiological agents of sore throat such as Group C Streptococcus and Fusobacterium necrophorum.

It is certainly true that there are many aetiological agents other than Group A beta haemolytic Streptococcus (GABHS), including a huge range of viruses and bacteria, and even non‐infective causes. However two factors influence the review:

a) The enormous focus on acute rheumatic fever as a complication, which for decades was the over‐riding indication, and the single reason proposed by researchers and clinicians for using penicillin for sore throat. This was the motivation for an enormous search to find the best way of identifying GABHS, (and incidentally the reason why your own work on predictors of GABHS was so important).

b) The availability of randomised controlled trials that addressed these agents.

In future updates, any new RCTs that address other aetiological agents will be eligible for inclusion, as can be seen from our inclusion and exclusion criteria.

Contributors

Anneliese Spinks (Feedback reply submitted 24 January 2014)

Summary

Thank you for your informative review. A previous review on, generally, the same topic was conducted by Robertson et al. (1) which included n = 10 trials. Would you comment on why the following two citations included in Robertson et al. do not appear either as included or excluded references in your review?

‐ Brock LL, Siegel AC. Studies on the prevention of rheumatic fever: the effect of time of initiation of treatment of streptococcal infections on the immune response of the host. J Clin Invest 1953, 32:630‐632.
 ‐ Houser HB, Eckhardt GC, Hahn EO, Denny FW, Wannamaker LW, Rammelkamp CH: Effect of aureomycin treatment of streptococcal sore throat on the streptococcal carrier state, the immunologic response of the host, and the incidence of acute rheumatic fever. Pediatrics 1953, 12(6):593‐606.

Thanks,
 Marlys LeBras BSP, ACPR, PharmD

References:

1. Robertson KA, Volmink JA, Mayosi BM. Antibiotics for the primary prevention of acute rheumatic fever: a meta‐analysis. BMC Cardiovascular Disorders 2005;5:11.

I do not have any affiliation with or involvement in any organisation with a financial interest in the subject matter of my comment

Reply

We would like to thank you for alerting us to the omission of these early research studies in our review. We will seek to redress this in our coming update by reviewing the studies against our inclusion / exclusion criteria and revising the results accordingly if these studies do meet the inclusion criteria.

Contributors

Anneliese Spinks

Summary

I had a hard time understanding the assumed and corresponding risks in the Summary of Findings table ‐ the groups seem mixed up. After doing a manual calculation and also checking earlier versions I'm now convinced that you confused "placebo" and "antibiotics".

Best regards,
 Jon Pallon

Reply

Thank you for drawing to our attention the error in the column headings. We have now corrected this mistake so that the information in the table is consistent with the review findings.

Contributors

Anneliese Spinks