Can 88% of patients with acute lower respiratory infection all be special?

BACKGROUND

Acute lower respiratory tract infection (LRTI), also referred to as ‘acute bronchitis’ and ‘acute chest infection’, remains one of the most common presentations to primary care internationally.¹ Usually defined as an acute cough with at least one of the following, sputum, chest pain, shortness of breath, and/or wheeze,² between 52% and 100% (median 88%) of patients are currently prescribed an antibiotic.³ Using conservative national morbidity survey estimates,¹ LRTI costs the UK NHS an estimated £190 million annually.⁴

Despite good evidence that antibiotics do not reduce the duration or severity of LRTI,⁵ they continue to be widely prescribed³ promoting bacterial resistance to antibiotics.⁶ Between 2002 and 2010, the total number of antibiotics dispensed per 1000 population in England increased by an alarming 25%,⁷ a statistic that cannot be explained by increasing primary care presentations: consultations for upper and lower RTIs fell by 19% in the same period.⁸

So why is our antibiotic prescribing increasing? There is evidence that a key concern is avoiding under-treatment.⁹ None of us wants to be seen to have withheld treatment from a patient who subsequently deteriorates, especially if they are hospitalised. Although rare, it damages doctor–patient relationships, and leads to complaints and medico-legal consequences. However, the rising use of antibiotics and antimicrobial resistance is also of concern, and is now at the top of the Chief Medical Officer (CMO) for England, the Department of Health and the National Institute for Health Research (NIHR) agendas. In March 2013, the CMO highlighted the rise of antimicrobial resistance as a threat to healthcare delivery,¹⁰ in August 2013 the NIHR launched a themed antimicrobial resistance call for research and in September 2013, the Department of Health published the UK Five Year Antimicrobial Resistance Strategy 2013 to 2018.¹¹

NEW RESEARCH EVIDENCE

This month’s BJGP includes two articles providing evidence to help improve antibiotic treatment decision making.^12,13 They are both from the humbly named ‘GRACE’ consortium (www.grace-lrti.org), an internationally respected group of primary care researchers who worked with hundreds of GPs and nurses from 12 European nations to conduct high quality research into LRTI in over 3000 adults. A strength of both studies is that LRTI was transparently defined to maximise generalisability: adults (≥18 years) were eligible if presenting to primary care for the first time with acute (≤28 days) cough as the main symptom, and non-infective cause (such as pulmonary embolism or heart failure) was judged very unlikely. Exclusions included suspected community-acquired pneumonia, penicillin allergy, or immunodeficiency. Patients with asthma and chronic obstructive pulmonary disease (COPD) were eligible, and constituted 15% of the final sample.

The article by Moore et al¹² is a secondary analysis of a previously published randomised controlled trial (RCT).¹⁴ The largest of its kind, 2061 patients were randomised to amoxicillin 1 g or placebo three times daily and asked to complete a validated diary¹⁵ to measure duration and severity of the most common LRTI symptoms. Symptom severity was measured using a scale of zero (‘no problem’) to six (‘as bad as it could be’) that included the interim score, three (‘moderately bad’). Despite the higher than standard UK treatment dose, the trial demonstrated no difference in overall duration of all symptoms rated ‘moderately bad’ or worse (≥3) or their severity at days 2–4. Although new or worsening symptoms occurred slightly less frequently in the amoxicillin group (16% versus 19%, number needed to treat = 30), nausea, diarrhoea or rash occurred more frequently (number needed to harm = 23) and there was one anaphylaxis in the amoxicillin group. Three (0.15%) patients needed hospitalisation for cardiovascular or respiratory problems, two in the placebo and one in the amoxicillin group. Figure 1 in the Moore article shows the strength of evidence for the lack of effect on duration of ‘moderately bad’ or worse symptoms.¹²

But, we hear you say, my patient is special they need antibiotics because they are a smoker/have green phlegm/a more severe illness/chronic lung disease/are going on a family holiday tomorrow/have a vitally important business meeting in Washington DC next week [delete as appropriate]. Moore et al¹² address ‘being-special’ by investigating patient subgroups who might differ from the full group, for whom antibiotics may offer extra benefits (or harms). Also referred to as ‘looking for interactions’, this article reports effects of antibiotics on patient subgroups defined by the presence/absence of factors that are of clinical concern: green phlegm; currently smoking; ‘significant past medical history’; longer illness duration prior to consulting; fever at presentation; and ‘minor’ chest signs (not suggestive of pneumonia).

Although there appear to be some statistically important findings (suggesting patients with significant past medical histories, shorter prior illness duration and those who don’t smoke gain from amoxicillin), Moore et al conclude that:

‘There is no clear evidence of clinically meaningful benefit from antibiotics in subgroups of patients with uncomplicated LRTI ...’.

¹²

So, why the caution?

STATISTICAL MINI-TUTORIAL

There are three main types of subgroup effects which could be considered as ‘interactions’:¹⁶ first, treatment effects in both subgroups in the same direction but of different sizes; second, a treatment effect only in one subgroup; and third, treatment leading to harm in one subgroup but benefit in the other. A trial with 80% power to detect a given size of an overall treatment difference would have only 29% power to detect an interaction of the same size, if the subgroups are of equal size.¹⁶ Where the interaction size is smaller, as is often the case in pragmatic trials, or where subgroup size is uneven, the power falls further. Recommendations for subgroup analyses include that significance (P-value) of individual subgroups should not be reported,¹⁷ that focus should be on interpretation of the confidence intervals, that the size and direction of the anticipated effects in planned subgroup analyses should be identified before analysis,¹⁷ and that a more conservative P-value should be used to take account of multiple testing (known as Bonferroni correction).¹⁸ Here, 18 interaction tests were carried out, so the Bonferroni correction would have lowered the P-value for ‘significance’ from 0.05 to 0.003. Together, the above means this study is likely to be underpowered to detect even large interactions between subgroup and antibiotic use and that only the interaction between past medical history and symptom severity would be deemed unlikely to be due to chance alone.

The second BJGP article from the GRACE consortium, by Hamoen et al,¹³ shows interesting and marked variations between 12 European nations in the use of symptomatic remedies prior to presentation, including (percentages given for extremes of range): antibiotics (0–26%); antitussives (5–21%); mucolytics (0–30%); expectorants (0–15%); antihistamines (0–18%); inhaled bronchodilators (0–5%); inhaled corticosteroids (0–11%); aspirin and other salicylic acid derivatives (0–19%); paracetamol (4–48%); and NSAIDs (0–15%). These mirror between-nation variations in the use of antibiotics,¹⁹ many without any accompanying evidence of benefit.

CONCLUSIONS

The main antibiotic trial article suggests patient safety was not improved by using antibiotics — if anything, their use may compromise safety — more patients experienced side effects than were prevented from experiencing worsening of their illness, and we already know about the dangers of anaphylaxis and antimicrobial resistance. The caveat is that patients with more severe illness may not have been recruited to the study, and the subgroup analysis suggests patients with ‘significant past medical histories’ (most of whom had asthma or COPD) may differentially benefit from antibiotics. Therefore, caution needs to be exercised for the more unwell, and we should continue to follow national guidelines for patients with COPD exacerbations, but most patients with acute chest infections (in whom pneumonia is not suspected and without significant past medical histories) should not receive antibiotics.

Further research is needed to see if there are other ‘special’ patient subgroups more likely to benefit from antibiotics (for example, those with haemoptysis and infective exacerbation of asthma), and to evaluate if other remedies can safely and effectively replace antibiotics. But given the strength of evidence to date, perhaps we should also ask ourselves if 88% of patients with acute LRTI can all be special?

Provenance

Commissioned; not externally peer reviewed.

Competing interests

The authors have declared no competing interests.