Main findings

Children aged 4–15 years with a history of low ATI consultation frequency at baseline (≤3 ATI consultations during 3 years) who underwent tonsillectomy experienced a net reduction in the mean ATI consultation rate of 3.67 over 6 years’ follow-up, representing 2.46 (95% CI 2.29 to 2.63, p<0.001) over the first 3 years' and 1.21 (95% CI 1.04 to 1.38, p<0.001) per child over the subsequent 3 years’ follow-up, compared with those who did not undergo surgery. This approximates to an annual reduction in the mean ATI consultation rate of 0.61 per child, per year, over 6 years’ follow-up. Tonsillectomy was also associated with much lower odds of ≥3 ATI consultations throughout the 6-year follow-up period.

Strengths and limitations of the study

This is among the first studies to explore the longer term impact of tonsillectomy in a general practice setting in the UK, covering 9 years. We analysed ATI consultation data for up to 6 years’ follow-up in both groups and, to our knowledge, such a long follow-up period has not been studied before. We used routine individual-level general practice data from a nationally representative population-based data set, which limited response bias. The large number of comparisons from a wide population base increased the power of the study.

However, there are also a number of limitations to our study. First, our measure of ATI frequency relied on visits to the GP rather than self-reporting. Although this is an objective measure, we recognise it relies on health-seeking behaviour (HSB) and could, for example, underestimate the frequency of ATI episodes since not every child attends with every ATI episode. Second, although we endeavoured to exclude all children who may have undergone tonsillectomy for the SDB spectrum, we are unlikely to have captured everyone who had surgery for this indication. Third, the losses to follow-up could have created attrition bias, but this is difficult to minimise and this also affects experimental studies. The losses to follow-up were similar in the tonsillectomy and comparison groups over the 3-year and 6-year follow-up periods and were both under 20% at 6 years’ follow-up. Fourth, there were higher baseline ATI consultation and antibiotic means in the tonsillectomy group. However, we adjusted for prior ATI consultations and antibiotics which made the two groups more comparable. Finally, the large size of the study population with demographic and comorbidity variables, which were available from the CPRD data set, allowed us to adjust for a number of potential confounding factors. However, as with most retrospective cohort studies, we were unable to study all known and unknown potential confounding variables and thus there could be residual confounding, which we could not control for.

Comparisons with other studies

A Dutch RCT among children aged 2–8 years who did not fulfil the Paradise criteria4 for severe recurrent throat infection frequency found the effects of tonsillectomy were more pronounced in children who had 3–6 throat infections in the year prior to entering the trial than in those with ≤2 throat infections.8 Among children in this latter group, the authors reported fewer throat infections (−0.21, 95% CI −0.06 to −0.36) and fewer sore throats (−0.60, 95% CI −0.30 to −0.90) per person year in the tonsillectomy group but concluded that these differences were not large enough to be clinically significant, although the median follow-up time was only 22 months.8 Although ours was a cohort study, we similarly found that although tonsillectomy led to a statistically significant reduction in ATI consultation rates among children with low baseline ATI, this was not necessarily clinically significant.

A UK study which analysed CPRD data for annual sore throat consultation rates found that the rates per 1000 registered patients declined by 50% between 1995 and 2000 among children aged 5–16 years.17 The authors reported sore throat consultation rates in 2000 of 59 and 76 per 1000 registered patients among children aged 5–10 and 11–16 years, respectively. However, that study did not investigate the annual consultation rates for an individual child. The authors proposed that the overall decline in consultation rates may reflect that patients were increasingly self-managing minor illnesses.17

We found the natural history among children of all ages who did not have tonsillectomy was for a gradual increase in the number of ATI consultations. This is contrary to previous studies, which reported that the natural history of sore throats is to attenuate over time.18–20 However, our study was based in a community setting and focused on a population of children with low baseline ATI who may have emerging natural history whereas other studies have selected children with higher baseline ATI from the outset. If we had followed children up for longer we may have observed a decline over time.

Clinical implications of the study

Overall, children aged 4–15 years, who underwent tonsillectomy compared with those who did not, experienced an 88% and 76% reduced risk of having ≥3 ATI consultations over 1–3 and 4–6 years’ follow-up, respectively. However, the observed decline in ATI consultations might also, in part, be attributed to adenotonsillectomy operations reducing rhinosinusitis or otitis media infections, which were coded as URTIs.21–23

There was no evidence of an association between tonsillectomy and deprivation level, which is in contrast to some recent studies and reports.24–27 However, we only had practice-level deprivation data and not individual-level data, which could account for the absence of an association.

Children in our study who underwent tonsillectomy in 2001 had ≤3 ATI consultations documented in general practice during the 3 years prior to their operation. This is markedly less than the minimum threshold recommended by national guidelines, reviews and the Paradise criteria for frequency of episodes prior to tonsillectomy.1 ,4 ,20 ,28 ,29 These findings may reflect lower thresholds being followed or perhaps represent suboptimal documentation of sore throat consultations prior to tonsillectomy in primary care settings. Although we assumed that parents would have consulted if the ATI was severe, the low consultation rate could reflect HSB, whereby some parents may not present with ATI episodes if, for example, antibiotics were not issued at prior consultations. However, this depends on the individual GP's attitudes and threshold for prescribing antibiotics for ATIs.

Other possible explanations for these low ATI consultation rates are that we did not have access to the ‘free text’ within individual medical consultation records, which may contain additional details about ATI episodes. Children may have attended with concurrent signs or symptoms and so other diagnostic codes, such as otitis media or specific viral URTIs, may have been recorded with details of the sore throat symptoms included within the ‘free text’. As previously mentioned, overall annual consultation rates for sore throat declined by 50% among children aged 5–16 years between 1995 and 2000 in the UK, which may further account for our lower ATI consultation rates at baseline.17 We did not have consultations from other primary care settings, direct accident and emergency department attendances or hospital admissions, which have been increasing over recent years and may not be well documented within general practice records.30 Therefore, we interpret our findings with the caveat that there could be other possible explanations for the low documentation rate of ATI consultations. However, the strength of our study is that we are consistent in comparing the ATI consultation frequency prior to and subsequent to the index date for each child in a particular general practice, which is where the vast majority of consultations with children occur.10 We recommend future studies should examine individual consultations in greater detail, as well as attendances in community and hospital emergency settings, to attempt to capture ATI consultations within different healthcare settings.

Our findings suggest that there is limited clinical benefit in performing tonsillectomy for children with low baseline ATI consultation rates. This reinforces existing guidance that only children with severe throat infection disease at baseline should be referred for tonsillectomy. Previous research has suggested that the operation is beneficial in the short term for those with more serious ATI morbidity at baseline, although its benefit in the longer term is still not clear and this still needs to be investigated. Tonsillectomy is a major operation so it is important that surgery confers significant individual and societal benefits, including fewer days off work by parents affected by sick children.

Our study also highlights the importance of diligent documentation of ATI consultations within general practice. There is a need to validate diagnostic screening tools among children, such as the Centor criteria and McIsaac scores, for group A β-haemolytic streptococcal throat infections. This may help towards more accurate diagnosis, recording and more appropriate management of ATI consultations in general practice.