Participants

Meeting attendees and their representation are shown in Tables 79 and 80. In the items below, co-investigators are referred to by initials (see Table 79) but other meeting participants by stakeholder group.

TABLE 79

Meeting attendees

TABLE 80

Stakeholder representation

Formal voting was performed via Turning Point software (www.turningtechnologies.co.uk/). Only those attendees who had previously completed rounds 1 and 2 of the Delphi survey were eligible to vote, but involvement in discussion of the switch criteria and outcome set was encouraged from all attendees. Two public contributors who were present voted separately using a paper-based voting system (as they had not been involved in the qualitative interviews). One was the patient and public representative from the study steering committee and had direct experience of a child suffering from the condition and the other was a ‘trials-experienced’ member of the Medicines for Children Young Persons Advice Group. Three clinicians attended the meeting who had not completed the online survey, votes again were collected using a paper-based voting system (two complete voting sets and one partial were produced). These votes have not been included in the results presented in this report.

Presentation of the service evaluation data

• The definitions of complex and simple cases of SA or OM were discussed, as well as the numbers of each case type. Forty-nine cases of misdiagnosis were recorded and the meeting queried if the patients were treated appropriately in each of the cases. The data captured did not cover this information.
• A query was raised regarding whether or not some infection sites were more common than others. Among the study population, the most commonly affected joints were hip, knee and ankle in simple cases and femur and elbow in complex cases (and that this reflects the historical literature).
• The potential issue of missing data was raised, as 7% of children with simple disease and 3% with complex disease did not appear to have received i.v. antibiotic treatment. Although concerning, there are reports of oral therapy alone being used in recent European cohorts to treat bone and joint infections (oral data from Spain presented at European Society for Paediatric Infectious Diseases, Leipzig, May 2015; J Saavedra-Lozano, 2015, personal communication).
• There seems to be a lack of detailed information about the date that the child/parents first noticed symptoms. The date of first presentation at the health-care service provider is recorded, but the meeting participants agreed that this is often days, if not weeks, after first symptoms appear.
• The definition of diagnostic surgery was discussed as many of the surgeries listed as diagnostic would normally be seen as therapeutic. For the service evaluation, diagnosis was defined as the start of i.v. antibiotic therapy and therefore any surgery completed before this was included in the diagnostic category.
• Surprise was expressed at the number of aspiration surgeries. Research nurses reported these based on surgeons’ notes and/or cross-checking discussions with surgeons.
• The participants discussed and concluded that lavage should fall into arthrotomy and that they would expect the numbers of arthrotomy procedures to be much higher as this is standard practice for simple SA. The lower numbers of arthrotomy procedures is very likely a result of their inclusion in the ‘Incision and drainage’ category.
• There were concerns over the low numbers of X-rays of abnormal bone, and it was hypothesised that this can be attributed to failure to follow the norms of standard practice in a few centres lying, thus skewing the results. This discrepancy could be caused by the department/specialist that the child is first referred to, or junior doctors who lack experience.
• The number of blood cultures taken prior to initiation of antibiotic therapy is not available in the existing data set.
• The preliminary results show that in only 42.3% of the 75.8% of children who underwent microbiology tests was at least one micro-organism detected, a figure that may have been expected to be higher.
• Results of the service evaluation suggest that standard practice in the UK is short-course i.v. antibiotic therapy (see Chapter 2). The data also imply that average timing of i.v./oral switch is between 1 and 2 weeks. The data set does not provide information about if location of infection impacted the timing of switch or length of overall antibiotic treatment.
• Concerns were shared regarding discrepancies in dosing of i.v. ceftriaxone. There was general agreement that in clinical practice a high (80 mg/kg) dose should be used for bone and joint infections (the manufacturer, Roche, in the European summary of product characteristics states up to 1000 mg/kg/day), but one participant noted that the BNFC allows a lower dose of 50 mg/kg.
• It was noted that only 40% of children underwent a blood test at the time of switching. It was queried whether or not the blood tests could have been done a few days before, which would be standard practice but not captured in the database. All agreed that this was an important question in relation to future discussions for the trial.
• The results found for length of stay in hospital were discussed; hospital stay could be influenced (1) as a result of patients being transferred to a district general hospital in the case of tertiary centres and (2) by the age of the patient.
• The group noted that there were not as many complications as might be expected, but that this could be due to the short length of follow-up.
• Discussed was whether or not pain at follow-up should be classified as a treatment failure.

Presentation of the qualitative study

• Interviews provided a reminder that both i.v. and oral antibiotic therapy can be traumatic for children; some may tolerate tablets/liquid well and find the i.v. insertion stressful, while others may not be bothered by an i.v. line but find tablets/liquid hard to take/keep down. Parents need to feel that therapy for their child is tailored to their individual case and this may make recruiting to a randomised trial difficult as it removes the element of personalised care.
• Parents expressed their concern that complex conditions might not be suited to a randomised trial and would remove any parental choice regarding the methods of treatment. The data indicated that the period before diagnosis causes a high level of anxiety and a longer safer standard treatment arm would be preferred.
• The qualitative researchers had thought carefully about how to explain a RCT to the children interviewed. They concluded that it is important that all information (patient information leaflet/consent and assent) is tailored to young persons’ needs and that PPI is important at this stage. Parents may be more concerned with safety specifics, but a child/young person may be more focused on what is required of them (e.g. blood test). It is sometimes important to discuss the trial with the child without the parent/guardian present and it was pointed out that there appeared to be a lack of children’s opinions in the qualitative research presentation.
• Timing of consent is important, as parents are unlikely to consent to a trial while waiting for their child to go in for surgery or come round from anaesthetic, for example. It was suggested that discussion of the trial should wait until treatment has started. Once the patient is settled, parents are more likely to be receptive to a trial.
• It was also pointed out that bone and joint infection is perhaps not on a junior doctor’s ‘radar’. Parents felt the that speed of diagnosis was often down to the treating medic, and therefore luck. It was pointed out that a trial would increase awareness of bone and joint infection.
• The group should consider any subgroups of patients that clinicians would not be comfortable randomising.

Discussion and voting results

Oral to intravenous antibiotic switch criteria

Results from voting on oral to i.v. antibiotic switch criteria are presented in Table 81. Overall, resolution of fever for ≥ 48 hours, tolerating oral input (medicines and food/drink) and pain improvement (rather than resolution) were considered important criteria by consensus following the meeting voting.

TABLE 81

Summary of switch criteria voting

Although there was no clear consensus regarding the absolute or relative (to presentation) of the CRP, there was agreement in both rounds 1 and 2 of the Delphi survey that CRP is an important parameter in around 60% of respondents, and among approximately 70% of those at the consensus meeting. In addition, the patient representative felt that a CRP < 20 mg/l was important but not critical.

Discussion about switching criteria:

• CRP levels are an important test for recognising improvement, but picking an arbitrary absolute CRP level may not be appropriate.
• A two-thirds reduction in CRP might not be sufficient reduction if the baseline CRP was very high.
• A very high CRP level could also indicate a complex case as opposed to simple and should therefore not be ignored.
• If CRP was used as a switch criterion, parents could find conflicting evidence online and not wish to consent and/or withdraw their child from the study.
• The overall trend should be more important than any particular fixed value.

Conclusion

Monitoring CRP in terms of absolute value and trend would be important despite consensus not being reached.

Core outcomes voting

A summary of core outcome consensus is presented in Table 82. Overall, the outcomes assessed as ‘consensus in’ voting include rehospitalisation or recurrence of symptoms while on oral antibiotics, treatment failure or recurrence of infection, disability at follow-up, being symptom free at 1 year, limb shortening or deformity, chronic OM or chronic arthritis, amputation or fasciotomy, death, need for paediatric intensive care and line infection.

TABLE 82

Summary of outcomes discussed

Discussion about core trial outcomes

• All would be uncomfortable taking part in a study that included only outcomes that had been voted as critical.
• Critical outcomes would form a core outcome set and other important outcomes would be included.
• Twelve out of 16 participants felt that an outcome should be included for both parental and child stress/anxiety.
• It was agreed that a barrier to collecting quality-of-life data would be that often in these cases the negative impacts may not be seen for several decades after recruitment.

Standard care

Concerns about standard care in a future RCT:

• The absence of any agreement on diagnosis, antibiotic dosing or defined standard care procedures as identified in the service evaluation is a concern for a future RCT. A study aiming to standardise care is a different research question for which it might be hard to attract funding as there is no specific clinically related research question. Part of the rationale for this Dinosaur study was to review the current UK practice and establish the parameters for current standard of care that can be tested against.
• Delay in diagnosing simple bone and joint infections overall in UK settings. The process of running a study could itself help to define the approach to these cases and improve diagnosis and treatment by raising awareness in participating centres. There is therefore a need to ensure that a study is accessible to all and does not take place only in tertiary centres.
• Is the service evaluation coherent with personal experience of clinicians? This is the largest prospective global study to date on childhood bone and joint infections, and part of the problem in this condition is that treatment has been based on personal experience and mostly retrospective small cohort studies.
• Should primary care trusts be included in the future study to ensure that standard care is consistent across the board? Running the future study would be the best way to increase awareness of these conditions and improve normal care.
• Is 3 months’ follow-up, as that used in service evaluation, long enough to capture long-term complications like growth problems? A relatively short follow-up is often standard practice (in routine clinical care of the simple cases being considered for a RCT) but that perhaps conducting this study could improve follow-up times for normal care.
• A future trial needs to consider the cost to the NHS within research costings.

Adherence to potential trial treatments

Adherence to oral antibiotics regimens potentially can influence/disrupt results for trial treatments involving oral antibiotic therapy at home. The items discussed were:

• Parents are often better than doctors or nurses at complying with antibiotic regimens.
• Parents in the qualitative research had reported high levels of adherence to antibiotic therapy as parents had a strong understanding of the consequences of not completing the course. This finding could be influenced by a study effect and selection bias.
• Clinical experience shows concerns about adherence.
• Older children and young adults are often given responsibility for their own treatment by parents and may report a higher level of compliance than is the case.
• Adult RCTs successfully used an oral treatment strategy.

Recruitment and inclusion criteria issues

• Parents often self-inform and have preferences for their child’s treatment (length of i.v. therapy, PICC line, liquid/tablet form).
• Clinician preference for one type of therapy over another has made it harder than anticipated to recruit to some paediatric clinical trials, for example the TORPEDO (Trial of Optimal Therapy for Pseudomonas Eradication in Cystic Fibrosis) trial.⁹² Despite these issues the trial has proceeded and achieves a consent rate of approximately 40%.
• Service evaluation data suggest that short-course i.v. therapy is now more common. Randomising patients to longer than normal treatment may reduce participation.
• Parents are sometimes less willing to take part in randomised trials because of the perceived lack of personalised care. The design and patient/parent information would have to clearly explain this issue, as apparently shorter therapy duration may be off-putting to parents. Parents interviewed in the qualitative research were generally happy with the timing of their child’s switch from i.v. to oral antibiotics as they felt they were well enough to be switched.
• Children with bony (rather than soft tissue) abnormalities on X-ray imaging would need to be excluded from participation in a clinical trial of simple OM or SA. The final feasibility assessment needs to take into account cases from the service evaluation in which no bony abnormalities were found.
• The study population will remain heterogeneous despite aiming for clear definitions because patients take different amounts of time to present at health-care service providers, doctors take differing amounts of time to diagnose the infection and infections progress at differing rates.

Developing the research question

A future study is thought to be important to reduce the use of antibiotics, reduce length of hospital stay and to avoid complications of i.v. therapy.

Possible research questions:

1. For how long should bone and joint infections be treated with antibiotics in total?
2. When should the switch from i.v. to oral therapy occur?
3. Should the study compare early with late switch rather than type of antibiotic treatment?
4. Early switch point of 48 hours compared with longer, such as 9 days (as the latter appeared to be the median time for switch in the data from the service evaluation).

Concerns noted:

• Assessing more than one research question in one study may impact on results, creating smaller groups of patients with more compounding factors and reducing the power of the study.
• The umbrella title of bone and joint infections covers two very different conditions: acute OM is perceived to have a lower risk of long-term problems due to treatment failure than SA, despite the fact that, historically, duration of antibiotic therapy has been shorter in OM than in SA. Differences in standard care procedures need to be examined from the service evaluation to decide if OM and SA need to be evaluated in different study groups.
• Setting an arbitrary switch date would not take different infecting organisms into account. Any defined switch timings should be within a range that is comfortable for clinicians and parents.
• Many clinicians would be uncomfortable switching as early as 48 hours. This would be possible only in a subgroup with mild disease, affecting the recruitment and power of the study.

Conclusion

The meeting concluded that a future study looking at switch timing should be prioritised over one looking at the effect of reducing the overall duration of antibiotic therapy.