Penicillin and meningococcal disease: case-control study

Meningococcal infection can be fatal in hours, and a delay in treatment can lead to excess mortality and morbidity. In the UK general practitioners follow guidelines that say they should give parenteral penicillin as soon as possible if they suspect meningococcal disease. But some studies have shown that giving antibiotics early can increase mortality. This study attempts to investigate this controversial matter.

What did the authors do?

The study design was an observational study. There are different types of observational studies, but essentially this means that patients are observed, and certain outcomes are measured. The researchers do not intervene in any way in an attempt to change the outcome.

This particular observational study is known as a case-control study. This is because the investigators are observing cases and controls. In this study the cases were the children who died as a result of meningococcal infection. The control group was children who survived meningococcal infection. Once the researchers identified the two groups they looked at how many children in each group had received penicillin before being admitted to hospital. This would allow them to compare the two groups to see if the use of penicillin was different between children who died and children who survived.

Confounding is an external variable that may effect the outcome of a study by influencing what you are trying to measure. For example, confounding may obscure the relationship between treatment and outcome in an observational study such as this one. The authors took several steps to try and control for confounding. They restricted the study to only children who had been diagnosed with meningococcal disease. They also matched the two groups so that the children were of a similar age and from the same geographical area. This is so that any differences found between the two groups are because of penicillin and differences in demographics are minimised.

The authors also discounted any children whose diagnosis of meningococcal disease was not suspected by the general practitioner. This is because the authors were looking to see if early treatment before admission to hospital caused the children’s outcome. If the general practitioner had not suspected the illness he or she would obviously not have considered giving treatment.

When they analysed the data they adjusted for several factors, including the sex of the child and the severity of their illness. When doing a case control study you must not adjust the analysis by something you have matched between the two groups because you have already tried to control for these confounders. The authors are careful not to do this in their study.

Observational studies are not always the best way to answer a research question because you cannot control factors that may influence the results. In this situation, however, it would have been difficult for the researchers to have performed a more robust study, such as a randomised controlled trial, which would allow them to control these factors.

Meningococcal infection is a severe illness, and current guidelines are to give penicillin as soon as it is suspected. It would not be ethical for the researchers to intervene with this process. There are also logistical difficulties with this question, which meant that a case control study was the best option. For example, because doctors have to act quickly if they suspect meningococcal infection it would not have been feasible to ask them to worry about enrolling their patient into a study and obtaining consent during the initial presentation.

Data collection

To ensure that the study provides robust results it is important to be accurate in the data collection. These authors did this by collecting and corroborating information from six different sources, including questionnaires with the parents, telephone interviews with the general practitioners, copies of primary care records, hospital referral letters, records of complaints made to health services, and the hospital case record. Because this study was retrospective the authors used all these different sources to make sure the sequence of events, the diagnosis, and the severity of the child’s illness was as accurate as possible. Any mistakes here could mean that the results aren’t true. For example, they didn’t want to include children who had not been diagnosed with meningococcal disease.

The authors wanted to look at the severity of the child’s illness to see if this made a difference to whether they received penicillin or whether it affected their outcome. To do this the authors needed a way to assess severity of illness objectively. They didn’t want to rely on the memory of the participants or on their subjective opinion because this can lead to bias in results. To do this they looked at the following criteria before hospital admission, 999 or blue light ambulance call, circulatory collapse, and loss of consciousness. Once children were admitted to hospital the severity of illness was assessed by calculating the Glasgow meningococcal septicaemia prognostic score (GMSPS). They categorised children as having severe illness if one or more of these criteria were present.

Once the authors had all the information they needed they calculated the likelihood of children in the two different groups dying from their illness. They did this using odds ratiosadjusted by logistic regression. Logistic regression takes into account variables that might affect the outcome of the study. The variables they used were previous use of antibiotics, sex, duration of illness, meningococcal serotype, presence of haemorrhagic rash, presentation with septicaemia, and severity as assessed by the general practitioner. This allows the authors to say more confidently that any differences found between the two groups were not caused by any of these factors.

What did they find?

As you can see from the table in the paper (reproduced here) children who received penicillin were seven times more likely to die. Out of those who survived the group that received penicillin were also more likely to have complications. However, we can also see that those with more severe disease were also more likely to have complications. This does not mean that giving penicillin leads to a worse prognosis: it is important to think of other factors that might have influenced the results. The researchers found that the commonest reason for general practitioners not to give penicillin was because the diagnosis was uncertain. The results show that more severely ill children were more likely to receive penicillin. This might explain why those who received penicillin had a worse outcome, as you would expect children who were more unwell to have a worse prognosis.

Figure 1 in the paper (reproduced here) also supports the theory that penicillin was not responsible for a worse prognosis. This graph shows the severity of illness in relation to time taken to arrive at hospital for the two groups. If it was true that administration of penicillin worsened prognosis you would expect this group to deteriorate in the time after receiving treatment to arrival at hospital. What this graph actually shows is that increased length of time between seeing the general practitioner and admission to hospital was weakly associated with decreased severity of disease, and this is the case for children who received or did not receive penicillin. This means that severely ill children were rushed to hospital faster than the more stable children.

Time from general practice consultation to hospital admission (hours)
Severity of illness on admission to hospital (GMSPS score) according to time from general practice consultation and whether or not penicillin had already been administered

The slopes of the two lines for penicillin and no penicillin are almost the same. Essentially the graph shows that there is no evidence that the combination of receiving penicillin and increased time to hospital led to increased severity at admission.

It may have been possible for the authors to eliminate the confounding of severity if they had allowed for more than two groups, or if they had been able to give more accurate information about this. In this study they divided the children into severely ill or not severely ill. If they had used a range of severity of illness they would have had more information and the results would have been more accurate. It was difficult with this study design to provide accurate information about the severity of the illness because it was performed retrospectively and relied on recall.

What does this mean?

The interpretation of this study is that children are more likely to receive penicillin if they are considered to be severely ill and if a general practitioner is confident in the diagnosis of meningococcal disease. Children who are more severely affected are also more likely to die from their infection or to have complication. Although this study set out to answer the question of how penicillin administration affects outcome in children meningococcal infection, it does not give a definitive answer.

The authors conclude that the answer to this question is only likely to be achieved by conducting a randomised controlled trial, however, this would be difficult because doctors follow national guidelines to give penicillin when they suspect a diagnosis of meningococcal disease. It is unlikely that a study which disagrees with these guidelines would be able to get ethical approval.

I thank John Fletcher, clinical epidemiologist, BMJ, for reviewing the draft of this article.

   

 Printable version      Download PDF    E-mail this to a friend    Respond to this article