I have a confession to make – I once requested a whole body CT (WBCT) in a paediatric patient.
[Yes, yes, I know!]
To be fair, we didn’t know he was a paediatric patient at the time (although he was obviously young, there is considerable variability in body habitus in peri-pubertal males that can make it very difficult to determine) and owing to his very unusual and associated-with-significant-force mechanism of injury, he was in no fit state to tell us he wasn’t an adult.
I have long justified this, despite my familiarity with and belief in the RCR guidelines for imaging in paediatric major trauma, but a paper published online in JAMA Pediatrics this week has given me reason to question myself (and what a healthy thing that is!).
Where can I find this paper?
As always, we would encourage you to read the paper yourself, appraise it critically and formulate your own conclusions about its findings before reading on. Click the image above to get to the JAMA page (sadly the paper is not open access) or the pubmed page here.
Tell me about the study
In this paper, the authors wanted to know whether whole body CT (WBCT) is associated with lower mortality among injured children, compared with selective CT scanning. This strikes me as a slightly unusual perspective; although I am aware that many health services have not adopted the RCR’s approach to imaging in paediatric trauma, I would have thought that the use of WBCT would imply higher mortality as it would be reserved for those children most seriously injured. I suppose what the authors are asking here is whether there is a benefit in WBCT in children in the context of identifying treatable injuries, which might otherwise have been missed or their diagnosis delayed with deleterious effects.
What did they do?
This was a retrospective, multicentre cohort study using a national database (the National Trauma Data Bank), examining the care of children over a four-year period (Jan 2010 to Dec 2014). The authors decided to look at children aged 6 months to 14 years sustaining blunt trauma, and included those receiving some sort of CT imaging within the first two hours of ED arrival. They hypothesise in the methods that having had a CT scan of some sort means the child was fit enough for transfer to CT and therefore could have undergone WBCT scanning; thus whether or not the scan was limited or a WBCT was clinician-determined (rather than prevented by severity of injury and need for ongoing resuscitation). They excluded transferred patients, so what we are looking at is the initial ED presentation following trauma, wherever that might be.
They have defined WBCT as head-pelvis CT (excluding the stipulation of cervical spine CT as this is not coded for specifically in their data set), versus selected CT if one or more (but not all) of these regions were imaged.
Unlike many papers (!), the primary outcome in this study makes a lot of sense and is arguably patient-oriented (and easy to measure); the authors looked at death in-hospital in the first 7 days after admission. This timeframe was limited in an attempt to differentiate between death from trauma and subsequent death from associated complications (like infection etc).
One of the most interesting methodological aspects of the paper was the use of propensity modelling, something I had never heard of (and have subsequently learned a little bit about!).
Propensity modelling is a way of correcting for a number of variables when comparing groups with the aim of controlling for treatment selection bias. It is sometimes used when randomising subjects isn’t possible but you want to correct for a number of confounding variables. The propensity score is the probability a subject (patient, in this case) would receive a treatment (WBCT, in this case) based on the characteristics of the patient and the department they are being treated in.
In this paper, the authors collected data around hypoxia and assisted respirations, paediatric trauma centre status and number of paediatric beds, paediatric trauma centre status and adult trauma centre status, paediatric trauma centre status and age, hypotension and PRBC transfusion, and PRBC transfusion and tube thoracostomy (chest drain).
The idea is that, just like randomisation occurs in prospective trials to try to even up the distribution of potential confounders between groups, propensity score matching should balance baseline covariates between the groups (as long as they were measured and included). Once you read a little deeper than this you get into complex statistical mathematics…!
There is a nice, fuller explanation in this (not FOAM) paper at JAMA. It describes nicely how the propensity score is used (most commonly):
The most common is propensity score matching, which involves assembling 2 groups of study participants, one group that received the treatment of interest and the other that did not, while matching individuals with similar or identical propensity scores. The analysis of a propensity score–matched sample can then approximate that of a randomized trial by directly comparing outcomes between individuals who received the treatment of interest and those who did not, using methods that account for the paired nature of the data.
There’s also a nice summary at Life in the Fast Lane.
So what did they find?
The study group found 42912 patients at 631 hospitals eligible for analysis. Of these, 8757 (20.4%) received whole body CT – these patients tended to be older and more seriously injured (measured by GCS <9 and/or hypoxia and/or hypotension and/or need for intervention such as chest drain, ventilatory support or blood transfusion). This seems to make logical sense in my world.
Interestingly, patients receiving WBCT were less likely to be seen at a specialist paediatric centre or university hospital. At this stage, it’s unclear whether that reflects an independent variable (are staff at smaller, non-specialist centres more likely to request WBCT, or are more seriously injured children more likely to present there in the first instance?).
Looking at their propensity matching, they found four of their variables to be statistically significant (hypoxia and assisted respirations, paediatric trauma centre status and number of paediatric beds, paediatric trauma centre status and adult trauma centre status, and PRBC transfusion and tube thoracostomy (chest drain)). When stratified by exposure group, the authors found that covariate differences between the WBCT and selective CT groups were no longer present.
405 children in the sample died within 7 days of ED arrival, a rate of 0.9% overall. As you might expect, higher mortality was seen with higher energy transfer in the mechanism of injury (pedestrian vs vehicle had the highest mortality at 1.9%) and in those presenting with a lower GCS (<9 was the cutoff used), with hypotension or requiring PICU admission – it’s no surprise that patients in these subgroups were more likely to die.
In the unadjusted analysis, patients receiving WBCT had higher mortality rates than those selectively scanned (RR 5.0, 95% confidence interval 4.1-6.1), but when the data was adjusted as per the propensity score this difference did not persist. The overall weighted RR fell to 0.8 (95% confidence interval 0.6-1.1) and all of the confidence intervals for the high mortality subgroups crossed 1 when propensity score weighting was used.
Children in the WBCT group also had significantly higher median injury severity scores but again, adjusting for propensity scores also removed evidence of this apparent association.
What does all this mean?
The authors tell us that they have found no evidence of WBCT being associated with lower mortality compared with selective CT, and that subgroup analyses did not show any survival benefit from WBCT in those patients at increased mortality risk. Their postulated expectation for this is twofold;
- That any additional injuries identified were non-life threatening
- That the benefit of WBCT did not offset the risk of a prolonged time in CT.
I’m not sure how much these hold true – in my experience the difference in time taken for WBCT compared with isolated regional CT is actually relatively small, particularly in trauma centres, and from my retrieval experience I would argue that the patient should be stable enough to be out of ED for an unlimited period of time if you are going to take them for scan, regardless of the amount of scanning you intend to do.
There are some things we can draw out from this data, which are of interest.
Firstly, the patients in the unadjusted analysis WBCT group had higher mortality. This makes perfect sense at face value – children with greater injuries are more likely to be injured in more regions and thus require more imaging.
Secondly, the significantly higher proportion of WBCT carried out in institutions with an adult trauma centre reflects our tendency to translate adult practice directly to paediatric practice where this might not be appropriate.
What we don’t know from this paper is whether other imaging modalities – such as plain radiograph or ultrasound – were utilised in place of CT in the selective CT group and whether they contributed relevant information. We also don’t know whether there was an element of inevitability of death for the most seriously injured patients; as in my original case (who did die), does it really matter how much radiation he was exposed to when he had unsurvivable injuries? The counter-argument to this is – how much does it matter that we know exactly the extent of those injuries? The study adds nothing concrete to our imaging-informed prognostication.
Should we use WBCT in practice?
This study does seem to support the RCR guidance in the context of applying a selective approach to CT use in paediatric blunt trauma. As the authors point out, with limited evidence of mortality reduction in WBCT use we should be mindful of the radiation exposure and stick to ALARA principles. It is unlikely we will see an RCT on this topic for ethical reasons (can you imagine arguing for randomising children to WBCT?) so this might be as good as it gets for now.
So for me, the bottom line is to revisit the RCR guidance and to treat the patient in front of me with imaging as indicated. Occasionally, that might mean a whole body CT but it certainly shouldn’t be our default position.