The liver is the second most commonly injured intra-abdominal organ after the spleen. However, liver injuries are more often severe than splenic injuries and are associated with higher mortality.1 2 In children, and infants in particular, the horizontal diaphragm and pliable rib cage render the liver an intra-abdominal organ which is afforded little protection from trauma by the bony ribs. Because of its vital role in metabolism and homeostasis, detection of significant injury to the liver is essential.

Abdominal CT scanning is regarded as the “gold standard” for detecting liver injuries. However, in a large series of paediatric trauma patients, 74% of abdominal CT scans were normal and only 3.6% of patients required a therapeutic laparotomy.3 In addition to their cost, CT scans in children are associated with a significant radiation burden4^–7 and with other risks associated with transfer and sedation.8

Alanine aminotransferase (ALT) is a cytosolic liver enzyme. When a blunt force is exerted on the liver, necrosis of the hepatocytes occurs with immediate leakage of their contents (including ALT) into the circulation. The ALT level rises immediately, and levels remain raised for at least 72–96 h.9 10 Most of the increase in ALT has been shown to be secondary to liver injury, with little effect from muscle injury.11 12 Consequently, ALT—which is an inexpensive and readily available biochemical test—in theory provides a good marker for liver injury. However, current data on the correlation of ALT with the presence and/or severity of liver injury are based on small series with variable ALT thresholds which have not generally been explicitly derived by empirical means;10 13^–19 these are summarised in table 1.

Unlike haemodynamically unstable patients where urgent resuscitation with appropriate imaging is necessitated, stable patients (for example, a child sustaining an isolated injury to the right upper quadrant following a handlebar injury) may be suspected of having an isolated liver injury. Such patients are highly unlikely to require operative management.20 In this situation, ALT could provide an alternative screening tool to CT scanning for the identification of liver injury, particularly where access to CT scanning is limited owing to cost or lack of facility, as occurs frequently in rural settings or in the developing world.

The aims of this study are to describe the spectrum of blunt liver injury in children, to determine the optimum threshold levels of ALT indicating the presence of liver injury, and to explore a possible correlation between liver injury severity and ALT levels.

METHODS

Setting and design

The Royal Children’s Hospital (RCH) in Melbourne, Australia is a 250-bed specialist paediatric hospital. It is the only paediatric trauma centre in the state of Victoria, treating approximately 32 000 inpatients annually and serving a population of approximately 5 million in the south-east corner of Australia. Since 1999 the hospital has maintained a dedicated trauma registry. Trauma patients are identified on admission and entered into the registry by a single full-time data manager (CP).

All children admitted to RCH during the 5-year period from 1 July 1999 to 30 June 2004 with radiological, operative or post-mortem evidence of liver injury following blunt or penetrating trauma were identified using the trauma registry. A comparison group was also identified, comprising all children admitted with evidence of intra-abdominal injury other than to the liver. The comparison group included patients admitted between 1 July 2000 and 31 December 2003.

All cases were reviewed to determine the first ALT level at presentation. The patients with identified ALT levels were further analysed. Patients with liver injuries who had an abdominal CT scan performed at the time of their injury were evaluated using the American Association for the Surgery of Trauma scoring system (AAST)21^–23 by two senior paediatric radiologists blinded to the ALT level, clinical history and previous radiology report (PR and SG). Data including demographics, interventions performed and outcome (including injuries sustained) were also retrieved for comparative analysis.

Data analysis

ALT levels in patients with liver injuries were compared with patients with other types of intra-abdominal injury by plotting the receiver operating characteristic (ROC) curves for ALT results in the presence or absence of liver injury. χ² testing (or Fisher exact test if indicated) including odds ratio calculations and two-sample t tests (with or without equal variance assumptions) were used to make additional comparisons between patients with liver and non-liver injuries, depending on whether variables were categorical or ordinal. Skewed data are presented with median and interquartile range (IQR). For a number of highly skewed variables (including ALT, length of stay, Injury Severity Score (ISS)24 and time to blood testing), logarithmic transformations were required. Spearman’s rho was used to evaluate the association between ALT and the severity of the liver injury. One-way ANOVA testing with post hoc Bonferroni multiple comparisons testing for pairwise comparisons was used to compare mean ALT levels across different AAST injury grades.

All data calculations were performed using Microsoft Excel 2003 (Microsoft Corporation, Redmond, 2003) and Stata Version 8.2 (StataCorp, College Station, Texas, USA, 2004).

RESULTS

Sixty patients were identified as having sustained a liver injury; 51 of these patients had ALT levels taken at presentation. One hundred and nineteen additional patients identified as having sustained an intra-abdominal injury other than to the liver were also retrieved; 65 of these had ALT levels recorded. In both groups, patients with ALT available were analysed further. CT scans had been performed in 48 of the 51 patients with liver injury (3 patients who died soon after arrival did not have a CT scan). Of these patients, only 42 had CT scans available for evaluation by AAST grading (as some films from referring hospitals were not obtainable).

Table 2 summarises the general characteristics of both groups. In both groups there was a preponderance of males. Patients with intra-abdominal injury to organs other than the liver were, on average, older than those with liver injury (p = 0.003), which is reflected in differences in the mechanism of injury between the groups (such as the higher incidence of motorcycle use among patients with non-liver injury).

Patients with liver injury tended to be significantly more severely injured than those with non-liver injury in terms of their median ISS (p = 0.016) and their likelihood of requiring mechanical ventilation (p = 0.007). This was due to a higher incidence of concurrent injuries to other body regions (p = 0.001), particularly the head and chest. However, group outcome measures assessed were not meaningfully affected by these differences. Eighteen of the patients with liver injury (35%) also had an injury to another abdominal structure. In the non-liver injury group, the spleen was the most commonly injured organ (45%) followed by the kidney and gut. Length of stay was similar between the groups (p = 0.264). More than 80% of patients in both groups were discharged home.

In the liver injury group only one patient had abdominal surgery for the liver injury itself, and a further two patients required other abdominal surgery (one kidney and one duodenal). In the non-liver injury group there were 14 operations (8 for bowel repair, 3 for partial pancreatectomy, 2 for renal repair and 1 exploratory laparotomy). Although more patients died in the group with liver injury, death rates were not statistically different between the groups (p = 0.097). Details of the five patients who died are summarised in table 3. Only one patient died as a consequence of liver injury. Although all of the liver injuries were at least AAST grade III, the remaining patients all died primarily of head injury.

The median time from hospital arrival to blood being logged into the laboratory for processing was 37 min (IQR 15–82 min) for the liver injury group and 58 min (IQR 23–150 min) for the non-liver injury group (p = 0.020).

Figure 1 shows the ROC curve generated for the sensitivity and specificity comparison for the association between ALT level and the presence of liver injury, with the area under the curve (0.945, 95% confidence interval (CI) 0.905 to 0.985) demonstrating that the test is a good discriminator for the identification of liver injury. Using ROC curve assessment, optimum ALT thresholds were determined to be ⩾162 IU/l (given by the threshold closest to the top left hand corner of the ROC curve) and ⩾104 IU/l (maximising the correct prediction of liver injury status with sensitivity ⩾95%). More complete descriptions of these and other thresholds used by previous investigators13^–15 are given in table 4.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

ROC curve showing sensitivity and specificity for ALT threshold values for combined liver and non-liver abdominal trauma patients (n = 116). ALT, alanine aminotransferase.

Spearman’s rank correlation coefficient for the association between ALT and AAST liver injury grade was 0.503 (p = 0.007, 95% CI 0.236 to 0.700). The scatterplot (fig 2) visually demonstrates this association. Table 5 summarises mean values and confidence intervals for (transformed) ALT levels for each AAST injury grade.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Scatterplot of association between ALT and AAST liver injury grade (n = 42). AAST, American Association for the Surgery of Trauma scoring system; ALT, alanine aminotransferase.

ANOVA testing for differences between mean transformed ALT levels was significant (p = 0.011). However, owing to small patient numbers at some AAST grades (particularly grades I and V), significant intergroup differences were only seen between grades II and IV (p = 0.035). When injuries were stratified into minor (AAST grades I and II) and clinically significant (grades III, IV and V) injuries, a significant (p = 0.004) difference was seen in mean transformed ALT levels using the two-sample t test. In considering only the clinically significant liver injuries in this sample, using an ALT threshold of 104 IU/l identified 100% of these patients (95% CI 89% to 100%) with a specificity of 70% (95% CI 59% to 79%), positive predictive value 57% and negative predictive value 100%.

DISCUSSION

Our paediatric series is the largest one yet published and identifies both liver injury and a subset of clinically significant liver injuries (AAST grades III, IV and V) following blunt trauma. In this sample, few patients with liver injury required operative intervention and only one died of liver injury itself. Most children with blunt liver injury can be managed by expectant observation.

Our data comparing liver and non-liver abdominal trauma indicate that the presence or absence of liver injury can be predicted with a sensitivity of 96% and a specificity of 80% by using an ALT level ⩾104 IU/l, and with a sensitivity of 90% and a specificity of 88% by using an ALT level ⩾162 IU/l. The sensitivity is even higher if only clinically significant injuries are being screened for, with 100% of AAST grade III, IV and V injuries in this sample being identified with an ALT threshold of ⩾104 IU/l. Only one patient was determined to have liver injury in the presence of a “normal” ALT level (<55 IU/l); this was not deemed clinically important (AAST grade II) and no operative intervention was required.

Only one of the previous studies investigating ALT levels in paediatric liver injury appeared to use an empirically determined ALT threshold, using a ROC curve to obtain a threshold value of >250 IU/l.13 However, their sample was comparatively small (less than half the size of the present study) and the authors did not attempt to differentiate between different levels of liver injury severity. Also, although their ROC curve evaluation method was stated in part (maximised sensitivity with a certain minimum specificity), this was not described further.

We also showed that there is a meaningful (although moderate) correlation between the severity of liver injury and the level of ALT when injury severity was stratified by AAST grade. Only one other study has shown a similar relationship. Miller et al16 in 1988 showed that patients with grade III or IV liver injury had a significantly higher ALT level than those with grade I or II injury, and only one other paper10 investigated a relationship, finding no correlation between injury severity and ALT level in a small sample. Although in the current study the relationship between ALT and AAST was statistically significant, there is probably too much variation at each level of both AAST and ALT for one variable to be usefully predictive of the other. Put another way, if a patient has a raised ALT level (of, say, 200 IU/l), then one can be reasonably certain that a liver injury exists but cannot be sure of the extent of that injury. If a patient has an ALT level of 500 IU/l, he/she is almost certain to have a liver injury and that injury is likely (although not certain) to be more severe than the injury to the patient with an ALT level of only 200 IU/l.

Although there were some patients with grade I liver injuries in our sample, there should probably have been far more to be truly representative of the entire population of patients with liver injury as minor injury is (intuitively) far more prevalent than significant injury. It is likely that the majority of these injuries pass without being clinically recognised, far less receiving a CT scan.

It is also worth noting that ALT levels and liver AAST grades on CT scan are not markers of eventual patient outcome, which is mainly correlated with the presence or absence of associated injuries seen25 or the need for surgery (which is itself usually determined by the need to gain haemodynamic control). Furthermore, for the majority of patients the mainstay of treatment is continuous clinical observation, bed rest and avoidance of contact sport, although determining appropriate length of stay and avoidance of contact sport remains controversial. Little is actually known about the risk levels and it must be assumed that many children with a grade I or II liver injury probably never present to a tertiary centre, let alone have a CT scan. It is therefore likely that there exists a large unidentified population without apparent harm.

Limitations of the study

In addition to the low number of patients with low-grade liver injuries, there are a number of other limitations in this study. Previous studies have used various ALT thresholds; within a normal healthy population ALT has a Gaussian distribution, but varies to some extent with age and gender.11 12 26 Although our hospital has routinely used ALT levels of <55 IU/l as a normal reference, there is a notable lack of standardisation between thresholds used in previous studies (table 1) which, in part, prompted the current study. ALT levels in our study were obtained at statistically different (p = 0.020) average times in patients with liver and non-liver abdominal trauma. As patients with liver injury had their ALT taken earlier than those with non-liver injury, they would intuitively have been even higher (on average) had they been taken with the same delay to ALT testing experienced by the non-liver injury group. It is therefore felt that this difference is not clinically relevant. ALT levels rise immediately after injury and remain raised for 72–96 h. There was no standard protocol for obtaining liver function tests (including ALT) in all patients with suspected abdominal trauma, and ALT was assessed more frequently in the non-liver abdominal trauma group than in the patients with liver injury. Liver function tests are performed early in a trauma resuscitation phase and often before full clinical evaluation of the apparent injuries the patient has sustained. The decision to perform liver function tests appeared to be on an “ad hoc” basis by the treating physician and did not seem to differ between groups. However, it could be surmised that, because the liver injury group were statistically more seriously injured, they were more likely to get an extensive blood investigation (including liver function tests). Because of the apparent “clinical randomness” with which liver function tests were performed, it is felt unlikely that any bias has been introduced which could affect the study. Eight of the nine patients who did not have their ALT level recorded at presentation had minor injuries of AIS grade 2 (AAST grade I or II), apart from one patient with AIS grade 5 injury who died from severe head injury soon after arrival.

Owing to the retrospective design, patient history and abdominal findings during physical examination where not taken into account. Finally, our data are representative of a setting where penetrating trauma is extremely rare and where FAST ultrasound scans are not employed.

CONCLUSIONS

ALT appears to be a valuable predictor for the presence of and guide to the severity of paediatric liver injury. In the current sample, an empirically determined ALT threshold of ⩾104 IU/l gave 96% specificity for the detection of liver injury and correctly identified all patients with clinically significant liver injuries. Although any unstable child warrants immediate evaluation, resuscitation and appropriate investigation (often involving CT scan) at a trauma centre, there is a significant group of stable patients in whom there is a clinical suspicion of injury but time is less critical. Although these cases still warrant close clinical observation, there is a role for the use of laboratory evaluation to provide best practice in the reduction of unnecessary radiation exposure, particularly in rural, remote or developing settings where access to CT scanning equipment is poor.