Invited commentary: “Identifying traumatic significant haemorrhage is challenging for patient with low and intermediate risk, not when bleeding is obvious”

We would like to thank the authors for their valuable comments on our study, wherein we investigated how pre-hospital lactate (P-LACT) measurements could be used to predict the need for (ongoing) in-hospital blood product transfusion in patients attended by HEMS with major traumatic haemorrhage.

As mentioned in our article, the algorithm we developed is a decision support tool, which means that it should be used in conjunction with other parameters, such as clinical gestalt in a heuristic approach to estimate transfusion requirements. The cut-off value of a P-LACT < 2.5 mmol/l used in our population yielded a sensitivity of 80% (corresponding to a low probability of major haemorrhage as the authors rightly mention), and hence was inadequate to be used in isolation. The SOP in our service states that a P-LACT < 2.5 mmol/l is used in conjunction with an SBP > 100mmHg to identify patients who have a low probability of major hemorrhage. This is supported by a recent publication of Gaessler et al. (2023) wherein the authors show that P-LACT and SBP are complimentary in terms of predictive probability [1].

To identify patients with a high likelihood of major haemorrhage requiring in-hospital transfusion, a P-LACT of 6.0 mmol/l was used, as at this this point the predicted probability curve (Fig. 2 in our original article) starts to flatten: using a higher cut-off would not have yielded a higher specificity, whereas a lower cut-off would have dropped specificity whilst not yielding a much higher proportion of the population meeting the cut-off criteria (n = 13, 6.7% for a lactate of 6.0 mmol/l vs. n = 17, 8.7% for a lactate of 5.5 mmol/l). Although we agree that it is likely that many patients with a lactate > 6.0 mmol/l will show clinical signs of shock, 5/13 patients had an SBP > 100 mmHg on first occasion, two of whom also did not exhibit tachycardia. In these patients P-LACT may still be a useful tool. Despite this however, the major challenge remains to identify the bleeding patients in the P-LACT group of 2.5-6 mmol/l, and serial measurements may be the way forward in this group.

Finally, we acknowledge that transfusion requirement is not always a good surrogate to use for outcome, especially not when confounding by indication may be present: using lactate may result in transfusing more patients in the pre-hospital setting, which again may result in a lower threshold to continue transfusion in-hospital. However, as 2/3 of the patients in our cohort received a massive transfusion (> 10 units PRBC within 24 h) rather than a major transfusion, we think transfusion requirement was a reasonable surrogate for risk of death from bleeding in our population. We agree however, that ideally outcome studies should be performed using hard endpoints to confirm this.