Criteria abstracted from The Users' Guide to Medical Literature, from the Health Information Research Unit and Clinical Epidemiology and Biostatistics, McMaster University

Highlighted lines and questions below provide links to the pertinent description of criteria in The EBM User's Guide, now available at the Canadian Centres for Health Evidence

Article Reviewed:

Please visit the new Evidence Based Journal Club Reviews

Blood lactate as an early predictor of outcome in paracetamol-induced acute liver failure: a cohort study

Bernal W, Donaldson N, Wyncoll D, Wendon J.

Lancet. 2002 Feb 16;359(9306):558-63. [abstract]

Reviewed by Parthak Prodhan MD, Natan Noviski, MD Department of Pediatric Critical Care, Massachusetts General Hospital, Boston, MA

Review posted June 1, 2004

I. What is being studied?

Study objective:

1. To evaluate if the use of arterial blood lactate measurement early after admission (median 4 h) and after fluid resuscitation (median 12 h) in patients with paracetamol (acetaminophen) -induced acute liver failure identifies those likely to die without liver transplantation.
2. To compare these threshold lactate levels with the existing King's College Criteria (KCH) criteria for prognostic accuracy and speed of identification of patients who will not survive without transplantation.

[The KCH selection criteria are currently, widely used as a guide for emergency liver transplantation in paracetamol-induced acute liver failure. In its current state the KCH criteria has good specificity when predicting the need for transplantation (1). Only about 15% survive without transplantation. However it is not sensitive enough, as 20% to 35% of those on medical management still die (2).]

Study design

Two single-center cohort studies in patients with paracetamol-induced acute liver failure admitted to a tertiary-referral intensive-care unit.

The first was a derivational (retrospective) cohort: this included an initial sample of 103 patients (median age 35 y) to determine threshold lactate values (early after admission and after fluid resuscitation) that best identified individuals likely to die without transplantation.

The second was a validation (prospective) cohort: this included a sample of 107 patients (median age 36 y).

Predictive value and speed of identification using the blood lactate level threshold at these two time points were compared with those of the existing KCH criteria.

II. Are the results of the study valid?

Note: These questions follow from Randolph AG et al. Understanding articles describing clinical prediction tools. Crit Care Med 1998;26:1603-1612. [abstract]

1. Was a representative group of patients completely followed up? Was follow-up sufficiently long and complete?

All patients were followed up until death, survival or liver transplantation. The follow up was complete and adequate (in both the initial retrospective derivational and the prospective validation studies) to evaluate the intended outcomes. The derivational (retrospective) cohort had some missing initial data, but it did not affect the derivation for statistical analysis. Similarly 14 surviving patients in the validation (prospective) group did not have post-resuscitation lactate concentrations measured. The study does not mention about how long did someone live to be labeled a survivor nor does it mention the duration of the follow-up period to define the outcome variables.

2. Were all potential predictors included?

Yes, all potential predictor's variables were included in this analysis. Potential predictor variables were taken from an established ICU program dataset for the first 24 hours of the hospital stay of each patient. This included demographic, biochemical and clinical variables used to calculate APACHE II and III scores. The lactate levels were measured early after admission at a median of 4 hours [interquartile range 3-6] and after fluid resuscitation at a median of 12 hours [interquartile range 7-15]. In the initial derivational cohort, when the timing of ingestion was known, these values were obtained at a median of 56 hours and 64 hours after the overdose respectively. N-acetylcysteine was infused in all admitted patients for 24 hours until the INR was below 2.

3. Did the investigators test the independent contribution of each predictor variable?

Yes. For the derivational (retrospective) cohort, first, a multiple linear regression model was used to model early lactate concentration, accounting for the effects of demographics, biochemical, and clinical variables. Thereafter, a univariate logistic regression analysis of admission clinical predictors of a fatal outcome was carried out. Further, the variables with significant association were introduced and analyzed by multiple logistic regression to determine the odds ratio for the variables independently associate with increased mortality.

4. Were outcome variables clearly and objectively defined?

Yes. The main outcome measured which was death without transplantation was clearly defined. However, the study does not mention if there were any deaths once the patients were transplanted and, if yes, whether these patients were included in the analysis. However, an appropriate analysis considering all patients transplanted as non-survivors was also carried out in this study.

III. What are the results?

1. What is(are) the prediction tool(s)?

Blood lactate levels is the prediction tool used.

The investigators demonstrated in the derivational (retrospective) cohort as well as the validation (prospective) cohort, a statistically significant relationship between higher median lactate levels (in both the both the early after admission samples and in post fluid resuscitation samples) and survival using a multivariate logistic analysis model.

In the derivational (retrospective) cohort, the median lactate was significantly higher in non-surviving patients than in survivors both in the early samples (8.5 [range 1.7-21.0] vs 1.4 [0.53-7.9] mmol/L, p<0.0001) and after fluid resuscitation (5.5 [1.3-18.6] vs 1.3 [0.26-3.2], p<0.0001).

Similarly in the validation (prospective) cohort, the median lactate concentration was significantly higher in non-surviving patients than in survivors both in the early samples (4.9 [range 1.04-26.5] vs 1.49 [0.50-7.37] mmol/L, p<0.0001) and after fluid resuscitation (4.17 [0.78-28.0] vs 1.7 [0.46-6.10], p<0.0001).

In the multivariate logistic analysis on the derivational (retrospective) cohort, the odds of a fatal outcome using the early lactate threshold of > 3.5 mmol/L is 43 times that of patients with blood lactate levels ≤ 3.5 mmol/L (95% CI 2.7-66). Similarly, the odds of a fatal outcome using the post-resuscitation lactate levels of > 3.0 mmol/L is 63 times that of patients with blood lactate levels ≤ 3.0 mmol/L (95% CI 10.4-385).

A receiver operating characteristic (ROC) analysis showed the best area under the curve with the early blood lactate threshold concentration > 3.5 mmol/L (sensitivity 82%, specificity 96%, accuracy 90%) and the post-resuscitation blood lactate concentration threshold of > 3.0 mmol/L (sensitivity 82%, specificity 96%, accuracy 90%) for identifying patients who would not survive. Accuracy is the degree to which, on average, a test represents the true value (that is, is unbiased). Accuracy is measured by the area under the ROC curve. An area of 1 represents a perfect test; an area of 0.5 represents a worthless test.

The application of the threshold blood lactate values of early and post-resuscitation lactate concentrations was explored prospectively in the validation cohort as a prediction tool and compared with the KCH criteria applied to the same cohort. The lactate levels per se were not used to determine who would need transplantation. Two assessors from the study (blinded to patients identity, outcome, and the variables required for the criteria that were not being assessed) independently reviewed clinical records of selected patients on two separate occasions to assess whether and at what time after admission lactate or KCH criteria were satisfied. Patients were judged to have met each of criteria only if there was agreement between the assessors.

2. How well does the model categorize patients into different levels of risk?

The authors have been successful in their attempt to use the early lactate threshold > 3.5 mmol/L and the post-resuscitation lactate threshold > 3.0 mmol/L as a prognostic indicator.

The table below shows how each of the indicator (alone or in combination) fare in terms of sensitivity, specificity, accuracy, positive likelihood ratio, negative likelihood ratio and the median time (in hours) from admission to criteria being fulfilled.

Table 1: Assessment of arterial blood lactate measurement and King's College Criteria (KCH) as prognostic indicators of death from validation sample

Indicator	Sensitivity	Specificity	Accuracy	PLR*	NLR*	Time* (h)
Early lactate threshold > 3.5 mmol/L	67%	95%	89%	13	0.35	4
Post-resuscitation lactate threshold > 3.0 mmol/L	76%	97%	93%	30	0.24	12
Either of these criteria	81%	95%	92%	16	0.21	4
KCH criteria	76%	94%	91%	15	0.25	10
KCH criteria and either lactate criteria	95%	91%	92%	11	0.05	4
Post-resuscitation + KCH criteria	91%	94%	93%	14	0.1	12

* PLR = Positive likelihood ratio; NLR = Negative likelihood ratio; Time(h) = median time from admission to criteria being fulfilled

In the context of prediction of outcome in acute liver failure, likelihood ratios indicate the extent to which fulfillment of a particular criteria will increase or decrease the pretest probability of death. A positive likelihood ratio (PLR) indicates the extent to which a positive test result will increase the pretest probability (20%) of death without transplantation, and the negative likelihood ratio (NLR) the extent to which a negative test will reduce this probability.

The above table clearly shows that the combination of KCH criteria and either lactate criteria or the post resuscitation + KCH criteria improves upon the KCH criteria. When the KCH criteria and either lactate criteria were combined together it was more sensitive (95% vs 76%), had a lower PLR (11 vs 15), but a better NLR (0.05 vs 0.21), than KCH criteria. It identified patients at risk of death earlier (4 h vs 12 h) than the KCH criteria. Addition of the post-resuscitation lactate concentration to KCH criteria increased sensitivity from 76% to 91% and lowered NLR from 0.25 to 0.10.This combination was most accurate to identify patients at risk for death (93%) but did not identify patients sooner than KCH criteria alone (12 h vs 10 h).

To decrease the proportion of patients who die without identification as transplantation candidates, any criteria should have a lower NLR than the currently used criteria, but an equivalent or better PLR. The simple addition of the early and post resuscitation blood lactate concentrations to the KCH criteria improved speed of identification and NLR value but decreased the PLR (see table 1 above). Thus, the combination of the post fluid resuscitation (median 12 h) arterial blood lactate of 3.0 mmol/L and the KCH criteria is the best compromise. It has a similar PLR to the KCH criteria; however this combination has a lower NLR. This would result in a post-test probability of death in a validation sample of less than 2%.

A simple method of deriving this number is to look at the Fagan likelihood ratio nomogram (3). The < 2 % value is reached by intersecting the pre-test probability (20% in this study) with the negative likelihood ratio (0.1 in this study) intersect for the combination of KCH criteria and post-resuscitation value > 3 mmol/ L on this nomogram. This gives a post-test probability of death at < 2%. In contrast, for the KCH criteria only, with a negative likelihood ratio of 0.25 and the pre-test probability of 20%, it gives a post-test probability of death for those not meeting the KCH criteria only between 5% to 10%.

When the patients who were transplanted were classified as non-survivors, the sensitivity, specificity, accuracy, PLR, and NLR for the lactate criteria were 86%, 95%, 93%, 17, and 0.15, respectively. KCH criteria gave values of 79%, 95%, 91%, 15, and 0.22. Lactate criteria identified non-surviving patients at a median of 4 h (3-13) after admission and KCH criteria did so at 12 h (3-32; p =0.003)

8 patients in the validation sample underwent transplantation. Median early lactate was 4.45 mmol/L and that post-resuscitation was 5.56 mmol/L. Lactate criteria were met at a median of 4 h (4-32 h) and the KCH criteria at 12h (4-32) after admission (p=0.13).

3. How confident are you in the estimates of the risk?

In the multivariate logistic analysis, the odds of a fatal outcome using the early lactate threshold of > 3.5 mmol/L is 43 times that of patients with blood lactate levels ≤ 3.5 mmol/L, the confidence intervals for the odds ratio on is wide, however the lower confidence interval is still above 1. Similarly, the odds of a fatal outcome using the post-resuscitation lactate levels of > 3.0 mmol/L is 63 times that of patients with blood lactate levels ≤ 3.0 mmol/L, the confidence intervals for the odds ratio on is very wide. However these values are significant as the lower confidence interval value is 10.4.

Despite the wide confidence intervals in estimates of lactate in both the derivational and the validation cohorts, this study shows the ability of lactate level estimation in improving upon the predictive abilities of an already widely utilized patient selection criteria.

IV. Will the results help me in caring for my patients?

1. Does the tool maintain its prediction power in a new sample of patients?

Yes, the tool [the blood lactate levels measured early after admission (median 4 h) and after fluid resuscitation (median 12 h)] does maintain its prediction power in a new prospective consecutively selected validation patient sample.

2. Are your patients similar to those patients used in deriving and validating the tool(s)?

The study included patients above the age of 16 years of age. However, the mean age for the initial derivational (retrospective) study and the validation (prospective) cohort were 35 years and 36 years respectively. Thus a subset of these patients (teenagers > 16 yr age) who are cared for in a pediatric intensive care unit would benefit from this study. However, these results cannot be entirely applied to all pediatric patients due to the age group in which this study was carried out.

3. Does the tool improve your clinical decisions?

Yes, this tool improves my clinical decision making process in patients with paracetamol-induced acute liver failure in identifying those likely to die without liver transplantation. Arterial lactate measurement could improve the speed and accuracy of selection of appropriate candidates for transplantation.

However the following caveats should be kept in mind while using this as a decision tool:

1. The results should not be extrapolated to levels of lactate measured soon after paracetamol ingestion.
2. The results should not be extrapolated to under-resuscitated patients.
3. This studies applicability to patients only above 16 yrs of age.
4. Further it is unknown if concomitant exposure to alcohol or other hepatotoxic agents which would confound the results as it was not reported as part of the results.

4. Are the results useful for reassuring or counseling patients?

This study attempts to improve identification, sensitivity and facilitate earlier transplantation in these patients. Arterial blood lactate measurement rapidly and accurately identifies patients who will die from paracetamol-induced acute liver failure. Its use could improve the speed and accuracy of selection of appropriate candidates for transplantation.

The authors thus propose a modification of the currently used KCH criteria for transplantation in paracetamol- induced acute liver failure.

Strongly consider listing for transplantation if:

• Arterial lactate concentration is > 3.5 mmol/L after early fluid resuscitation.

List for transplantation if:

• Arterial pH is below 7.3 or arterial lactate concentration is above 3.0 mmol/L after adequate fluid resuscitation.

Or concurrently

• Serum creatinine is above 3000 micromol/L, INR is above 6.5, and there is encephalopathy of grade 3 or greater.

The model may be useful to treating physicians for counseling patients above 16 years of age with paracetamol-induced acute liver failure. This guides the treating physician for the identifying the need for liver transplantation in these patients and to identify those likely to die without liver transplantation. However this model requires validation in the age group below 16 years of age.

References:

1. O'Grady JG, Alexander G, Hayllar KM, et al. Early indicators of prognosis in fulminant hepatic failure. Gastroenterology 1989;97:439-45. [abstract]
2. Shakil A, Obaid KD, Mazariegos GV, et al. Acute liver failure: clinical features, outcome analysis, and applicability of prognostic criteria. Liver Transpl. 2000;6:163-9. [abstract]
3. Sackett D, Strauss S, Richardson WS et al. Evidence based medicine: how to practice and teach EBM, 2nd ed. Edinburgh: Churchill Livingstone, 2002.

 

Comments

Back to the EB Journal Club Index