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

Serial blood lactate measurements predict early outcome after neonatal repair or palliation for complex congenital heart disease.

Charpie JR, Dekeon MK, Goldberg CS, Mosca RS, Bove EL, Kulik TJ.

J Thorac Cardiovasc Surg. 2000 Jul;120(1):73-80. [abstract]

Reviewed by Avihu Gazit MD, Washington University School of Medicine, St. Louis Children's Hospital

Review posted March 6, 2005

I. What is being studied?

Study objective:

To determine whether a change in lactate level over time was predictive of poor outcome defined as death within the first 72 hours or the need for ECMO in infants < 1 month of age undergoing complex cardiac surgical palliation or repair.

Study design

A prospective observational study

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?

Forty-two neonates who required cardiopulmonary bypass for repair or palliation of complex congenital heart disease within the first month of life were followed for 72 hours. An additional 4 infants were included from a pilot study and followed for 72 hours as well. The authors divided the patients into two groups. The first group included 34 patients who underwent a Norwood type 1 palliative surgery. The second one included 12 patients who underwent a biventricular surgical repair or palliation. The reason for that is the significantly increased risk for catastrophic events involving imbalance in pulmonary and systemic circulation and coronary ischemia in the single ventricle type palliations compared to the biventricular repairs/palliations.

The study plan required a relatively short follow-up which allowed for excellent compliance. All patients were followed for 72 hours or until death. This short period of time although representative of the majority of sudden death episodes does not represent the long term outcome of either single ventricle or two ventricle repair/palliation.

2. Were all potential predictors included?

Laboratory predictors were initial lactate level, an increasing lactate level over time, and bicarbonate levels. The change in lactate level over time was dichotomized as < 0.75 mmol/L per hour and ≥ 0.75 mmol/L per hour. Clinical predictors were cardiopulmonary bypass time, circulatory arrest time and performance of Norwood operation. Other possible predictors that were not included are weight, hemodynamic and respiratory status before surgery (need for inotropic support or prostaglandin E1,and mechanical ventilation), associated conditions and malformations, and adequacy of the repair as demonstrated by transesophageal echocardiogram.

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

No, the investigators describe the isolated (bivariate analysis) contribution of initial lactate level, increasing lactate levels over time, bicarbonate level, cardiopulmonary bypass time and circulatory arrest time. Although they discuss using multivariable analysis in the methods, they do not present the results of this analysis in depth (they do state that bypass and circulatory arrest times, in multivariable analysis, were not independent predictors of outcome).

4. Were outcome variables clearly and objectively defined?

Outcome definition is clear and objective. A good outcome was defined as survival beyond postoperative day 3, and a poor outcome was defined as death or the need for ECMO. Although these outcomes are easy to measure, there is variability in these at the clinical level. In other words, one physician might have put a patient on ECMO while another might not. In both cases the infant might have survived - at least for three days. Conceivably some infants could be kept alive with very aggressive support for another day and then died at day 4, but they would be deemed as having a "good outcome.

III. What are the results?

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

A. All patients combined

1. Initial lactate levels

1. Mean initial lactate level was significantly greater in patients who died or who required ECMO support (9.4 ± 3.8 mmol/L) then in patients with a good outcome (5.6 ± 2.1 mmol/L; p=.03).
2. Initial lactate level > 6 mmol/L had a low positive predictive value (38%) for poor outcome.

2. Increasing lactate levels

Table 1

	Poor outcome	Good outcome
Lactate increase ≥ 0.75 mmol/L/hour	8	0
Lactate increase < 0.75 mmol/L/hour	1	37
Totals	9	37

1. A change in lactate level of 0.75 mmol/L/hour or more was associated with death or the requirement for ECMO support (p < 0.0001) and predicted a poor outcome with an 89% sensitivity value, a 100% specificity value and a 100% positive predictive value. In other words, almost 90% of the patients with a poor outcome were identified with this test and of those with a positive test, all had a poor outcome.
2. A change in lactate level < 0.75 mmol/L/hour has a negative predictive value for poor outcome of 97% (p < .0001). In other words, 97% of patients with a lactate rise < 0.75 nmol/L/hr had a good outcome.

3. Bicarbonate levels:

1. Admission bicarbonate levels were significantly lower in patients with a poor outcome (20 ± 2.3 mmol/L) compared with patients with a good outcome (25 ± 3.8 mmol/L; p = 0.0003).
2. By 18 hours, bicarbonate mean levels were not significantly different. A minimal correlation was found between admission lactate levels and bicarbonate measures.

4. Cardiopulmonary bypass time and circulatory arrest time:

No substantial difference in cardiopulmonary bypass time and circulatory arrest time between patients with a good outcome (87 ± 40.2 minutes and 45.6 ± 12.3 minutes respectively) and patients with a poor outcome (84.9 ± 20.3 minutes and 44.2 ± 8.2 minutes, respectively). Multivariate analysis (with an adjustment for both Norwood operation and change in lactate level over time) also did not demonstrate an association of cardiopulmonary bypass time or circulatory arrest time with outcome.

B. Single ventricle palliation group.

1. Increasing lactate levels

Table 2

	Poor outcome	Good outcome
Lactate increase ≥ 0.75 mmol/L/hour	7	0
Lactate increase < 0.75 mmol/L/hour	1	26
Totals	8	26

1. A change in lactate level of 0.75 mmol/L/hour or more was associated with death or the requirement for ECMO support (p < 0.0001) and predicted a poor outcome with an 87% sensitivity value, a 100% specificity value and a 100% positive predictive value. In other words, almost 90% of the patients with a poor outcome were identified with this test and of those with a positive test, all had a poor outcome.
2. A change in lactate level < 0.75 mmol/L/hour has a negative predictive value for poor outcome of 96% (p < 0.0001). In other words, 96% of patients with a lactate rise < 0.75 nmol/L/hr had a good outcome.

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

The authors suggest a dichotomized model that will give early indication for poor outcome based on one predictor, a maximum increase in lactate level greater than 0.75 mmol/L/hour. This cut-off value was derived from an earlier pilot study. No attempt was done to adjust the level of increase using the data derived from the current larger study with new ROC curves. In any case, even using this cut-off value shows a very good correlation between the predictor and poor outcome in patients who undergo a Norwood 1 type palliation. What we do not know is whether other findings at the time, e.g., perfusion, heart rate, blood pressure, mixed venous saturations, etc., would have been equally powerful predictors.

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

The tables below outlines the 95 % confidence intervals (1) for the sensitivity, specificity and positive predictive values, correlating the rise in lactate level (≥ 0.75 mmol/L/hour) in both groups (table 3), and in the single ventricle palliation group (table 4) to poor outcome.

Table 3: All patients

Statistic	Value	95% confidence interval
Sensitivity	0.889	0.678-0.889
Specificity	1	0.949-1
Positive predictive value	1	0.762-1
Negative predictive value	0.974	0.924-0.974

Table 4: Single ventricle palliation group

Statistic	Value	95% confidence interval
Sensitivity	0.875	0.644-0.875
Specificity	1	0.929-1
Positive predictive value	1	0.736-1
Negative predictive value	0.963	0.895-0.963

For both tables, the 95% Confidence interval for the sensitivity is wide and suggests that larger sample is needed in order to assess more accurately the sensitivity of the predictor. The same applies to the positive predictive value. The 95% confidence intervals for the specificity and negative predictive value are narrower and suggest that the predictor may be applied more accurately to describe good outcome. In other words, we can be fairly confident that if the lactate level is not rising more than 0.75/hr, then the patient will have a good outcome (or at least survive the first 72 hours!).

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?

Not applicable. They did not test this in a separate sample of patients.

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

The patients cared for in our PICU include the same spectrum of diagnoses described by the authors. Unfortunately most of our single ventricle repairs are done with the Sano modification (2), a factor that might change the incidence of the catastrophic events significantly. (The Sano modification uses a non-valved polytetrafluorourethane shunt between the right ventricle and the pulmonary artery, instead of the classical systemic to pulmonary or "modified BT" shunt.) Therefore it will be hard to use this model in our unit. There are also many unit and surgeon-specific aspects of the care of these complex patients that may affect outcome.

3. Does this tool improve your clinical decisions?

In order for this tool to improve my clinical decisions a study needs to be conducted in a single ventricle palliation using the Sano modification and in a larger number of patients. That will probably require the participation of several centers.

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

Taken as part of a much broader clinical assessment that will include information on the adequacy of the repair and the amount of inotropic support needed initially, the model suggested by the authors might give an indication as to the outcome of the patients, definitely not as a sole predictor. I would not use this model to reassure or counsel patients.

References

1. Statistical Methods for Rates and Proportions (2nd Ed.) Section 5.6, by Joseph L. Fleiss (Pub: John Wiley & Sons, New York, 1981)
2. Sano S, Ishino K, Kado H, Shiokawa Y, Sakamoto K, Yokota M, Kawada M. Outcome of right ventricle-to-pulmonary artery shunt in first-stage palliation of hypoplastic left heart syndrome: a multi-institutional study. Ann Thorac Surg. 2004 Dec;78(6):1951-8; discussion 1951-8. [abstract]

 

Comments

Back to the EB Journal Club Index