PREDICTION TOOL

Criteria abstracted from The Users' Guides to the Medical Literature series in JAMA

Intensive Care Med 2007 33(11):1892-9.[abstract]

Review posted April 03, 2008

1. What is being studied?:
   
   
   1. The study objective:
      
      Independent association of initial venous lactate levels with 28 day in hospital mortality (independent of other variables) in adults presenting to the emergency department and admitted to the hospital in the early stages of suspected clinically significant infection
      
      
   2. The study design:
      
      Prospective, observational cohort study in an urban, university tertiary care hospital.  
      
      
2. Are the results in the study valid?

Primary questions:

1. Was a representative group of patients completely followed up?
   
   Yes. Both the groups (survivor and non-survivor) were followed completely for 28 days. Patients were identified by review of the daily list of emergency department admissions. Outcome (28 day mortality) was collected subsequently using the hospital information system.
   
   
2. Was follow-up sufficiently long and complete?
   
   Yes. The patients were followed for 28 days. Out of 1287 patients with suspected infection who had measurement of venous lactate levels, 73 died in the hospital within 28 days.
   
   
3. Were all potential predictors included?
   
   Yes. The authors included all the pertinent demographic and clinical data pertinent for the study. In addition to the potential predictors that represent data that is easily obtainable on ED presentation, they also used Rivers’ entry criteria (1). Some of the significant criteria included age, several comorbidities, all vital signs, and all routine laboratory data (Table 1 of the article).
   
   
4. Did the investigators test the independent contribution of each predictor variable?
   
   Yes. A large number of variables showed significant association with 28-day in-hospital mortality in univariate comparisons. But due to limited number of deaths in the non-survivor group, the authors could include only six binary predictors without risking overfitting in final multiple logistic regression model.
   
   
5. Were outcome variables clearly and objectively defined?
   
   Yes – 28 day mortality.

3. What are the results?
   
   
   1. What is(are) the prediction tool(s)?
      
      The final multiple regression model (Table 2 of the article) consisted of lactate and five other variables, namely age, blood pressure, malignancy, platelet count, blood urea nitrogen level and their association with mortality.
      
      
   2. How well does the model categorize patients into different levels of risk?
      
      Using a lactate level of <2.5 mmol/l as the reference group, the model compares the odds ratio of death for patients with lactate level of 2.5-4.0 mmol/l and ≥ 4.0 mmol/l. When controlled for age, blood pressure, malignancy, platelet count, and blood urea nitrogen level, lactate levels in the two groups remained strongly associated with mortality. The odds ratio for the group with lactate levels of 2.5-4.0 mmol/l was 7.1 with 95% CI of 3.6 -13.9. Similarly, the odds ratio for group with lactate levels of ≥ 4.0 mmol/l was 2.2 with 95% CI of 1.1-4.2
      
      
   3. How confident are you in the estimates of risk?
      
      Debatable
      
      Lactate levels of ≥ 4 mmol/l
      Though p value for this group was significant and odds ratio well above 1, the CI was very wide.  A very wide interval may indicate that more data should be collected before anything very definite can be said about the parameter.
      
      Lactate levels of 2.5-4.0 mmol/l
      Again the p value was significant and odds ratio well above 1 (though not as high as in the other group), the lower limit of the CI was close to 1, and CI very wide.
      
      The model also calculated the p values and adjusted odds ratio (with confidence intervals) of mortality rates for other variables in the multiple regression model as age, systolic blood pressure, malignancy, platelet count and blood urea nitrogen. Systolic blood pressures, malignancy, platelet count and blood urea nitrogen were also found to be independently associated with mortality in addition to lactate levels. The only other variable with stronger association with death in this model was systolic BP <70 mm Hg  (with an OR of 8.5 and 95% CI of 3.9-18.4). This model had acceptable discrimination (AUC 0.87) and calibration (Hosmer and Lemeshow chi-square  of 6.2).
      
      
4. Will the results help me in caring for my patients?
   
   
   1. Does the tool maintain its prediction power in a new sample of patient?
      
      Not tested. Furthermore, the study had several limitations, all of which are noted by the authors, which were as follows:
      
      
      1. In spite of the large sample size (of 1287 patients), there were only 73 deaths, which could support only six binary predictors (though there were more predictors in univariate regression model) and had to ignore the other ones.
         
      2. Though, this study was a prospective, observational cohort study, the predictor variables were collected from medical records, which may have been incorrectly collected and may have caused some misclassification.
         
      3. The patients with elevated lactate or shock often received evidence based sepsis therapies, which could have again altered the results
         
      4. Some patients may have received some resuscitative therapies prior to lactate sampling, which could again skew the results
         
         
   2. Are your patients similar to those patients used in deriving and validating the tool(s)?
      
      No. My patient population is less than 18 years of age while the study population had a median age of 63 and 74 in survivors and non-survivors respectively.
      
      
   3. Will the results lead directly to selecting or avoiding therapy?
      
      No. Due to lack of confirmatory studies (in pediatrics) and limitations of the study (mentioned above), I would be very cautious in selecting or avoiding therapy based on this study. Lactate levels can be used as an adjunct to other markers to modify the therapy, but further studies are needed in pediatrics to quantify the role of lactate in sepsis.
      
      
   4. Are the results useful for reassuring or counseling patients?
      
      No. Further research is required before counseling patients and families with regard to mortality and risk of death with concomitant lactic acidosis only. However, trends in serum bicarbonate and lactate somewhat characterize the recovery of children with septic shock (2)
     

References

1. Rivers E, Nguyen B, Havstad S, Ressler J, Muzzin A, Knoblich B, Peterson E, Tomlanovich M. Early goal- directed therapy in the treatment of severe sepsis and septic shock. N Eng J Med. 2001; 345: 1368-1377
   
   
2. Dugas MA, Proulx F, de Jaeger A, Lacroix J, Lambert M. Markers of tissue hypoperfusion in pediatric septic shock. Intensive Care Medicine. 2000;26(1):75-83.

Last Updated: April 03, 2008

Fill out our Feedback Form | Editorial Notes | About Us