Special Feature: Surrogate Outcome Measures in Critical Care: It’s the Mortality, Stupid!

By Gordon D. Rubenfeld, MD MSc

The ultimate goal of medicine is to improve health in ways that matter to patients. A variety of outcomes are important to patients, including symptoms, quality of life, duration of life, quality of dying, the effect of their health care on their loved ones, and the cost of medical care. Because of the importance of these outcomes to patients, they are referred to as "patient-centered." Ideally, clinicians will offer, insurers will pay for, and patients will have the opportunity to use treatments that have been shown to improve patient-centered outcomes.

Much of clinical research, particularly in critical care, studies other outcomes. In infectious disease studies, we look at "microbiologic success" or "bacterial clearance." In hypertension studies, we look at blood pressure. In cholesterol studies, we look at cholesterol level. These outcomes do not measure variables that matter directly to patients. They don’t measure how patients feel, their quality of life, or how long they live. These variables are often selected because we think they predict—or are surrogates for—more important patient-centered outcomes. We measure blood pressure in a hypertension study because we think that lowering blood pressure will improve survival and prevent diseases like heart attacks and strokes that affect patients’ quality of life. We measure the effect of lipid-lowering drugs on cholesterol because we think that lowering cholesterol improves patient-centered outcomes, not because having a low LDL cholesterol is good in its own right.

Critical care research relies heavily on surrogate outcomes. We measure whether noninvasive ventilation has an effect on intubation because we believe that intubation causes complications that increase mortality and that being intubated is uncomfortable. We measure oxygen delivery because we think that treatments that improve oxygen delivery will reduce organ failure, which causes death. We measure sedation and agitation as scored by clinicians because we think it reflects patient comfort. We measure gas exchange because we think treatments that improve gas exchange will increase survival.

History has not been kind to these assumptions. Outside of critical care, surrogate outcome variables have received harsh criticism. In fact, we cannot assume that any surrogate outcome is reliably related to patient-centered outcomes. Our best held hypotheses about physiologic relationships, regardless of how well reasoned and sensible, are shown to be wrong—repeatedly. Two of the most common examples are the Cardiac Arrhythmia Suppression Trial (CAST) and inotropic therapy in congestive heart failure.

No hypothesis made more sense than the one that initiated the CAST study. Sudden death after myocardial infarction from cardiac arrhythmia is strongly associated with the presence of premature ventricular contractions and other ventricular dysrhythmias in the postmyocardial infarction period. Suppression of these dysrhythmias should prevent sudden death after myocardial infarction, which is presumed to be due to a cardiac rhythm disturbance. Effective drugs exist to suppress these dysrhythmias. However, in a large, randomized, clinical trial, drug therapy that was effective at suppressing the dysrhythmia was associated with increased mortality.1 Arrhythmia suppression is not a valid surrogate for mortality.

The pathophysiology of congestive heart failure is well known. Cardiac output is decreased from a failing left ventricle leading to the well-recognized clinical scenario of dyspnea, exercise intolerance, and pulmonary congestion. Drugs (eg, milrinone) can increase cardiac output. Unfortunately, while this drug does increase cardiac output and improves exercise tolerance, it also increases mortality.2 Cardiac output and exercise tolerance are not valid surrogates for mortality.

If either of these studies had not looked at mortality or had not been sufficiently statistically powered, we might have concluded that these treatments were effective at improving the surrogate outcome without knowing about their harmful effects on mortality.

Critical care is not immune to the perils of surrogate outcomes. Table 1 indicates a series of studies in which the surrogate outcome was shown to improve, but the treatment had no effect on patient-centered outcomes. Perhaps the clearest lesson of the last several years is the experience with ARDS. A variety of treatments—including inhaled nitric oxide, inhaled prostacyclin, liposomal prostaglandin E1, prone positioning during mechanical ventilation, partial liquid ventilation, and tracheal gas insufflation—have shown improvements in gas exchange in patients with ARDS. To date, none of these treatments has been shown to improve patient-centered outcomes, despite, in the cases of inhaled nitric oxide and prone positioning, large multicenter clinical trials.3-5 Contrary to the beneficial effects on gas exchange noted in these treatments, lung protective ventilation for ARDS, a treatment that uses low tidal volumes and allows carbon dioxide to "permissively" build up, generally and intentionally worsens the surrogate outcome of gas exchange in patients with ARDS.6 Despite its negative effect on patient physiology, mortality is significantly improved. Therefore, gas exchange does not appear to be a valid surrogate outcome for mortality in ARDS.

Even if surrogate measures are not valid predictors of patient-centered outcomes in studies of critical illness, variables other than death, cost, and quality of life are important in clinical research. Clinical research to understand mechanisms of critical illness requires a broad range of biochemical and physiologic as well as patient-centered variables. Phase II or hypothesis-testing studies will continue to use surrogate variables to identify promising treatments to study in larger studies. When a class of treatments has been shown to yield patient-centered benefits, surrogate outcome studies may be used cautiously to extend the results to other members of the same class without repeating patient-centered studies.

Studies of surrogate outcomes have repeatedly provided misleading information about patient-centered treatment effects in many areas of clinical investigation. Surrogate outcomes may have even worse performance in critical illness because the causal pathways in critical illness syndromes are so poorly understood. Current understanding of the body’s response to injury, infection, and hypoperfusion stresses the complexity of this response and the heterogeneity of the response depending on the age and comorbidity of the patient.7 If surrogate outcomes fail in single-organ diseases like cardiac dysrhythmia and cancer, it is difficult to imagine how they would perform better in the less well-characterized critical illness syndromes.

The appeal of surrogate outcomes, particularly in a physiologically oriented field like critical care, is understandable. Designing studies to address patient-centered outcomes requires larger, longer, and more expensive clinical trials than is the case with surrogate outcome studies. Nevertheless, ample evidence exists to make clinicians pause before adopting any therapy based on data showing improvements in surrogate outcomes alone.


Validity of Surrogate End Points—Examples from Critical Care

Disease Treatment Effect on Surrogate Effect on Patient
ARDS Prone ventilation5   Improved oxygenation No effect on mortality
ARDS Inhaled nitric oxide3,4 Improved oxygenation No effect on mortality
ICU anemia Blood transfusion8 Improved hematocrit Increased mortality
Critical illness Hemodynamic goal directed therapy9,10 Increased oxygen delivery No effect or increased mortality
Critical illness Human growth hormone11 Improved nitrogen balance Increased mortality, prolonged duration of  intensive care
Sepsis Ibuprofen12 Reduces levels of prostacyclin and thromboxane, decreases fever and lactic acidosis No effect on mortality
Sepsis  Recombinant human interleukin-1-receptor antagonist13,14     Improved survival time and reduced short- term mortality No effect on mortality


1. Echt DS, et al. Mortality and morbidity in patients receiving encainide, flecainide, or placebo. The Cardiac Arrhythmia Suppression Trial. N Engl J Med. 1991;324(12):781-788.

2. Packer M, et al. Effect of oral milrinone on mortality in severe chronic heart failure. The PROMISE Study Research Group. N Engl J Med. 1991;325(21): 1468-1475.

3. Dellinger RP, et al. Effects of inhaled nitric oxide in patients with acute respiratory distress syndrome: Results of a randomized phase II trial. Inhaled Nitric Oxide in ARDS Study Group. Crit Care Med. 1998;26(1):15-23.

4. Lundin S, et al. Inhalation of nitric oxide in acute lung injury: Results of a European multicentre study. The European Study Group of Inhaled Nitric Oxide. Intensive Care Med. 1999;25(9):911-919.

5. Gattinoni L, et al. Effect of prone positioning on the survival of patients with acute respiratory failure. N Engl J Med. 2001;345(8):568-573.

6. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. The Acute Respiratory Distress Syndrome Network. N Engl J Med. 2000;342(18):1301-1308.

7. Marshall JC. Complexity, chaos, and incomprehensibility: Parsing the biology of critical illness. Crit Care Med. 2000;28(7):2646-2648.

8. Hebert PC, et al. A multicenter, randomized, controlled clinical trial of transfusion requirements in critical care. Transfusion Requirements in Critical Care Investigators, Canadian Critical Care Trials Group. N Engl J Med. 1999;340(6):409-417.

9. Gattinoni L, et al. A trial of goal-oriented hemodynamic therapy in critically ill patients. SvO2 Collaborative Group. N Engl J Med. 1995;333(16):1025-1032.

10. Hayes MA, et al. Elevation of systemic oxygen delivery in the treatment of critically ill patients. N Engl J Med. 1994;330(24):1717-1722.

11. Takala J, et al. Increased mortality associated with growth hormone treatment in critically ill adults. N Engl J Med. 1999;341(11):785-792.

12. Bernard GR, et al. The effects of ibuprofen on the physiology and survival of patients with sepsis. The Ibuprofen in Sepsis Study Group. N Engl J Med. 1997;336(13):912-918.

13. Fisher CJ, Jr., et al. Recombinant human interleukin 1 receptor antagonist in the treatment of patients with sepsis syndrome. Results from a randomized, double-blind, placebo-controlled trial. Phase III rhIL-1ra Sepsis Syndrome Study Group. JAMA. 1994;271(23): 1836-1843.

14. Knaus WA, et al. Use of predicted risk of mortality to evaluate the efficacy of anticytokine therapy in sepsis. The rhIL-1ra Phase III Sepsis Syndrome Study Group. Crit Care Med. 1996;24(1):46-56.