Keep 'em cool!

A new feature of VicCritCare is the assessment of some of the biggest papers in the critical care world - past and present.  This week's guest reviewer is Dr James Malycha.

James is an Intensive Care Registrar from Melbourne, Australia.  He has a love of obscure humour.

Treatment of comatose survivors of out of hospital cardiac arrest

Stephen A. Bernard N Engl J Med, Vol. 346, No. 8 · February 21, 2002

Summary

A well designed randomised, not blinded study showing clear improvement in neurological outcomes in patients who have out of hospital VF arrests who are promptly cooled (within 2 hours) to 33 degrees and treated in well resourced tertiary hospital, although no improvement in mortality found.

Background

Poor outcomes for OOHCA

5-35% survival

Survivors have poor outcomes due to anoxic brain injury

Study question: Does cooling help, as suggested by animal studies?

Hypothesis: Cerebral ischaemia may persist for hours post-resus. Hypothermia may help reduce damage caused by this phenomenon.

Existing data: Only other human trials  are retrospective and uncontrolled.

Methods

Design

Inclusion – VF, ROC at scene, 4 ED’s

Exclusion - < 18yo male, < 50yo female, SBP < 90 despite resus, causes other than cardiac, ROSC had to occur at scene

Primary outcome – survival to hospital discharge good enough to go home or rehab

Secondary outcome – heamodynamic, biochemical and heamotological effects of hypothermia

77 patients in total – 43 hypo, 34 normothermic

Controlled for Pa02, K, creatinine, CK, MAP.

Protocol 

Both groups given antiarrythmic (lignocaine bolus then infusion)

PACatheter. Some not (roughly 20% in each group)

Sedated and paralysed (all in hypo, as needed in normo) – midaz and vec.

Outcome 

In ICU – withdrawal, trachy or extubated depending on progress

On wards -Rehab physician assessed needs on discharge from hositpal. Blinded to initial treatment.

Stats

Well powered. Achieved significance.

Aimed to achieve 15 – 50% improvement in primary outcome b/w groups.

Results

84 pt over 33 months

Characteristics of groups similar

More men in normothermic group?? 79 vs 58 %

No difference in mortality

Difference in ‘positive outcome’ 

23% improvement in normal or minimal disability group – stat significant

Also shown with odds ratios. Both p values 0.046

Weaknesses

No women < 50yo.

More men in normothermic group.

Not blinded at scene or in ED/ICU (?impossible).

Randomisation using odd/even days is not ideal.

Application to ICU

Strong applicability to out of hospital cardiac arrest who gain ROSC in the field and are promptly cooled by emergency care givers.

Steroids for refractory septic shock

Can we use steroids to dampen the immune system in order to modify the “cytokine storm” associated with septic shock?  This is a plausible concept that has been vigorously studied over many years.  

The syndrome of Relative Adrenal Insufficiency (RAI) is thought to contribute to mortality, and therefore the question arises: if we replace endogenous cortisol with hydrocortisone, can we reduce mortality? 

The CORTICUS Trial (NEJM 2008: 358(2); 111-124)

Are results valid?

- multicenter, randomized, double blind, placebo controlled

- Hydrocortisone 50mg QID for 5 days, 50 BD for day 6 - 8, then 50 daily for days 9 to 11

- 1 year follow up

- primary end-point rate of death at 28 days

- power calculation: 400 per group

- ITT analysis

- similar patient groups

Results?

- 251 patients received hydrocortisone, 248pts placebo

- baseline difference in death at 28 days in non-responders 38% versus responders 29% to SynACTHen test

- no difference for hydrocortisone v placebo overall in death at 28days

- wide confidence intervals reflecting limited sample size, not powered to find treatment effect

- In those who got Etomidate, 60.4% did not respond to SynACTHen, while in those who did not receive Etomidate only 43.4% (p=0.004) did not respond, and there was a significantly higher death rate

- duration of time until reversal of shock was shorter among all patients who got hydrocortisone (p<0.001), however total number of patients with reversal of shock unaffected

- increased risk of new episode of sepsis or septic shock with OR = 1.37 in those who received hydrocortisone

Can I apply the results to my patient group?

- Yes, similar population, similar baseline mortality

- treatment is feasible, but benefits of hydrocortisone do not clearly outweigh the harms

Critical Appraisal Guide for Studies of Therapy

This is a basic tool you can use to assess a study of therapy, and is based on the CONSORT statement


Are the results of the study valid?

• Was the assignment of patients to treatments randomised? Was the randomisation concealed?
• Were the groups similar at baseline?
• Was follow-up of patients sufficiently long and complete?
• Were all patients analysed in the groups to which they were randomised?
• Were patients and clinicians blinded?
• Were the groups treated equally, apart from the intervention

What are the results?

• Was the study powered to find a difference between the two groups?
• How large was the treatment effect?
• How precise was the estimate of treatment effect?

How Can I apply the results to my patient care?

• Can the results be applied to my patient population?
• Is the treatment feasible in my setting?
• Were all clinically important outcomes considered?
• Are the likely treatment benefits worth the potential harms and costs?
• Should this study change my current practice?

For more information on critical appraisal, see www.consort-statement.org

Introduction to Statistics

TYPES OF DATA

• Categorical (qualitative)
• Nominal (eg. type of car)
• Ordinal (eg. Stage of cancer - rank order)
• Numerical (quantitative)
• Discrete (eg. number of children)
• Numbers are real and can subtract/divide at will
• Continuous (eg. cholesterol level)

Choice of statistical method to summarize data depends on type of data

Mean - arithmetic average of the observations (used for numerical data only)

Median - it is the middle observation in a data-set (can be used for rank ordered data, less sensitive to extremes than is the mean)

Mode - the most common value in a data-set (used only when the number of possible responses is small)

Geometric mean - used when data are heavily right skewed

Relationship between mean, median, and mode depends on the shape of the distribution of data.  In general, you should use the Mean for symmetric numerical data, and the Median for skewed numerical or ordinal data

Measures of spread

• lower (25%) and upper (75%) quartiles

Standard deviation

• Spread of data around the mean
• useful for symmetric data
• how far each observation is from the mean
• calculate deviation from mean
• then calculate variance
• then calculate standard deviation

Standard deviation is important because, when data follows a normal distribution (bell-shaped curve)

• 68% of data fall between mean -1 SD and mean +1 SD
• 95% of data fall between mean -2 SD and mean +2 SD
• 99.7% of data fall between mean -3 SD and mean +3 SD