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

Human albumin administration in critically ill patients: systematic review of randomised controlled trials.

Cochrane Injuries Group Albumin Reviewers.

Br Med J 1998; 317:235-240. [full-text]

Reviewed by Al Torres, MD, MS, University of Illinois College of Medicine at Peoria

Review posted May 23, 1999

---

I. Are the results of the study valid?

A. Primary questions:

1. Did the overview address a focused clinical question?

No. The purpose of the study was "to quantify the effect on mortality of human albumin solution in the management of critically ill patients with hypovolaemia, from injury or surgery, burns, and hypoproteinaemia". The overview was not focused because albumin is not a drug but a fluid that is given according to a protocol for an effect. Its usage can vary by concentration, amount, and titration to effect. The authors pooled all kinds of albumin given in varying amounts for varying unspecified effects.

2. Were the criteria used to select articles for inclusion appropriate?

Unclear. The criteria reported were incomplete. The authors searched for "relevant" randomised controlled trials but do not define relevant. They do define a randomised controlled trial as a "trial in which the subjects followed were assigned prospectively to one of two (or more) interventions by random allocation or some quasi-random method of allocation". They did a subgroup analysis on those trials with "adequate" blinding and into the defined distinct subgroups (burns, hypovolemia, hypoproteinemia). The authors did not place many restrictions on the studies they included, nor did they attempt to standardize the length of follow-up, timing of the intervention or presence of significant comorbidity (e.g., MODS).

B. Secondary questions:

3. Is it unlikely that important, relevant studies were missed?

Yes. Besides computerized literature searches of a number of databases (Medline, Embase, Cochrane Controlled Trials Register, the Medical Editors Trial Amnesty Registry, and BIDS Index to Scientific and Technical Proceedings), the authors also hand searched 29 international journals and proceedings of several international meetings on fluid resuscitation. The authors also checked the reference lists of all the included trails and contacted the investigators of identified trials to ask them about any other published or unpublished trials that may have been conducted. There were no language restrictions. The authors claim they avoided publication bias by not detecting funnel plot asymmetry in the included trials (described below).

4. Was the validity of the included studies appraised?

No. The only validity criterion assessed was allocation concealment. The authors made no attempt to assess if the intervention (albumin) and the control groups were equivalent (e.g., age, comorbidities, severity of illness) or whether they were treated equally. Follow-up was not consistent across studies. The authors did not assess if the trials performed an intention-to-treat analysis.

5. Were assessments of studies reproducible?

Unclear. Two reviewers independently extracted data, and disagreements were resolved by consensus. It is not stated how much disagreement there was and agreement was not analyzed using a kappa statistic.

6. Were the results similar from study to study?

Yes. The authors performed chi squared analysis to assess for heterogeneity amongst the trial results. There was no heterogeneity (p > 0.2) when the trials were tested together or in the three defined groups.

The authors also used funnel plots - plots of the trials' effect estimates against sample size - to assess the validity of their meta-analysis (1). The funnel plot is based on the fact that precision in estimating the underlying treatment effect will increase as sample size of the included studies increase. Results from small studies will scatter widely at the bottom of the graph, with the spread narrowing among larger studies. In the absence of bias, the plot will resemble a symmetrical inverted funnel. However, funnel plots do have their limitations (2). Publication bias of positive trials (those favoring the intervention) would be detected by an asymmetric funnel, weighted at the bottom toward the right, representing that only small trials which favored the intervention were published. It is often difficult to visualize a funnel plot with only a few randomized controlled trials. Second, the unit of analysis is the randomized trial, not the patients, limiting the power of this method when only a few trials are included. And finally, since it is recommended that more attention be paid to the shape of the funnel plot rather than any statistical measure of asymmetry, this method should be viewed as non-specific and as a partially validated screening test for bias in meta-analysis (2).

II. What are the results?

1. What are the overall results of the review?

The authors used a fixed effects model to calculate the summary relative risk and 95% confidence intervals. It appears that studies were weighted by sample size, although this is not explicitly stated. The relative risk of dying if a critically ill patient received albumin was 1.68 for all trials. The relative risk of dying if a hypovolaemic patient received albumin was 1.46; burn patient, 2.4; hypoalbuminaemic patient, 1.69. The 13 trials with adequate blinding found similar results: overall, 1.61; hypovolaemic 1.39, burns, 2.47; hypoalbuminaemic, 1.71. The pooled difference in risk of death was 6% if you received albumin versus no albumin or crystalloid solutions only.

2. How precise were the results?

Fair. The relative risk 95% confidence intervals included one for the hypovolaemia trials (0.97 to 2.22) and the lower value was close to one for the burns (1.11 to 5.19), hypoalbuminaemia (1.07 to 2.67), and for all the trials combined (1.26 to 2.23). Recall that the 95% confidence intervals represent the range of relative risks of dying for 95 out of 100 trials with identical design but different sample sizes. If the confidence interval of a relative risk includes one, the risk may be equivalent for the treatment and the control groups, making the relative risk of the trial being appraised insignificant. The relative risk 95% confidence intervals were wide for the burns trials but this was not surprising since this subgroup only included three trials and had the smallest combined sample size of the three subgroups.

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

1. Can the results be applied to my patient care?

No. None of the studies reviewed included patients admitted to a PICU. Details regarding other interventions, severity of illness, and timing of the intervention were not presented making the results less convincing. Also, although the intervention was defined as human albumin, the administration forms used varied considerably (e.g., 25% albumin infusion, albumin in TPN). Albumin is not a single dose drug but is a fluid that is titrated to effect. The goal of therapy was not the same across trials so it is unclear how to apply the results.

2. Were all clinically important outcomes considered?

No. The authors point out mortality as the least likely to be biased and was, therefore, the only outcome evaluated. The differences in co-morbid conditions causing the need for albumin and the variation in follow-up between studies (e.g., 1 day to hospital discharge) make even mortality a suspect outcome.

3. Are the benefits worth the harms and costs?

Unknown. Another recently published systematic review of colloid vs. crystalloid studies found an absolute increase in mortality associated with colloid administration in critically ill patients of 4% (95% CI, 0 to 8%) (3). With human albumin being "over 30 times more expensive than crystalloid solutions" and no evidence that human albumin helps critically ill patients survive any better than crystalloids, the benefits may not be worth the risks. However, use of a specific concentration of albumin titrated to achieve an effect may improve patient outcome. This could not be deciphered from this overview.

References:

1. Egger M, Smith GD, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:629-634. [full-text]
2. Naylor CD. Meta-analysis and the meta-epidemiology of clinical research. Meta-analysis is an important contribution to research and practice but it's not a panacea (editorial). BMJ 1997;315:617-619. [full-text]
3. Schierhout G, Roberts I. Fluid resuscitation with colloid or crystalloid solutions in critically ill patients: a systematic review of randomised trials. BMJ 1998;316:961-965. [full-text]

Related Reviews:

Schierhout G, Roberts I. Fluid resuscitation with colloid or crystalloid solutions in critically ill patients: a systematic review of randomised trials. Br Med J 1998; 316: 961-5. (full-text)

Review, by M. Verive

Choi PT, Yip G, Quinonez LG, Cook DJ. Crystalloids vs. colloids in fluid resuscitation: a systematic review. Crit Care Med 1999; 27:200-10. [abstract]

Review, by M. Sanchez-Mendiola

Comments

Back to the EB Journal Club Index