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

Crystalloids vs. colloids in fluid resuscitation: a systematic review.

Choi PT, Yip G, Quinonez LG, Cook DJ.

Crit Care Med 1999; 27:200-10. [abstract]

Reviewed by Melchor Sanchez-Mendiola, Hospital Central Militar, Mexico City, Mexico

Review posted August 23, 1999

---

I. Are the results of the study valid?

A. Primary questions:

1. Did the overview address a focused clinical question?

Yes. The overview tried to answer the question: In adults requiring fluid resuscitation, does the infusion of isotonic crystalloid fluids vs. colloids improve mortality, length of stay or pulmonary edema ? It is important to emphasize that the terms "crystalloids" and "colloids" are very broad, and include a variety of different solutions, with different components, different concentrations, different molecular sizes and electrical charge, factors that affect the physico-chemical and pharmacological properties of the different fluids studied. The review did not include studies that used hypertonic solutions. determine the effect on mortality of resuscitation with colloid solutions compared with resuscitation with crystalloid solutions. Additional information on hospital stay and renal and pulmonary complications was sought in trials among surgical patients, although this was not the main focus of the overview.

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

Yes. Two reviewers independently screened all citations and classified them in primary studies, review articles or other. Afterwards all primary studies were reviewed in a blinded manner by 2 reviewers for inclusion, evaluating only the abstract and methods section, blinded to the journal, authors, year of publication, results and discussion. The criteria were the same as those used to structure the focused clinical question: population - adults requiring fluid resuscitation; intervention - isotonic crystalloid vs. colloid fluid infusion; outcomes - mortality, length of stay, pulmonary edema; study design - randomized clinical trials.

B. Secondary questions:

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

No. The Medline and CINAHL (Nursing and Allied Health) databases were searched from 1966 to November 1996 (more than two years before the publication date of the review, no mention is made of the specific software and version that was used to do the search). No mention is made of other probably relevant databases (Embase, Cochrane Controlled Trials Register, Lilacs, etc.), or if it was restricted to english language trials (all the references are in english, and that introduces country and language bias). The search strategy was: "resuscitation(textword or subject)", "fluid therapy (textword or subject)", and "comparative study or prospective studies or clinical trials", limiting the search to "human". Reference lists of the articles retrieved were reviewed to identify additional citations, however, no mention is made of "grey" literature (unpublished data, proceedings, etc.). In addition, there is a very interesting review related to the subject published electronically in June 16, 1999 (1), that identifies sixteen papers not included in this review that were cited in two very recent (2,3) systematic reviews. This reference explores the virtues and defects of the three most recent and important meta-analyses about fluid resuscitation.

4. Was the validity of the included studies appraised?

Yes. All selected studies were assessed for methodologic quality by two investigators independently based on methods (randomization, blinding), population (patient selection, population description), intervention (description, co-interventions), and outcomes (mortality and pulmonary edema). A scoring system was used (0,1 or 2 for each characteristic of the study) similar to the one described by Detsky et al (4). The methodologic quality of each is described, with scores ranging from 4 to 12 (maximum score 16). There were no double blind studies in all the papers evaluated, caregivers were unblinded in all the studies, and only 3 studies blinded the outcome assessors.

5. Were assessments of studies reproducible?

Yes. Agreement between reviewers on inclusion of studies and methodologic quality was measured by the kappa statistic with quadratic weights. The weighted kappa on agreement for inclusion was high (0.76), and disagreement was resolved by consensus. Agreement about the methodologic quality of the studies was moderate (kappa of 0.54).

6. Were the results similar from study to study?

No. In testing a priori hypotheses about heterogeneous results, they compared odds ratios between subgroups of studies. The Breslow-Day method was employed to test for differences in summary odds ratios between subgroups. To account for interstudy heterogeneity they used the random effects model, with the assumption that the analyzed studies were part of a larger group of studies comparing crystalloids and colloids in resuscitation; this model inherently assumes interstudy heterogeneity and results in wider confidence intervals and more conservative estimates of treatment effects than the fixed effects model.

The results were not similar among the different studies. For example, in the outcome of pulmonary edema, only six studies evaluated it, the definition was vague (clinical, chest x-ray or not described), and only one study defined it with more objective criteria such as PAOP > 25. The results for the effects of crystalloids vs. colloids on pulmonary edema were very different, with relative risk ranging from 0.25 to 4.69, and their 95% confidence intervals from 0.02 to 89.88. The variability in results in pulmonary edema is probably due to the different population studies (burns, sepsis, heart or vascular surgery), and the fact that there was variation in the definition of the outcome.

Regarding mortality (two studies did not report mortality figures), there was great variability, with relative risk ranging from 0.07 to 2.43, and the 95% CI from 0.00 to 90.08. The variability in mortality is due to the different populations studied (from trauma and burns, to sepsis, elective open-heart surgery and patients receiving IL-2 therapy), with probably very different primary diseases, and widely different prognostic factors (age, APACHE scores, nutritional status, comorbidities).

II. What are the results?

1. What are the overall results of the review?

Of 105 articles identified, 17 primary studies met the inclusion criteria and were included in the overview, with a total of 814 patients. Five studies evaluated trauma subjects, four studies had patients undergoing cardiac surgery, five studies patients with vascular surgery, three studies assessed other critically ill nontrauma patients. Risk ratios were used in the overview to avoid the magnification of the apparent size of treatment effects seen with odds ratios, mainly when baseline effects are infrequent. The overall relative risk for mortality of crystalloids vs. colloids in the 15 studies that reported it was 0.86 (95% CI 0.63-1.17), showing that the use of crystalloids or colloids did not affect mortality in the overall evaluation of the studies.

Subgroup analysis showed a statistically significant effect size on mortality favoring the use of crystalloids in the trauma subgroup ( RR 0.39, 95% CI 0.17-0.89). No other subgroup analyses demonstrated differences in mortality, and there was homogeneity of effect size across each subgroup. Only six studies measured pulmonary edema as outcome, and the overall results for these studies were RR 0.84 ( 95% CI 0.25-2.45 ). Length of stay or ICU stay were not different between groups (6 studies).

2. How precise were the results?

The results of the study do not demonstrate an important difference in the outcomes evaluated (mortality, pulmonary edema or length of stay) between the different groups treated with crystalloids vs. colloids for fluid resuscitation. The trend suggestive of less mortality in the group treated with crystalloids was not statistically significant; the confidence intervals for the relative risks cross unity. The suggestion in the subgroup analysis that the trauma patients had lower mortality when treated with crystalloids should be interpreted with caution, and viewed as data that should generate hypotheses for more studies. The studies don't have enough power to exclude clinically important differences in mortality or lung edema, because as the authors state, "to show a 10% difference in treatment effect, assuming 15% baseline mortality, an alpha of 0.05, beta of 0.20, a randomized trial with 5,748 patients would be required...". On the other hand, the duration, timing and specific nature of the fluid interventions were different between studies, co-interventions were frequently not described, and the patient populations were very different, making interpretation of the results difficult.

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

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

Probably not. The meta-analysis excluded studies that involved pediatric patients, and even though older children and adolescents may be physiologically similar to adults, there is no description in the paper of the age of the patients, and no information of how many elderly subjects were included in the studies. The information about the clinical characteristics of the patient populations is very scanty (e.g. "Patients with pulmonary edema...Ref.35), so there is no way to know how much our pediatric critically ill patients differ from the subjects included in the overview. The evidence does not support the conclusion that we should abandon the use of colloids for fluid resuscitation. Probably the use of specific types of fluids is important, but its influence is lost among the effects of all the other interventions, its sequence, the fact that several types of fluids are used in the same patient, illness severity, vasopressor support, nutrition, infections, etc.

2. Were all clinically important outcomes considered?

No. Mortality, LOS and pulmonary edema were assessed, but not other outcomes such as morbidity (specific for different age groups, multiple system organ failure), or several physiologic variables that were not summarized in the article (the authors refer us to the original papers), so there is no specific information about them.

3. Are the benefits worth the harms and costs?

No difference was convincingly shown between the treatments in regard to mortality, pulmonary edema and LOS, so there are no definite benefits. Costs were not reported, so there is no information to evaluate this aspect of medical care.

Additional comments:

There has been an increase recently in the debate about colloid vs. crystalloid fluids in resuscitation, triggered by the publication and instantaneous Internet availability of two other systematic reviews published in the British Medical Journal in 1998 (2,3), that have produced a fascinating response from physicians, researchers and the general public. An example of this is an electronic letter to BMJ in regard to the albumin review by Iain Chalmers, of the UK Cochrane Centre, saying "In brief, I would attempt to sue anyone who gave me an albumin infusion. And, as for any attempt to secure my informed consent to take part in a randomized trial (or my assent on behalf of someone I cared for who was unable to give informed consent) - forget it!". In the first review (2), an excess mortality of 4% was found for patients resuscitated with various colloids versus various crystalloids on several clinical situations; the second paper (3) concluded that there was an excess mortality of 6% in albumin-treated patients with various clinical conditions, versus non-albumin treated patients. Some of the studies in these reviews included pediatric patients. In the Choi et al. paper, there was no difference between crystalloid and colloid resuscitation with respect to mortality or pulmonary edema. There are some differences between the reviews that are clinically relevant; one of these is the inclusion of studies with hypertonic crystalloids pooled with isotonic crystalloids and compared with colloids in the Schierhout review (2). The patient populations stratification was different in the three studies as well, for in the Schierhout paper the groups were trauma, surgery, burns, and "other", while in the Cochrane review (3), they were hypovolemia, burns, or hypoalbuminemia. In Choi et al., the groups were burn trauma, penetrating trauma, critical care, cardiac surgery, and vascular surgery. The heterogeneity in the population groups and therapeutic interventions is perhaps the main obstacle for accepting the conclusions from these studies at face value, at least from the clinician point of view. It is very difficult for practising physicians, that manage these type of patients in an everyday basis, to be convinced and change their therapeutic behavior on the basis of a systematic review (a retrospective study by nature) that refutes their clinical wisdom, regardless of the statistical significance of the results.

References:

1. Webb AR. Crystalloid or colloid for resuscitation. Are we any the wiser? Crit Care 1999; 3:R25-R28.
2. Schierhout G, Roberts I. Fluid resuscitation with colloid or crystalloid solutions in critically ill patients: A systematic review of randomized trials. BMJ 1998; 316:961-4. [full-text] [link to PedsCCM EB Journal Club review]
3. Cochrane Injuries Group Albumin Reviewers. Human albumin administration in critically ill patients: A systematic review of randomized controlled trials. BMJ 1998; 317:235-40. [full-text] [link to PedsCCM EB Journal Club review]
4. Detsky AS, Naylor CD, O'Rourke K et al. Incorporating variations in the quality of individual randomized trials into meta-analysis. J Clin Epidemiol 1992; 45:255-65.

Comments

Back to the EB Journal Club Index