Intended for healthcare professionals

Letters

Fluid resuscitation with colloid or crystalloid solutions

BMJ 1998; 317 doi: https://doi.org/10.1136/bmj.317.7153.277 (Published 25 July 1998) Cite this as: BMJ 1998;317:277

Comparing different studies is difficult

  1. J Watts, Consultant anaesthetist
  1. Burnley General Hospital, Burnley BB10 2PQ
  2. Department of Clinical Biochemistry, University Hospital Birmingham NHS Trust, Birmingham B29 6JD
  3. Western Infirmary, Glasgow G11 6NT
  4. Department of Anaesthesia, St George's Hospital, London SW17 0QT
  5. Department of Anaesthesia and Intensive Care, West Middlesex University Hospital, Isleworth, Middlesex TW7 6AF
  6. Royal Brompton Hospital, London SW3 6NP
  7. Guy's Hospital, London SE1 9RT
  8. Department of Epidemiology and Public Health, Institute of Child Health, University College London Medical School, London WC1N 1EH
  9. Hereford Hospitals NHS Trust, County Hospital, Hereford HR1 2ER

    Editorial by Offringa and Paper p 235

    EDITOR—The debate over giving crystalloids or colloids has been raging since the 19th century, when Cohnheim and Lichtheim found gastric mucosal oedema in patients who had been resuscitated with saline and Starling suggested that albumin could prevent oedema. 12The meta-analysis by Schierhout and Roberts, which does not support the continued use of colloids for volume replacement in critically ill patients, makes a useful contribution to this debate but does not settle it.3

    A recent review by Hankeln and Beez comes to the opposite conclusion—that colloids are more effective than crystalloids for optimising physiological variables related to flow in critically ill patients and maintaining the delivery of oxygen to the tissues2; they say that this is related to the persistence of colloids in the circulating plasma volume, as opposed to their distribution throughout the total body water.4Although colloids are more expensive than crystalloids, their effect on the circulating volume lasts much longer. The real problem is the difficulty in comparing different studies, because of differences in case mix, resuscitation protocols, and volumes and types of fluids used and, therefore, in making firm conclusions about patient outcome.

    In all cases of hypovolaemia the main priority is to restore the circulatory volume as quickly and efficiently as possible to prevent impairment of organs due to ischaemia and hypoxaemia.5 Maybe we will never have a definitive answer to this question, in which case many practitioners will continue to administer a judicious mix of both types of fluid according to their own experience.

    References

    Newer synthetic colloids should not be abandoned

    1. Peter Gosling, Consultant clinical biochemist.
    1. Burnley General Hospital, Burnley BB10 2PQ
    2. Department of Clinical Biochemistry, University Hospital Birmingham NHS Trust, Birmingham B29 6JD
    3. Western Infirmary, Glasgow G11 6NT
    4. Department of Anaesthesia, St George's Hospital, London SW17 0QT
    5. Department of Anaesthesia and Intensive Care, West Middlesex University Hospital, Isleworth, Middlesex TW7 6AF
    6. Royal Brompton Hospital, London SW3 6NP
    7. Guy's Hospital, London SE1 9RT
    8. Department of Epidemiology and Public Health, Institute of Child Health, University College London Medical School, London WC1N 1EH
    9. Hereford Hospitals NHS Trust, County Hospital, Hereford HR1 2ER

      EDITOR—In their meta-analysis of trials that compared colloids with crystalloids in critically ill patients Schierhout and Roberts found increased mortality in patients treated with colloids and concluded: “this systematic review does not support the continued use of colloids for volume replacement in critically ill patients.”1

      Their finding of increased mortality in patients treated with colloid is not surprising, since of the 26 studies reviewed, 12 used albumin solutions, eight low molecular weight dextrans, and three gelatin solutions. It is now well recognised that critically ill patients have increased capillary permeability, which allows molecules such as albumin and water to pass into the interstitial space, thus compromising tissue oxygenation.2 For example, every 1 g of albumin that leaks from capillary to interstitium is accompanied by 18 g of water. This condition is termed clinical capillary leak syndrome and represents a stage through which all patients pass in the development of systemic inflammatory response syndrome and organ failures. For this reason, volume replacement treatments based on low to medium molecular weight colloids (for example, gelatins and albumin) which do not remain in the vascular space will be ineffective and may worsen interstitial oedema.

      The “colloids” in the studies reviewed ranged in molecular weight from 30 000 Da to over 1£106 Da, and their ability to remain in the vascular compartment, particularly during periods of capillary leak, varies.3Gelatins, for example, are lost from the circulation within an hour or two, while some hydroxyethyl starches with a large molecular weight are retained for 48 hours. Comparative studies of albumin with hydroxyethyl starches with a molecular weight of around 250 000 show the hydroxyethyl starches to be beneficial in terms of haemodynamic variables, oxygen delivery, and non-specific inhibition of the acute inflammatory response. 45

      The authors are correct in their recommendation for randomised controlled trials of colloids in patients at risk of clinical capillary leak syndrome and systemic inflammatory response syndrome. But it would be a pity if their conclusions, which are based on an uncritical amalgamation of studies of ineffective colloids, lead to the abandonment of the use of newer synthetic colloids which are well retained by the leaky vasculature in critically ill patients. Such an interpretation would encourage excessive administration of salt and water, leading to fluid overload, interstitial oedema, poor oxygenation, and organ failure.

      References

      One conclusion could be that hypertonic saline is better than colloids in trauma

      1. Andrew Makin, Specialist registrar in anaesthetics,
      2. Louie Plenderleith, Consultant anaesthetist.
      1. Burnley General Hospital, Burnley BB10 2PQ
      2. Department of Clinical Biochemistry, University Hospital Birmingham NHS Trust, Birmingham B29 6JD
      3. Western Infirmary, Glasgow G11 6NT
      4. Department of Anaesthesia, St George's Hospital, London SW17 0QT
      5. Department of Anaesthesia and Intensive Care, West Middlesex University Hospital, Isleworth, Middlesex TW7 6AF
      6. Royal Brompton Hospital, London SW3 6NP
      7. Guy's Hospital, London SE1 9RT
      8. Department of Epidemiology and Public Health, Institute of Child Health, University College London Medical School, London WC1N 1EH
      9. Hereford Hospitals NHS Trust, County Hospital, Hereford HR1 2ER

        EDITOR—Schierhout and Roberts's meta-analysis of fluid resuscitation with colloid or crystalloid solutions in critically ill patients is inaccurate.1They quote four trials with adequate concealment and say that there was an increase in the absolute risk of death of 4% with colloids.

        Their summary table lists four trials with a concealment of 3. We assume that these are the four trials in question. Three of these trials compared colloid with hypertonic saline and account for 39 deaths of the total difference of 50 deaths (78%) between colloids and crystalloid. We also note that these three papers are from trauma studies. It thus may be more accurate to conclude that hypertonic saline (7.5%) is better in trauma than colloids.

        References

        Eight studies should have been excluded

        1. G R McAnulty, Locum consultant in anaesthesia and intensive care,
        2. R M Grounds, Consultant in anaesthesia and intensive care
        1. Burnley General Hospital, Burnley BB10 2PQ
        2. Department of Clinical Biochemistry, University Hospital Birmingham NHS Trust, Birmingham B29 6JD
        3. Western Infirmary, Glasgow G11 6NT
        4. Department of Anaesthesia, St George's Hospital, London SW17 0QT
        5. Department of Anaesthesia and Intensive Care, West Middlesex University Hospital, Isleworth, Middlesex TW7 6AF
        6. Royal Brompton Hospital, London SW3 6NP
        7. Guy's Hospital, London SE1 9RT
        8. Department of Epidemiology and Public Health, Institute of Child Health, University College London Medical School, London WC1N 1EH
        9. Hereford Hospitals NHS Trust, County Hospital, Hereford HR1 2ER

          EDITOR—In their systematic review Schierhout and Roberts make several assertions that cannot be justified by their evidence, and their conclusions are unwarranted.1

          To draw valid conclusions from data reported in a variety of studies, like must be compared with like. Studies that include uncommon or experimental treatments should not be used to make inferences about conventional practice. Of the 19 trials that met the criteria for eligibility for analysis, five involved the use of hypertonic solutions. In these five trials mortality was 36% (90/250) among patients who received colloids and 25% (51/201) among patients who received crystalloids. Among the remaining studies (in which the resuscitative fluids more closely reflected those in clinical use) mortality was 18% (78/435) for colloids and 16% (67/429) for crystalloids (not significantly different). The study by Goodwin et al in patients with burns is the only trial in which any trend towards increased mortality in patients receiving colloid was reported.2Significant increases in lung water were found in the colloid group, but pulmonary capillary wedge pressure, although measured, was not used to guide resuscitation.

          Only five studies compared commonly used resuscitation fluids (Metildi et al used 5% albumin solution,3not 50% salt-poor albumin as reported by Schierhout and Roberts). Of these studies, three were of surgical patients; mortality was low but favoured the use of colloids as a resuscitative fluid (3% mortality (l/33) for colloids v 7% (3/42) for crystalloids). The remaining two studies were of seriously ill patients receiving intensive care, who are pathophysiologically distinct. Despite this, these studies showed opposite outcomes. Rackow et al found a mortality of 61% (11/18) for colloids and 75% (6/8) for crystalloids.4Metildi et al found a mortality of 60% (12/20) for colloids and 46% (12/26) for crystalloids.3 In Metildi et al's study, however, the proportion of patients without sepsis was higher in the crystalloid group; 20% (4/20) in the colloid group died within 48 hours of entry, compared with 12% (3/26) in the crystalloid group. Patients in critical care units have a wide range of conditions, and studies of mortality in small numbers of patients are misleading if they do not include data allowing comparison of different treatment groups in terms of severity of illness

          Of the 19 trials quoted, at least eight were in unusual groups or of unusual treatments. These eight studies should not have been included in the analysis.

          References

          Use of dextran-70 for fluid resuscitation has been dying out

          1. Pramod P Bapat, Locum consultant.,
          2. Geoff J Raine, Senior registrar.
          1. Burnley General Hospital, Burnley BB10 2PQ
          2. Department of Clinical Biochemistry, University Hospital Birmingham NHS Trust, Birmingham B29 6JD
          3. Western Infirmary, Glasgow G11 6NT
          4. Department of Anaesthesia, St George's Hospital, London SW17 0QT
          5. Department of Anaesthesia and Intensive Care, West Middlesex University Hospital, Isleworth, Middlesex TW7 6AF
          6. Royal Brompton Hospital, London SW3 6NP
          7. Guy's Hospital, London SE1 9RT
          8. Department of Epidemiology and Public Health, Institute of Child Health, University College London Medical School, London WC1N 1EH
          9. Hereford Hospitals NHS Trust, County Hospital, Hereford HR1 2ER

            EDITOR---Schierhout and Roberts have reviewed studies of different colloids compared with crystalloids in critically ill patients.1It is inappropriate to group all the colloids in one category. Although the authors accept that colloids and crystalloids have different effects on a range of important physiological variables, they presume that individual differences between different colloids are of no clinical importance.

            Human albumin, plasma, starch, dextran, and modified gelatin have differing pharmacokinetics and pharmacodynamics.2 Dextran-70 affects coagulation by a diminution of platelet adhesiveness, depression of factor VIII activity, and increased fibrinolysis. 23In the studies quoted by Schierhout and Roberts, use of dextran-70 is clearly associated with a higher mortality. In the studies specifically comparing dextran-70 with crystalloids, 96 of 351 patients who received dextran-70 died, compared with 56 of 301 who received crystalloids. The difference is clinically apparent and significant when analysed with Student's t test (P<0.01). On the other hand, in the studies that compared other colloids with crystalloids, 72 of the 304 who received colloids died, compared with 61 of the 319 who received crystalloids. This difference is not significant. We did not include the study by Tølløsfrud et al in the above analysis because they included dextran-70 and other colloids in one group, and compared them with crystalloids.4

            Use of dextran-70 for fluid resuscitation has gradually been dying out and largely replaced by use of other fluids. In our clinical practice we use a mix of crystalloids, starch, or gelatin solutions followed by packed cells and other blood products if necessary. We believe that our practice reflects the current British use of fluids during resuscitation.

            Perhaps the only interpretation one can draw from this review is that use of dextran-70 for fluid resuscitation in critically ill patients is associated with a higher mortality than use of other colloids or crystalloids. Future studies should focus not only on individual fluids but also on the timing of resuscitation and targets set for cardiovascular variables during resuscitation.

            References

            Conditions and patient groups were too heterogeneous to allow meaningful comparisons

            1. D L A Wyncoll, Specialist registrar in intensive care,
            2. R J Beale, Consultant intensivist,
            3. A McLuckie, Consultant intensivist.
            1. Burnley General Hospital, Burnley BB10 2PQ
            2. Department of Clinical Biochemistry, University Hospital Birmingham NHS Trust, Birmingham B29 6JD
            3. Western Infirmary, Glasgow G11 6NT
            4. Department of Anaesthesia, St George's Hospital, London SW17 0QT
            5. Department of Anaesthesia and Intensive Care, West Middlesex University Hospital, Isleworth, Middlesex TW7 6AF
            6. Royal Brompton Hospital, London SW3 6NP
            7. Guy's Hospital, London SE1 9RT
            8. Department of Epidemiology and Public Health, Institute of Child Health, University College London Medical School, London WC1N 1EH
            9. Hereford Hospitals NHS Trust, County Hospital, Hereford HR1 2ER

              EDITOR—In their systematic review about fluid resuscitation with crystalloids or colloids in critically ill patients, Schierhout and Roberts conclude that the use of colloids is associated with a 4% absolute increase in mortality and that, as colloids are also considerably more expensive than crystalloids, they should no longer be used outside randomised controlled trials.1This conclusion is flawed and is not supported by their analysis.

              Firstly, the definitions of crystalloid and colloid are far too loose, and invalid assumptions about the equivalence of different fluid preparations have been made. Included in the crystalloid treatment arm are hypertonic saline solutions (four studies); isotonic, hypotonic, or unspecified solutions (21 studies); and Ringer's solution with 3.5% gelatin (one study). Included in the colloid arm are albumin solutions (12 studies), dextran solutions (nine studies), gelatin solutions (three studies), starch solutions (five studies), and whole blood (one study); many studies used mixtures of colloids. These groups are far too heterogeneous to allow a meaningful comparison.

              The patient populations are also widely disparate, and the authors seem to have ignored the influence that different strategies for giving fluid may have had. A recent postal survey of British intensive care units showed that dextran solutions are almost never used,2and a German survey reported that <1% of units ever used dextran solutions and that <2% used albumin solutions.3This suggests that the practice described in the studies in this analysis is not relevant to practice in Europe today.

              Nearly all the studies included in the mortality analysis had low mortality and wide confidence intervals. Not surprisingly, the cumulative analysis also has a confidence interval spanning unity (0.98 to 1.45). Despite this, the authors are sufficiently confident to conclude that colloids should no longer be used routinely rather than accepting that the true result of their analysis is uncertainty. This paper exhibits all the worst abuses of evidence based medicine. It is written without clinical insight into the subject being considered, and the results have been used to produce a sweeping generalisation, which is scientifically inept. We are concerned that such a flawed undertaking should have been supported by the Intensive Care National Audit and Research Centre and paid for with public funds from the NHS R&D Programme.

              References

              Authors' reply

              1. G H Schierhout (gschierh{at}iafrica.com), Lecturer in epidemiology,
              2. Ian Roberts, Director, child health monitoring unit
              1. Burnley General Hospital, Burnley BB10 2PQ
              2. Department of Clinical Biochemistry, University Hospital Birmingham NHS Trust, Birmingham B29 6JD
              3. Western Infirmary, Glasgow G11 6NT
              4. Department of Anaesthesia, St George's Hospital, London SW17 0QT
              5. Department of Anaesthesia and Intensive Care, West Middlesex University Hospital, Isleworth, Middlesex TW7 6AF
              6. Royal Brompton Hospital, London SW3 6NP
              7. Guy's Hospital, London SE1 9RT
              8. Department of Epidemiology and Public Health, Institute of Child Health, University College London Medical School, London WC1N 1EH
              9. Hereford Hospitals NHS Trust, County Hospital, Hereford HR1 2ER

                EDITOR—Several of the correspondents interpreted our data in different ways, and in some cases this involved a reanalysis of the data. We believe that this is a major strength of systematic reviews. Because the reviews make explicit what the data are, including a full description of the participants in the trials, the interventions, and the outcomes, clinicians are ideally placed to make their own judgments about what they consider that the data mean.

                As Watts notes, the colloid versus crystalloid debate has raged for nearly a century. The fact that the American College of Surgeons recommends the use of crystalloid in the management of trauma in preference to colloid in its recent guidance on advanced trauma life support implies that this is still the level of debate.1Because we made the nature of the colloid explicit in our review the debate may now progress to the effectiveness of specific colloids, and reviews of this nature are currently being prepared by the Cochrane Injuries Group.

                We note with concern the mechanisms outlined by Gosling—in particular, the propensity of albumin to pass into the interstitial space, where it may compromise tissue oxygenation. Albumin is widely used, although the evidential basis for its use has yet to be defined.

                Although it was stated in the methods section and obvious from the table, several correspondents (Makin et al and McAnulty et al) failed to understand that trials were included only if the treatment (colloid) group differed from the comparison group only in the treatment of interest. Trials comparing colloid in hypertonic crystalloid with hypertonic crystalloid alone are unconfounded trials of the addition of colloid. Any effect of the hypertonic crystalloid would be produced in both the treatment and comparison arms, and any difference in mortality would thus be attributed to the colloid or to the play of chance.

                Wyncoll et al claim that the trials included in our review are not relevant to today's practice. The lack of contemporary trials is not a weakness of our review but a weakness of the evidential basis of intensive care medicine. If relevant studies had been conducted we included them. What our review has shown is the dearth of properly controlled evaluations on which to base current intensive care. We agree with Wyncoll et al that our analysis shows uncertainty about the use of colloids, and, given this uncertainty, we reassert our conclusion that further trials assessing the effectiveness of colloids (including the newer synthetic colloids) are urgently required. Not to act on this uncertainty risks the long term abuse of public funds through the daily use of more costly and possibly more harmful resuscitation regimens.

                Finally, we wish to clarify the contribution of the Intensive Care National Audit and Research Centre to our review. Because of the centre's handsearching activities, a large number of previously inaccessible randomised trials are now available to those interested in the evidential basis of intensive care medicine. This is an important contribution and deserves to be acknowledged. However, the centre is no more responsible for the conclusions of our review than is the National Library of Medicine for providing access to randomised trials on Medline.

                References

                Virtually identical article had appeared in Cochrane Library

                1. A P Corder, Consultant surgeon
                1. Burnley General Hospital, Burnley BB10 2PQ
                2. Department of Clinical Biochemistry, University Hospital Birmingham NHS Trust, Birmingham B29 6JD
                3. Western Infirmary, Glasgow G11 6NT
                4. Department of Anaesthesia, St George's Hospital, London SW17 0QT
                5. Department of Anaesthesia and Intensive Care, West Middlesex University Hospital, Isleworth, Middlesex TW7 6AF
                6. Royal Brompton Hospital, London SW3 6NP
                7. Guy's Hospital, London SE1 9RT
                8. Department of Epidemiology and Public Health, Institute of Child Health, University College London Medical School, London WC1N 1EH
                9. Hereford Hospitals NHS Trust, County Hospital, Hereford HR1 2ER

                  EDITOR—I read the meta-analysis by Schierhout and Roberts1with a sense of déjà vu as a virtually identical article had appeared in the Cochrane Library.2Acknowledgment of the identical nature of the two articles would have saved me reading what I had already read and would, no doubt, prevent unnecessary requests for reprints by other parties. I am disappointed that the article should appear in the BMJ without obvious and specific acknowledgment that it had already been published in the Cochrane Library.

                  We apologise that we didn't make it clear that the systematic review by Schierhout and Roberts had already appeared in the Cochrane Library. We have made it clear that the systematic review that we are publishing today on the use of human albumin in critically ill patients is also about to appear in the Cochrane Library.

                  Duplicate publication is always a problem when it is not declared, but there are many circumstances in which duplicate publication will help with the dissemination of important material. We have agreed with the Cochrane Collaboration that we are willing to publish in the BMJ reviews that are appearing in the Cochrane Library. There are several reasons for this policy. Firstly, many readers of the BMJ do not have easy access to the Cochrane Library: the BMJ gives the reviews a much higher profile. Secondly, many authors of systematic reviews find very important the value added by the BMJ's peer review process, technical editing, and design. Thirdly, many people do not find the Cochrane Library easy to use. Fourthly, authors of reviews receive greater academic credit through publication in the BMJ. All of these reasons are likely to change, and we need to keep this policy under constant review.—EDITOR.

                  References

                  View Abstract