Exclusion of suspect data raises question mark over safety of common plasma substituteBMJ 2013; 346 doi: https://doi.org/10.1136/bmj.f1132 (Published 19 February 2013) Cite this as: BMJ 2013;346:f1132
Hydroxyethyl starch, a commonly used plasma substitute, is associated with increased risk of death and acute kidney injury in critically ill patients needing an increase in blood fluid volume, according to a new meta-analysis1 that excluded trials from an investigator some of whose research has been retracted because of scientific misconduct.
“This study highlights the serious implications of scientific misconduct on patient safety,” warned Massimo Antonelli, professor of intensive care medicine at the Università Cattolica del Sacro Cuore, Rome, Italy, in an accompanying editorial.2 He added that it illustrated the importance of updating guidelines regularly as the evidence base changes.
Synthetic colloids such as hydroxyethyl starch were approved in the 1960s without evaluation of their efficacy or safety in large phase III trials. Subsequent studies have generated conflicting results. The controversy increased when studies by Joachim Boldt, who worked at Klinikum Ludwigshafen, Germany, were retracted after an inquiry by the state medical association found data fabrication and scientific misconduct in his research.3 Despite this, hydroxyethyl starch is still commonly used for volume resuscitation.
The meta-analysis, published in JAMA, included 38 trials comparing hydroxyl starch with crystalloids, albumin, or gelatin in more than 10 000 critically ill patients receiving acute volume resuscitation. Researchers analysed results with and without seven studies published by Boldt and colleagues, which have not been retracted because they were published before 1999 and misconduct investigations were limited to more recent trials.
Results including Boldt’s studies showed no reduction in mortality with hydroxyethyl starch compared with other resuscitation solutions (risk ratio 1.07, 95% confidence interval 1.00 to 1.14; absolute risk 1.2%, −0.26% to 2.66%).
Analysis excluding the seven Boldt trials, which involved 590 patients, showed that hydroxyethyl starch was associated with increased mortality among 10 290 patients (risk ratio 1.09, 95% confidence interval 1.02 to 1.17; absolute risk 1.51%, 0.02% to 3.0%). Hydroxyethyl starch was also associated with increased renal failure among 8725 patients (1.27, 1.09 to 1.47; 5.45%, 0.44% to 10.47%).
“Our study should prompt physicians to reflect on their use of hydroxyethyl starch in all clinical conditions. In the absence of evidence to suggest hydroxyethyl starch is effective and safe, it should not be routinely prescribed for acute volume resuscitation,” concluded lead author Ryan Zarychanski, assistant professor and haematologist and critical care physician at University of Manitoba and CancerCare Manitoba, Winnipeg, Canada.
Zarychanski added that hospitals should consider the appropriateness of unrestricted use of hydroxyethyl starch on their formularies and clinical guidelines, and that previous meta-analyses should be updated.
Antonelli agreed, “Thoughtfully planned, adequately powered and rigorously conducted trials are needed to assess the safety of hydroxyethyl starch 130.04 [the most recent formulation] versus crystalloids. Until then, this study indicates that the harm can outweigh the benefit and recommendations should be revised accordingly.” He added: “The design of the meta-analysis is clever and addresses a very important issue: how the inclusion or exclusion of papers associated with research misconduct can shift the balance toward harm.”
Zarychanski said, “The story of hydroxyethyl starch is larger than academic misconduct and is not unique . . . It is not good enough for these products to ‘do that which they were designed to do.’ These innovations should impact clinical outcomes that matter to patients, and be safe.”
Cite this as: BMJ 2013;346:f1132
Research: Hydroxyethyl starch 130/0.38-0.45 versus crystalloid or albumin in patients with sepsis: systematic review with meta-analysis and trial sequential analysis, doi:10.1136/bmj.f839.