Clinical Trials in Volume Resuscitation with Hydroxyethyl Starch: Focus on Risk of Bias

Dear Editor,

Harm from guideline implementation may arise if the recommendations are based on low-quality evidence (1). Ünal and Reinhart provided a review of fluids used for resuscitation in the Turkish Journal of Anaesthesiology and Reanimation, discussing the data on safety and adverse effects and updates from scientific and regulatory bodies of hydroxyethyl starch (HES) (2). Despite this high-quality data documenting the harm caused by HES in vulnerable patient populations, continued licencing of HES was considered a sad day as far as patients’ safety is considered.

Ünal and Reinhart (2) criticised the fact that the results of an observational study (RaFTinG), performed between 2010 and 2011 and fully published only in 2018 (3), were presented by the German Society of Anesthesiology and Intensive Care Medicine to the European Medicines Agency as a supporting evidence for the continued licencing of HES (https://www.oegari.at/web_files/dateiarchiv/1143/HES_Open_letter_EC_03.2018pdf.pdf). The study assessed the relationship between intravenous fluid therapy (crystalloids and colloids) and 90-day mortality, intensive care unit mortality and acute kidney injury and renal replacement therapy in patients in the intensive care unit. After a full multivariate adjustment, it reported that 6% HES 130/0.4 has no remarkable negative effects on 90-day mortality (3). Because the unadjusted and baseline-adjusted results clearly exhibited remarkably higher risks of these outcomes in patients treated with HES, owing to potential over-adjustment, it is requested that the fully adjusted results must be interpreted with caution.

However, over-adjustment may not be the only risk of bias of RaFTinG. On the basis of the MINORS scoring tool assessment of observational study quality (4), the research question of RaFTinG is not precisely addressed; no information is given if patient inclusion was consecutive; data were collected prospectively, but if the entire protocol was established before the beginning of the study remains uncertain; crucial protocol information, for example, primary and secondary endpoints, in the trial registry (US National Library of Medicine, registry number NCT01122277) is missing; follow-up period for renal endpoints that were restricted to the length of intensive care unit stay is not sufficiently documented; loss to follow-up was 22.5% (1,022 of 4,545 patients), which exceeds the proportion of 20.1% (707 of 3,523) experiencing the major endpoint of day-90 mortality; and prospective calculation of the study size is missing. In total, the MINORS score of 5 is clearly below the possible ideal score for non-comparative studies of 16 (4). These limitations fundamentally call into question the study’s internal and external validity.

Pre-publication of statistical analysis plans is necessary to provide confidence to readers that the analysis is driven by the desire to test a hypothesis, rather than by a data-mining exercise (5). In the case of RaFTinG, with no pre-publication analysis plan and a 5-year period between data collection and reporting, data mining is a typical risk.

Many of the standard critical care interventions were introduced into the clinical practice without evidence from randomised controlled trials and systematic reviews with overall low risk of bias (5). As exemplified with the continued licensing of HES despite the absence of evidence for its benefit, the use of HES based on the results of a biased research carries the risk of ongoing harm to patients.