Introduction

Gelatin, a degradation product of collagen [1], is a plasma expander used in patients in emergency departments, intensive care units, and operating and recovery rooms. Being one of the first synthetic colloids used for fluid resuscitation in humans [2], it came into more widespread use when different modifications of the gelatin molecules became available (oxypolygelatin [3], succinylated or modified fluid gelatin [4], and urea-linked gelatin [5]). These products entered the market long before current regulatory rules requiring proof of efficacy and safety of drugs came into existence [6].

Generally, gelatin solutions are considered to have no dosage-related side effects, specifically not impairing surgical hemostasis [7, 8], and being less harmful to the kidneys than other non-protein colloids [9, 10].

It was therefore surprising that retrospective analyses in observational studies found the use of gelatin to be associated with increased renal impairment and the need for transfusion products in patients with severe sepsis [11] and cardiac surgery [12]. Considering that the use of non-protein colloids is not associated with improved clinical outcomes [13, 14], potentially harmful effects of gelatins should be carefully explored.

Cumulative effects of gelatin on mortality were previously assessed by two Cochrane groups as part of larger meta-analyses that included studies until 2007 or 2008, respectively. Perel et al. [13] compared gelatin to crystalloid fluids in 506 critically ill adult patients, whereas Bunn et al. [15] compared gelatin with albumin or PPF in 636 patients thought to need volume replacement. No review found a significant effect on mortality. However, effects on renal function or blood product use were not systematically explored. In view of the emerging safety concerns, we considered that a systematic overview of the current clinical evidence on the safety of gelatin would be timely.

Methods

Eligibility criteria

Inclusion criteria

Inclusion criteria were: (1) prospective, randomized controlled trials, (2) hospital or pre-hospital setting, (3) patients with acute hypovolemia with need for fluid resuscitation, (4) study fluid was administered for resuscitation, (5) at least one intervention group received gelatin solution, and (6) at least one intervention group received another resuscitation fluid.

Exclusion criteria

Exclusion criteria were: (1) only volunteers or blood donors were recruited, (2) administration of fluid was solely for the purpose of volume preloading before anesthesia including volume loading for neural block, acute normovolemic or hypervolemic hemodilution without subsequent intra- or postoperative use, (3) retracted studies [16], (4) unsuitable control fluids, i.e., other synthetic colloids (as those may have similar risk profile), and (5) study results published in a language other than English, German, French, or Italian.

Search strategy

Three electronic databases were searched: Ovid Medline (1948–May 2011), EMBASE (1947–May 2011), and the Cochrane Central Register of Controlled Trials (CENTRAL) (through May 2011). In addition, reference lists from studies that met the inclusion criteria and from published systematic reviews were hand searched. Authors were contacted for further clarification if necessary.

The search terms used in MEDLINE or EMBASE are provided in the Electronic Supplement. Cochrane CENTRAL was searched using MeSH descriptor Polygeline (exploded).

Study selection, validity appraisal, and data extraction

Two reviewers independently screened the results of the search. Full-text manuscripts of potentially eligible articles were obtained and assessed independently against inclusion and exclusion criteria. Two out of five reviewers read the full-text reports and independently extracted the data into a datasheet. Differences were then compared and resolved by agreement or referral to a third reviewer. If results of one study were included in different publications, all extracted data were collated under one publication, and the others were removed from the datasheet.

Study validity was assessed by a tool developed by the Cochrane Collaboration’s tool for assessing risk of bias based on “low-high-uncertain” responses to the following seven domains: randomization, allocation concealment, blinding of participants, blinding of outcome assessment, incomplete outcome data, selective reporting, and other sources of bias [17].

Extracted data included: type and volume of gelatin, types and volumes of control fluids, fluid regimen, observation period, clinical condition, patient numbers, potential conflict of interest (COI) and type of funding, and reported outcomes. Outcomes of interest were defined as all-cause mortality at a time reported by the investigators, number of patients exposed to allogeneic blood products, acute kidney injury (AKI) by RIFLE criteria [18] or new need for renal replacement therapy (RRT), anaphylaxis, or itching. These outcomes were assessed in studies where gelatin was compared to suitable control fluids (albumin, plasma protein, or crystalloid); synthetic colloids such as hydroxyethyl starch, dextran, or other gelatins were defined as unsuitable control fluids because their comparative safety was not in question for this systematic review. Where possible, subgroup analyses were performed for different patient populations (i.e., critically ill, trauma, surgical patients), and subgroups with higher risk for gelatin effects were analyzed [high-dose gelatin (≥30 ml/kg), study periods >24 h].

Data analysis

The primary analysis of the data was descriptive, determining the proportion of studies meeting each of the criteria. Data were analyzed using SPSS 17.0 for Windows (SPSS Inc, Chicago, IL). Summarized fluid volumes were calculated as median values from all reported mean or median fluid volumes or from median/mean values for cumulative dose and body weight, if available.

For studies with more than one suitable control group, all control comparisons were pooled for the overall meta-analysis.

The relative risk of death and the relative risk of allogeneic blood product transfusion were calculated using a random-effects model (RevMan 5.1, Cochrane Collaboration). We specified subgroup analyses by study duration (>24 h vs. shorter periods), total gelatin dose (≥30 ml/kg versus less), and study population (critically ill patients vs. others). The statistical program used excludes studies with no events from the pooled estimate of relative risk.

Results

Descriptives

The structured search yielded 1,288 reports, of which 210 were read in full. The final sample contained 40 RCTs (full list is provided in Electronic Supplement). Studies were published between 1976 and 2010. The study flow graph is shown in Fig. 1.

Fig. 1
figure 1

Study flow

Risk of bias is shown in Fig. 2. Three studies had a a low risk of bias in all domains [1921], and five studies had a low risk in all domains except blinding [2226].

Fig. 2
figure 2

Risk of bias in included studies

Characteristics of all included studies are shown in Table 1. The total number of 3,275 patients was included in 40 studies, and 1,265 patients overall received gelatin. Median sample size in the gelatin groups was 15 patients (range 10–249). Two trials [27, 28] were multi-centered. In 32 RCTs (80 % of RCTs, n = 1,946 patients, 59 % of all 3,275 patients), the study period was ≤24 h. The total gelatin dose was 17 ml/kg (median, range 6–57). Twenty-nine RCTs (73 % of RCTs, 2,001/3,275, 61 % of all patients) investigated elective surgical patients, mostly undergoing cardiac surgery (18 RCTs, 819 patients). Three RCTs (723 patients) investigated critically ill adults. Two RCTs (59 patients) were in emergency department patients, and six RCTs (492 patients) were in children (Table 1).

Table 1 Characteristics of included studies

Outcomes

Mortality

Mortality rates were provided in 15 RCTs with 1,766 patients. Seven RCTs were in critically ill patients, and 8 RCTs were in trauma, emergency, or elective surgery. Overall mortality was similar in gelatin or control groups (RR 1.12, 95 % confidence interval CI, 0.87–1.44) (Fig. 3a).

Fig. 3
figure 3

Forest plots of pooled estimates

Exposure to allogeneic blood products

Eight RCTs with 712 patients reported the number of patients with transfusion of allogeneic blood products. Overall RR for exposure to allogeneic blood products was 1.28 (CI 0.89–1.83) (Fig. 3b).

AKI and need for RRT

Only three RCTs (n = 172 patients) reported the occurrence of AKI, ARF, or new need for RRT [25, 29, 30]. In two of these studies, gelatin was used in high dosage; one RCT was in critically ill adults and one in critically ill children. There was no difference in the risk of AKI between groups with a wide confidence interval around the RR estimate (Fig. 3c).

Subgroups

Higher doses of gelatin were used in four RCTs that reported deaths; no difference was seen between fluid groups. Exposure to allogeneic transfusions could be assessed in one RCT with high-dose gelatin; the RR was 7.62 (95 % CI 1.05, 55.55, only 9 transfusion events) [31] (Table 2).

Table 2 Subgroup outcomes

Study periods longer than 24 h were found in six RCTs that reported mortality, and in two RCTs that reported exposure to allogeneic transfusions. No difference was seen between groups (Table 2).

Other outcomes

Ten RCTs reported anaphylactoid reactions, but events occurred only in two RCTs (1/44 and 5/56 in the gelatin group, and 0/44 and 0/111 in the control groups) [21, 32]. No RCTs reported itching.

Discussion

This systematic review included all controlled studies that randomized adult and pediatric patients with acute hypovolemia due to surgery, trauma, severe infection, or critical illness to receive either gelatin or suitable albumin or crystalloid control fluids for resuscitation. We did not consider synthetic colloids as suitable control fluids because they have a similar risk profile and may all be associated with coagulopathy, renal dysfunction, and anaphylaxis [33, 34], whereas albumin and crystalloids are not associated with these side effects. Main outcomes were mortality, need for transfusion of allogeneic blood products, and renal failure reported as AKI according to RIFLE criteria or new need for RRT.

There was no difference in the overall risks of death and of patient exposure to allogeneic transfusions. However, all point estimates were on the one side of the potential range, and their confidence ranges were wide. Considering blood product use, the effect estimate by fixed effect model achieved significance (RR 1.36 [1.02, 1.81]), while the more conservative random effect model did not (RR 1.28 [0.89, 1.83]). RR for exposure to transfusion was increased in the subgroup of high-dose gelatin use (RR 7.62 [1.05, 55.55]), but here only one RCT was included [31], and the result is very imprecise. Almost all studies had considerable risk of bias and small sample sizes. The study period was 24 h or less in three-quarters of studies, and two-thirds investigated elective surgical patients, mostly from cardiac surgery, making inferences regarding the use of those fluids in critically ill patients more difficult.

Except for three studies [25, 29, 30], no RCT assessed renal failure in terms of AKI or RRT in comparison to a suitable control fluid. There are only few reports of renal nephroses associated with gelatin [35, 36] and one case report of acute renal failure after vascular surgery [37]. Recently, retrospective analyses found that AKI occurred more frequently in patients with sepsis as well as cardiac surgical patients who received gelatin compared to patients who had received only crystalloids [11, 12]. In each case we consider the evidence regarding influence of gelatins on renal outcome in critically ill patients as having low quality. Two previous fluid trials with critically ill patients found that the older 6 % starch solution had more deleterious effects on kidney function than a 3 % gelatin solution [9, 10]. Small trials that compared third-generation starches with 4 % gelatin solution in cardiac surgical patients found no difference in renal function [38, 39], whereas a retrospective sequential analysis of 346 patients with severe sepsis found that AKI occurred in 70 % of patients with HES 130/0.4 (adjusted p = 0.002) and in 68 % of patients with 4 % gelatin (adjusted p = 0.025) compared to 47 % of patients receiving only crystalloids [11].

The strength of this review is the comprehensive approach that identified RCTs published over the last 35 years, the systematic focus on potentially harmful effects, and the selection of studies with suitable control fluids. A limitation of its conclusion is the overall poor quality of the evidence coming from analysis of the included studies. This results from the design limitations of numerous studies, bringing their validity into question, imprecision stemming from the small numbers of events, indirectness about to the clinically important use of those drugs in critically ill populations, short observation periods, and relatively low doses used. In our view, the available data do not allow definitive inferences about the safety of gelatins in the populations of interest, especially when gelatins are used in larger doses over longer periods of time.

Conclusion

Gelatins were introduced into clinical practice before legislation in the aftermath of the thalidomide tragedy made clinical proof of safety mandatory [6]. Despite over 60 years of clinical experience with its use, the safety of gelatin in all settings in which it is used cannot be reliably assessed and confirmed. We suggest the need to investigate and establish such safety.