MAJOR bleeding and need for transfusions are common complications of surgical procedures and have been associated with poor postoperative and long-term outcomes.1,2 Efforts to minimize the use of limited resources such as blood products are essential, and the most obvious and probably the most effective strategy is to improve surgical techniques and hemostatic management.

The most extensively studied hemostatic agents include recombinant activated factor VII, aprotinin, desmopressin (DDAVP), and the antifibrinolytic lysine analogs aminocaproic acid and tranexamic acid. The safety of aprotinin has recently been questioned,3 and the safety of recombinant activated factor VII in major surgery is still under investigation.

DDAVP, originally developed and licensed for the treatment of inherited defects of hemostasis, given by slow intravenous infusion at a dose of 0.3 μg/kg, acts by releasing ultralarge von Willebrand factor multimers from endothelial cells, leading to an enhancement of primary hemostasis.4,5 In 1986, Salzman et al  .6 suggested that DDAVP reduces blood loss and transfusion requirements by approximately 30%, as compared with placebo, during complex cardiac surgery. Subsequent attempts to reproduce these findings encountered variable results: Many published trials were not adequately powered to assess clinically relevant outcomes, and several reviews concluded that although DDAVP helps to reduce perioperative blood loss, its effect is too small to influence other more clinically relevant outcomes, such as the need for blood products.7–10 A recent review states that there is little evidence that DDAVP is efficacious in conditions other than mild hemophilia A and von Willebrand disease,1 and the most recent meta-analysis10 concluded that, because there is no clear beneficial effect of using DDAVP to reduce blood loss or minimize transfusion with allogenic blood, further placebo-controlled trials of DDAVP as an adjunct to surgery in patients who do not have bleeding disorders seem to be unwarranted. Moreover, safety issues have been repeatedly raised: The main concern is that the use of a procoagulant drug may induce thromboembolic complications, especially in patients at high risk.

To address these issues, we conducted a systematic review and meta-analysis of data pooled from existing trials with the purpose of determining the impact of DDAVP on transfusion needs and to evaluate the risk of adverse events, namely thromboembolic complications.

Pertinent studies were independently searched in BioMed Central, CENTRAL, and PubMed (updated May 1, 2008) by four trained investigators (G.L., E.B., F.P., O.F.). The full PubMed search strategy, including as key words desmopressin  , DDAVP  , 1-desamino-8-D-arginine vasopressin  , surgery  , intervention  , and operation  , was developed according to Biondi-Zoccai et al  .11 and is available in the 1. Further hand or computerized searches involved the recent (2003–2008) conference proceedings from the European Society of Anesthesia, International Anesthesia Research Society, American Society of Anesthesiologists, European Society of Intensive Care Medicine, Society of Cardiovascular Anesthesiologists, European Association of Cardiothoracic Anesthesiologists, International Society of Thrombosis and Hemostasis, and American College of Chest Physicians congresses. In addition, we used backward snowballing (i.e  ., scanning of reference of retrieved articles and pertinent reviews) and contacted international experts for further studies. No language restriction was enforced, and non–English-language articles were translated before further analysis.

References obtained from database and literature searches were first independently examined at the title/abstract level by four investigators (G.L., E.B., F.P., O.F.), with divergences resolved by consensus, and then, if potentially pertinent, retrieved as complete articles. The following inclusion criteria were used for potentially relevant studies: (1) random allocation to treatment, (2) comparison of DDAVP versus  placebo, and (3) performed in adult surgical patients. The exclusion criteria were (1) case-matched studies, (2) duplicate publications (in this case, only the article reporting the longest follow-up was abstracted), (3) nonhuman experimental studies, and (4) no outcome data. Four investigators (G.L., E.B., F.P., O.F.) selected studies for the final analysis by independently assessing compliance to selection criteria. Divergences from the selection criteria were resolved by consensus (table 1).

Baseline, procedural, and outcome data were independently abstracted by five investigators (G.L., E.B., F.P., O.F., C.M.), with divergences resolved by consensus (table 1). Specifically, we extracted study design, population, clinical setting, DDAVP dosage, and treatment duration. At least two separate attempts at contacting original authors were made in case of missing data.

The primary endpoint of our analysis was to determine the effect of DDAVP on the overall transfusion of blood products (packed erythrocytes, fresh frozen plasma, platelets). Secondary endpoints included death, reoperation for bleeding, the number of patients receiving transfusion of blood products, and the effect of DDAVP on postoperative bleeding (defined as the maximum bleeding after DDAVP administration as described by authors). The possible side effects of DDAVP, such as hypotension during administration, myocardial infarction, and other thromboembolic complications, were collected as well.

Binary outcomes from individual studies were analyzed according to the Mantel–Haenszel model to compute individual odds ratios (ORs) with pertinent 95% confidence intervals (CIs), and a pooled summary effect estimate was calculated by means of fixed effect or random effect methods, when appropriate (i.e  ., fixed effect if I²was < 50% and random effect if I²was > 50%).11 Weighted mean differences and standardized mean differences with 95% CIs were computed for continuous variables.12 Statistical heterogeneity and inconsistency was measured using, respectively, Cochran Q tests and I²values.13 According to Higgins et al  .,14 I²values around 25, 50, and 75% were considered representing respectively low, moderate, and severe statistical inconsistency. Unadjusted P  values are reported throughout. The risk of small study bias (including publication bias) was assessed by visual inspection of funnel plots and computing the Egger test.15 In addition, subanalyses according to type of surgery (cardiac vs  . noncardiac) were performed as sensitivity analyses for all major endpoints. Statistical significance was set at the two-tailed 0.05 level for hypothesis testing and at 0.10 for heterogeneity testing. Computations were performed with SPSS 11.0 (SPSS, Chicago, IL) and RevMan 4.2 (a freeware available from The Cochrane Collaboration, Oxford, United Kingdom).12 

This study was performed in compliance with The Cochrane Collaboration and the Quality of Reporting of Meta-Analyses (QUOROM) guidelines.

Database searches, snowballing, and contacts with experts yielded a total of 348 citations (fig. 1). Excluding 290 nonpertinent titles or abstracts, we retrieved in complete form and assessed according to the selection criteria 58 studies. A total of 20 studies were further excluded because of their nonexperimental design, including the use of historic controls or studies with matched controls,16–21 double publication,22–25 1studying a pediatric population,26–33 not using placebo as control,34 or because there were no outcome data and further details could not be obtained by the authors.35 Three studies including comparison with other hemostatic drugs (tranexamic acid, aprotinin) were split,36–38 and only the comparison between DDAVP and placebo was considered. We finally identified 38 eligible randomized clinical trials (published in the period 1986–2004), which were included in the final analysis.6,36–712; Four of these 38 trials, reporting data on different patients’ populations, were split to correctly report the data and were considered as two distinct studies, 54,55,60,70 achieving a final number of 42 studies (table 1).

The 42 included randomized controlled studies (RCTs) comprised 2,488 patients (1,255 randomized to DDAVP and 1,233 to placebo) (table 1). Twenty-eight trials were performed in cardiac surgery, 6 in orthopedic surgery,38,46,49,55,66 3 in plastic surgery,52,70 2 in vascular surgery,42,58 1 in abdominal surgery,71 1 in orthognathic surgery,50 and 1 in nonspecified elective surgery.57 

DDAVP dosage varied slightly across studies (table 1), being mostly a single 0.3-μg/kg dose (only 6 authors repeated it) administered in 15–30 min during surgical hemostasis (only 8 studies administered immediately before surgery). All studies but one43 evaluated the prophylactic use of DDAVP. Three studies randomized patients according to their coagulative status as documented by thromboelastography,60 hemoSTATUS,44 and bleeding time.43 

Twelve studies reported positive results (with 1 study interrupted after an interim analysis because of excessive positive results), and 26 reported no results (with 1 study interrupted ad interim  because of the risk of increased myocardial infarction in the DDAVP group).

Most studies seemed to be of excellent quality (all of them were RCTs), as testified by the double-blind design specified by 33 of 38 trials. Only 1 RCT used a multicenter design,40 a feature that does not strictly impact on internal validity, but usually increases external validity of a trial. All trials but 1 were published as full articles. All investigators compared DDAVP versus  placebo. Only one author published more than 1 article as first author.37,64 

Overall analysis showed that, in comparison with placebo, DDAVP was associated with reduced requirements of blood product transfusion (standardized mean difference =−0.29 [−0.52 to −0.06] units per patient; P  = 0.01; fig. 2) in the overall analysis and in the noncardiac surgery patients. Furthermore, DDAVP treatment lead to significant reductions in the overall analysis in the total bleeding volume (weighted mean differences =−79 [−134 to −24] ml per patient; P  = 0.005; fig. 3), which were confirmed in subanalyses focusing on cardiac and noncardiac procedures. Despite these promising results, other clinical endpoints were not statistically different between DDAVP and control, suggesting that the overall patient population was still not large enough to achieve an adequate statistical power. Specifically, we found similar rates of patients undergoing reoperation for bleeding (fig. 4) or receiving transfusion blood products (fig. 5) in both cardiac and noncardiac groups, with trends toward a reduced transfusion of platelets in patients undergoing cardiac surgery (fig. 6).

The adverse event most frequently associated with DDAVP treatment was postadministration hypotension (OR = 4.84 [2.31–10.13]; P  < 0.001; fig. 7), although this side effect was transient and not related to any serious adverse effect, as stated by the authors. The significant improvement in the rates of all major endpoints was obtained without a significant increase in thromboembolic adverse events: death, myocardial infarction, and any other thrombosis (all P  > 0.05 at overall analyses as well as selecting only cardiac or noncardiac patient; figs. 8–10). In particular, we recorded 31 of 816 (3.8%) myocardial infarctions in the DDAVP group versus  23 of 793 (2.9%) in the placebo group (OR = 1.27 [0.73–2.20]; P  = 0.4) and 26 of 899 (3.2%) other thromboses in the DDAVP group versus  22 of 877 (2.5%) in the placebo group (P  = 0.5). Overall, the thromboembolic complications were 57 of 1,002 in the DDAVP versus  45 of 979 in the placebo group (P  = 0.3). Mortality was 9 of 771 (1.2%) in the DDAVP group versus  7 of 768 (0.9%) in the placebo group (P  = 0.6).

The significant reduction in transfusion of blood products was confirmed in the overall population and in noncardiac surgery patients when only double-blind RCTs were analyzed (P  < 0.05 for both analyses).

This meta-analysis pooled data from 38 small, underpowered trials and demonstrated that DDAVP is associated with a slight but significant reduction in transfusion of blood products in the overall studied population as well as in the subgroup of patients who underwent noncardiac surgery. Twelve trials included in this analysis, although underpowered, demonstrated positive trends, which is consistent with the overall results of our meta-analysis.

Despite improvements in surgical skills and strategies, bleeding remains a problem in surgery. Surgical blood loss and associated transfusions of allogenic blood products may be reduced by pharmacologic manipulation of hemostasis. The choice among hemostatic agents is ultimately based on the clinician’s sense of the expected therapeutic efficacy, the safety profile, and the costs—a balance that may vary depending on the characteristics of individual patients and specific clinical settings.1 The data available so far indicate that aprotinin, lysine analogs, and recombinant factor VIIa are potent hemostatic agents, whereas DDAVP had no documented efficacy.1 This meta-analysis showed that DDAVP has a small effect in reducing the volume of transfused blood products but no effect on the proportion of individuals who receive transfusions. The clinical impact of such a finding is questionable but should be considered in the context of the low cost of DDAVP: At a single-dose regimen of 0.3 μg/kg, the cost of a treatment may be considered in the range of $5–10 US. Nonetheless, because there is still uncertainty of harm from death and myocardial infarction, probably the use of DDAVP should be limited to patients with refractory microvascular bleeding after surgery.

No meta-analysis of controlled clinical trials involving the prophylactic or therapeutic use of DDAVP supported the clinically significant effect of DDAVP in reducing transfusion needs after surgery. Hence, we conducted a meta-analysis of 38 studies comparing DDAVP with placebo, evidencing, for the first time, a statistically significant reduction in requirements of blood products (standardized mean difference =−0.29 [−0.52 to −0.06] units per patient; P  = 0.01) in patients undergoing surgery. Our results are in contrast with the recent meta-analysis of Carless et al  .,10 the only one to include patients undergoing noncardiac surgery: The authors concluded that there seemed to be no clear beneficial effect of using DDAVP to reduce blood loss or minimize transfusion with allogenic blood and that further placebo-controlled trials of DDAVP as an adjunct to surgery in patients who do not have bleeding disorders seemed to be unwarranted. The most important difference with this previous work is in the number of identified trials: 38 in our meta-analysis versus  25 in Carless’ one. This discrepancy should be attributed to the inclusion of 1 recently published article,63 of 2 studies that reported data on bleeding but not on blood transfusion,39,65 and to a better research strategy that permitted identification of 10 additional studies.43,46,49,52,55–58,70 2In addition, we imputed effect estimates and the corresponding SEs from included studies by means of an unbiased approach recently reported and validated.72 3Furthermore, we evaluated the use of numerous blood products (packed erythrocytes, fresh frozen plasma, and platelets), whereas Carless et al  . only considered blood transfusions. Because DDAVP improves platelet adhesion and because all blood products have risks and side effects when transfused, we are convinced that our approach is rational.

Similar considerations apply to previous systematic reviews such as Levi et al  .,7 Laupacis et al  .,73 Cattaneo et al  .,9 and Fremes et al  . 8 : The failure of previous RCTs and meta-analyses to demonstrate the effects of DDAVP on blood products should be attributed to lack of power. In fact, these authors did not include very recent reports, such as those published after 199838,44,61,62,71; they did not include patients undergoing noncardiac surgery; and, with the exception of Cattaneo et al  .,9 they did not focus on DDAVP but also reviewed other pharmacologic strategies to decrease excessive blood loss. None of these authors found a reduction in transfusion of blood products in patients receiving DDAVP: Levi et al  .,7 Laupacis et al  .,73 and Cattaneo et al  . 9 identified a decrease in perioperative blood loss either in the general population7 or only in patients receiving aspirin,73 with excessive bleeding9 defined as a mean blood loss in placebo-treated patients within the upper fourth of the blood loss distribution (≥ 1,180 ml/24 h).

DDAVP is a synthetic analog of the neurohypophyseal nonapeptide arginine vasopressin. It is known that this hormone causes the appearance of larger von Willebrand factor multimers in addition to increased concentrations of von Willebrand factor multimers and factor VIII; a role for DDAVP has thus been indicated in patients with mild hemophilia or von Willebrand disease. DDAVP has also been shown to shorten bleeding time in other conditions, including uremia, chronic liver disease, and aspirin ingestion.

The use of any drug that potentiates hemostasis inevitably carries a risk of thrombosis, particularly in patients with atherosclerosis or other risk factors for thrombosis. Early case reports suggested a potential relation between DDAVP and thromboembolic complications, and 2 previous meta-analyses on pharmacologic strategies to decrease excessive blood loss in cardiac surgery confirmed that perioperative myocardial infarction was more common in DDAVP-treated patients when compared with a placebo-treated group: Laupacis et al  .73 suggested a trend toward an increased incidence of myocardial infarction (4.4% vs  . 1.6%) in the DDAVP group studying 12 trials and 793 patients (OR, 1.85; 95% CI, 0.74–4.62), whereas Levi et al  .7 found that DDAVP was associated with a 2.4-fold increase in the risk of perioperative myocardial infarction in 7 trials (OR, 2.39; 95% CI, 1.02–5.60). A larger and more updated analysis of 1,833 patients, however, showed no increased risk of thromboembolic complications. Mannucci et al  .74 stated that in 31 reports on the use of DDAVP in surgery published up to 1993, 57 thromboembolic events were reported (33/956 = 3.4% in DDAVP-treated patients vs  . 24/877 = 2.7% in placebo-treated patients; P  = 0.38), with 19 myocardial infarctions versus  13 in the placebo group, 8 and 2 cases of ischemic stroke, 1 and 2 cases of arterial thromboembolism, and 5 and 7 cases of venous thromboembolism. Our results are consistent with these findings and confirm a similar incidence of thromboembolic events: 57/1,002 = 5.4% versus  45/979 = 4.6% in the DDAVP and placebo groups, respectively. Nonetheless, it should be acknowledged that RCTs are not necessarily the best study design to assess the harm of an intervention. The sample size is usually too small even with the combining of studies to adequately address harm. Cohort studies, registry studies, use of administrative data, or some active surveillance techniques are often better. Also, thromboembolic events are not easy to detect, and some of the studies included in this meta-analysis may not have had a systematic approach for detecting them. The recent lessons that have been learned from aprotinin3 should make the medical community very cautious about making pronouncements on safety. Concerns about safety remain unanswered, even if this meta-analysis itself does not raise a major concern in this regard.

The only statistically significant difference in side effects between the two groups was the mild, transient arterial hypotension observed by seven authors (37/442 = 8.4% vs  . 9/419 = 2.1%; P  < 0.001) that is consistent with the reported increase in cardiac output and decrease in systemic vascular resistance observed in patients receiving DDAVP.63 This effect is related to a mild vasodilating effect of DDAVP and is not associated with changes in filling pressure, heart rate, or right ventricular function. The etiology of this vasodilation is uncertain but is related to the rate of administration. Histamine release does not seem to be the mechanism of the response, even with rapid administration of the drug. A more likely possibility is that DDAVP interacts with a subgroup of vasopressin receptors that cause peripheral vasodilatation.63 Whether DDAVP acts as a vasodilator in the coronary vasculature is unknown.

As suggested by recent guidelines75 and a RCT,44 it would be wise to administer DDAVP using point-of-care tests for platelet function in patients with excessive postoperative bleeding. Most probably, there are groups of patients who may benefit more from DDAVP due to dysfunctional platelets. Overall perioperative platelet dysfunction from congenital (mild von Willebrand disease), pharmacologic (local anesthetics, nonsteroidal antiinflammatory drugs, heparins, ephedrine, and others), or physical (hypothermia, turbulent blood flow, artificial surfaces, and blood loss) causes may be around 3–5% of operated patients, making platelet dysfunction the most common hemostatic defect.76 Patients who may benefit from DDAVP include those who undergo prolonged operations with cardiopulmonary bypass,6 those with excessive postoperative bleeding (e.g  ., > 1,180 ml/24 h),9 those taking platelet-inhibiting drugs,73 those at high risk of excessive bleeding as identified by tests of hemostatic function,17,60 or those with chronic renal failure or liver dysfunction.

The limitations of systematic reviews and meta-analyses are well known77 and include the level of uniformity among study populations as well as the primary endpoints in each of the studies, which on the other hand were often of high internal validity. A particular limitation of our analysis is the lack of uniform reporting of bleeding, which may have indeed decreased the overall statistical power of the meta-analysis. Nonetheless, we strived to comply with the most stringent guidelines of The Cochrane Collaboration and of the QUOROM statement.

Only an individual patient-data meta-analysis or a large and adequately powered RCT could provide a sounder and more rigorous appraisal of the clinical role of DDAVP in this clinical setting.

A heterogeneity factor in our study is the presence of different surgical procedures, which can raise the problem of pooling the data together. Nonetheless, most of these studies focused on patients undergoing cardiovascular procedures, liver surgery, and major orthopedic procedures that are associated with severe bleeding.1 

Probably the most important limitation of this meta-analysis (and of all the RCTs conducted in this setting) is the absence of midterm follow-up: Patients receive prothrombotic drugs to reduce transfusion rate and limit midterm complications (lung dysfunction and increased rate of perioperative infections), but no study had a follow-up longer than 30 days, and most studies limited their observation to the perioperative period.

The fact that so many patients were required (roughly 2,500 in this meta-analysis) to demonstrate a beneficial effect of DDAVP suggests that the benefit is quite small or quite variable or that only a small subset of patients benefit.

A large multicenter RCT with a 1-yr follow up is currently recruiting patients with excessive microvascular bleeding after cardiac surgery (NCT00337766).

DDAVP slightly reduced blood loss (almost 80 ml per patient) and transfusion requirements (almost 0.3 units per patient) in surgical patients, without reduction in proportion of patients who received transfusions. This meta-analysis suggests the importance of further large randomized controlled studies in patients with or at risk of perioperative pathologic microvascular bleeding.

(randomized controlled trial[pt] OR controlled clinical trial[pt] OR randomized controlled trials[mh] OR random allocation[mh] OR double-blind method[mh] OR single-blind method[mh] OR clinical trial[pt] OR clinical trials[mh] OR (clinical trial[tw] OR ((singl*[tw] OR doubl*[tw] OR trebl*[tw] OR tripl*[tw]) AND (mask*[tw] OR blind[tw])) OR (latin square[tw]) OR placebos[mh] OR placebo*[tw] OR random*[tw] OR research design[mh:noexp] OR comparative study[mh] OR evaluation studies[mh] OR follow-up studies[mh] OR prospective studies[mh] OR cross-over studies[mh] OR control*[tw] OR prospectiv*[tw] OR volunteer*[tw]) NOT (animal[mh] NOT human[mh]) NOT (comment[pt] OR editorial[pt] OR meta-analysis[pt] OR practice-guideline[pt] OR review[pt])) AND ((desmopressin OR davp OR “1-desamino-8-D-arginine vasopressin”) AND (surgery OR intervention* OR operation*)).