Prothrombin complex concentrate for reversal of vitamin K antagonist treatment in bleeding and non‐bleeding patients
Abstract
Background
Treatment with vitamin K antagonists is associated with increased morbidity and mortality. Reversal therapy with prothrombin complex concentrate (PCC) is used increasingly and is recommended in the treatment of patients with bleeding complications undertaking surgical interventions, as well as patients at high risk of bleeding. Evidence is lacking regarding indication, dosing, efficacy and safety.
Objectives
We assessed the benefits and harms of PCC compared with fresh frozen plasma in the acute medical and surgical setting involving vitamin K antagonist‐treated bleeding and non‐bleeding patients. We investigated various outcomes and predefined subgroups and performed sensitivity analysis. We examined risks of bias and applied trial sequential analyses (TSA) to examine the level of evidence, and we prepared a 'Risk of bias’ table to test the quality of the evidence.
Search methods
We searched the following databases from inception to 1 May 2013: Cochrane Central Register of Controlled Trials (CENTRAL); MEDLINE (Ovid SP); EMBASE (Ovid SP); International Web of Science; Latin American and Caribbean Health Sciences Literature (LILACS) (via BIREME); the Chinese Biomedical Literature Database; advanced Google and the Cumulative Index to Nursing and Allied Health Literature (CINAHL). We applied a systematic and sensitive search strategy to identify relevant randomized clinical trials and imposed no language or date restrictions. We adapted our MEDLINE search strategy for searches in all other databases. We reran the search in October 2014 and found one potential new study of interest. We added this study to a list of ‘Studies awaiting classification', and we will incorporate this study into the formal review findings at the time of the review update.
Selection criteria
We included randomized controlled trials (RCTs), irrespective of publication status, date of publication, blinding status, outcomes published or language. We contacted investigators and study authors to request relevant data.
Data collection and analysis
Three review authors independently abstracted data and resolved disagreements by discussion. Our primary outcome measures were 'overall mortality longest follow‐up' and 'overall 28‐day mortality'. We performed subgroup analyses to assess the effects of PCC in adults in terms of various clinical and physiological outcomes. We presented pooled estimates of the effects of interventions on dichotomous outcomes as risk ratios (RRs), and on continuous outcomes as mean differences (MDs), with 95% confidence intervals (CIs). We assessed risk of bias by assessing trial methodological components and risk of random error through TSA.
Main results
We included four RCTs with a total of 453 participants and determined that none of these trials had overall low risk of bias. We found six ongoing trials from which we were unable to retrieve further data. Three trials provided data on mortality. Meta‐analysis showed no statistical effect on overall mortality (RR 0.93, 95% CI 0.37 to 2.33; very low quality of evidence). We were unable to associate use of PCC with the number of complications probably related to the intervention (RR 0.92, 95% CI 0.78 to 1.09; very low quality of evidence). Lack of transfusion data and apparent differences in study design prevented review authors from finding a beneficial effect of PCC in reducing the volume of fresh frozen plasma (FFP) transfused to reverse the effect of vitamin K antagonist treatment. The number of new occurrences of transfusion of red blood cells (RBCs) did not seem to be associated with the use of PCC (RR 1.08, 95% CI 0.82 to 1.43; very low quality of evidence). Still, the included studies demonstrate the possibility of equally reversing vitamin K‐induced coagulopathy using PCC without the need for transfusion of FFP. No effect on other predefined outcomes was observed.
Authors' conclusions
In the four included RCTs, use of prothrombin complex concentrate does not appear to reduce mortality or transfusion requirements but demonstrates the possibility of reversing vitamin K‐induced coagulopathy without the need for transfusion of fresh frozen plasma. All included trials have high risk of bias and are underpowered to detect mortality, benefit or harm. Clinical and statistical heterogeneity is high, and definitions of clinically important outcomes such as adverse events are highly dissimilar between trials. Only weak observational evidence currently supports the use of PCC in vitamin K antagonist‐treated bleeding and non‐bleeding patients, and the current systematic review of RCTs does not support the routine use of PCC over FFP. Additional high‐quality research is urgently needed.
PICOs
Plain language summary
Prothrombin complex concentrate for reversal of vitamin K antagonist treatment in bleeding and non‐bleeding patients
Prothrombin complex concentrate (PCC) is a drug that contains a source of proteins involved in the human blood clotting process. Patients medicated with vitamin K antagonists (blood thinning drug) have low blood levels of these important blood clotting proteins. Therefore these patients will be at increased risk of spontaneous and traumatic bleeding events. Also, when these patients experience a bleeding event, this will lead to progressive loss of these important blood clotting proteins. This process causes a vicious circle, thereby increasing risks of illness and death.
In the present Cochrane systematic review, we assessed the benefits and harms of prothrombin complex concentrate in vitamin K antagonist‐treated bleeding and non‐bleeding patients who are undertaking acute surgical intervention. We searched the databases until 1 May 2013. We identified four randomized trials (453 participants) involving neurological and cardiac surgical settings, as well as medical reversal of vitamin K‐intoxication among participants. We found six ongoing trials but were unable to retrieve these data. We reran the search in October 2014 and found one new study of potential interest. We added this study to a list of ‘Studies awaiting classification', and we will incorporate this study into the formal review findings at the time of the review update.
We could not identify any beneficial effects of prothrombin complex concentrate on death. In our predefined outcomes, we identified a decreased volume of fresh frozen plasma transfused for reversal of vitamin K antagonist treatment. We could not identify statistical differences in reduced blood loss, harm or numbers of adverse events in the group treated with PCC. However, all trials were of low quality and small, and all were characterized by a high level of heterogeneity. Therefore, evidence in support of PCC in vitamin K antagonist‐treated bleeding and non‐bleeding patients receiving acute surgical intervention remains weak.
Authors' conclusions
Summary of findings
| Prothrombin complex concentrate versus any comparator for reversal of vitamin K antagonist treatment in bleeding and non‐bleeding patients | ||||||
| Patient or population: patients with vitamin K antagonist‐induced bleeding requiring acute intervention | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | Number of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Control | Prothrombin complex concentrate | |||||
| Overall mortality ‐ longest follow‐up | Study population | RR 0.93 | 421 | ⊕⊝⊝⊝ | Effect of the use of PCC on mortality was uncertain. None of the included studies were powered to detect differences in mortality. Large confidence intervals. Few participants and few events | |
| 84 per 1000 | 78 per 1000 | |||||
| Moderate | ||||||
| 91 per 1000 | 85 per 1000 | |||||
| Number of new occurrences of RBC transfusion | Study population | RR 1.08 | 376 | ⊕⊝⊝⊝ | Number of new occurrences of blood transfusion did not seem to be associated with use of PCC. Large confidence intervals. Few participants and few events | |
| 319 per 1000 | 344 per 1000 | |||||
| Moderate | ||||||
| 300 per 1000 | 324 per 1000 | |||||
| Number of complications probably related to the intervention | Study population | RR 0.92 | 442 | ⊕⊝⊝⊝ | Assessment of safety was not uniform, which raises concerns of underreporting. Large confidence intervals. Few participants and few events | |
| 573 per 1000 | 527 per 1000 | |||||
| Moderate | ||||||
| 563 per 1000 | 518 per 1000 | |||||
| Transfusion of RBCs | Mean transfusion of RBCs in the intervention groups was | 370 | ⊕⊝⊝⊝ | Trial sequential analysis of PCC vs FFP on quantity of RBCs transfused from 2 trials led to rejection of an intervention effect of 125 mL based on sparse data and repetitive analyses. Large confidence intervals. Few participants and few events | ||
| *The basis for the assumed risk (e.g. median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence. The quality of the evidence is downgraded 3 levels based on risk of bias in the included studies and the relatively small overall number of participants. | ||||||
| aOnly poster and abstract published, open‐label study. Industry sponsored. Industry funded, small sample size. Describes several breaches of the study protocol. dImprecision of results (large confidence intervals). Indirectnes of evidence (few participants and few events). CI = confidence interval. FFP = fresh frozen plasma. PCC = prothrombin complex concentrate. RBC = red blood cell. RR = risk ratio. | ||||||
Background
Description of the condition
Severe bleeding can cause haemorrhagic shock, which is a lethal state of tissue hypoxia ultimately leading to cellular ischaemia and hence organ dysfunction, failure and death (Bruce 2008). An increasing number of patients receive oral anticoagulation therapy with a vitamin K antagonist such as warfarin or phenprocoumon to treat or prevent new occurrences of venous thromboembolism. However this approach requires special consideration with regard to co‐morbidity, interacting medications and potentially lethal haemorrhagic complications. The primary medical indications are atrial fibrillation, venous thromboembolism and mechanical heart valves (Leissinger 2008; Levy 2008; Vang 2011). Large‐scale epidemiological studies have reported new occurrences of major bleeding complications from vitamin K antagonist treatment ranging from 1.1% to 1.5%, with gastrointestinal, urinary and intracranial sites most frequently involved (Bruce 2008; Franchini 2010; Leissinger 2008; Palareti 1996). This potentially lethal condition requires rapid reversal of oral anticoagulant therapy, which traditionally has been executed by large‐volume transfusions of fresh frozen plasma. However, this strategy presents important safety concerns such as volume overload, dilutional coagulopathy, blood group specificity, lack of viral inactivation and risk of transfusion‐related acute lung injury (Kor 2010; Leissinger 2008; Ozgonenel 2007).
Description of the intervention
Inhibition of the biosynthesis of vitamin K‐dependent coagulation factors can be reversed by several treatments. Withholding oral anticoagulation therapy while administering vitamin K induces hepatic de novo synthesis of vitamin K‐dependent coagulation factors. In this context, normalization of haemostasis lasts hours to days. Fresh frozen plasma (FFP) generally has been considered the 'standard of care' for reversal of vitamin K antagonist therapy (Leissinger 2008). It provides partial reversal of the coagulopathy through replacement of exogenous factors II, VII, IX and X but presents important safety concerns and challenges involving 'dosage/international normalized ratio (INR) correction' and time needed to prepare the infusion (Demeyere 2010; Holland 2006). Prothrombin complex concentrate (PCC), originally developed for the treatment of patients with haemophilia B (Ostermann 2007), is derived from the cryoprecipitate supernatant of large plasma pools after removal of antithrombin and factor XI. Prothrombin complex concentrate represents a source of vitamin K‐dependent coagulation factors II, VII, IX and X and the naturally occurring anticoagulants activated proteins C and S. It has a final overall clotting factor concentration approximately 25 times higher than that of normal plasma (Franchini 2010). Increasing availability of purified and recombinant factor IX has led to expanded use of PCC for other indications (e.g. reversal of vitamin K‐induced bleeding disorders) (Franchini 2010).
Prothrombin complex concentrate, which has come into focus increasingly with regard to acute reversal of vitamin K antagonist bleeding disorders, constitutes treatment for two complex clinical subgroups with a similar need for acute intervention but with different therapeutic indications (Levi 2009; Pindur 1999; Schick 2009; Vigué 2009).
-
Vitamin K antagonist‐treated patients with a bleeding complication requiring acute surgical intervention (e.g. intracranial bleeding).
-
Vitamin K antagonist‐treated patients with supranormal INR levels and no ongoing bleeding complications who have the need for acute surgical intervention. .
How the intervention might work
Acute administration of PCC to reverse vitamin K‐induced coagulopathy over the past few years has reduced the need for blood transfusions, thereby reducing risks of volume overload, haemodilution coagulopathy and transfusion‐related acute lung injury, as well as possible transmission of viral infection (Bruce 2008; Franchini 2010; Riess 2007; Schick 2009).
Prothrombin complex concentrate theoretically offers several advantages over FFP: decreased time required for initiation of goal‐directed therapeutic interventions (e.g. surgical intervention), as thawing of plasma is bypassed (Taberner 1976; Van Aart 2006); reduced infusion volume, as PCC treatment of 1 to 2 mL/kg corresponds to a volume of approximately 15 mL/kg of FFP, thereby diminishing the risk of a compromised vulnerable cardiovascular system (Franchini 2010); lack of blood group specificity, reducing the risk of transfusion‐related complications; and finally, viral inactivation, which may minimize the risk of pathogen transmission.
Prothrombin complex concentrate seems to have a faster INR correction time than FFP, and several studies have suggested a role for PCC in the setting of acute neurosurgery in reducing clinical progression of intracerebral haemorrhage (Boulis 1999; Franchini 2010; Steiner 2011).
Why it is important to do this review
Use of FFP constitutes a well‐established procedure to reverse vitamin K antagonist‐induced bleeding disorders (Ansell 2008). However, avoiding blood transfusions may result in a beneficial effect in terms of possibly reduced mortality and morbidity (Koch 2006a; Koch 2006b; Theusinger 2009). Infusion of coagulation factor concentrates in low volumes might offer a potential treatment for patients experiencing vitamin K antagonist‐induced bleeding complications and people with supranormal INR levels requiring acute surgical intervention. Prothrombin complex concentrate is already included in American and European guidelines for reversal of acute life‐threatening bleeding related to elevated INR (Ansell 2008; Baglin 2006). However, it is important to stress that rapid reversal of an elevated INR is not necessarily associated with better clinical outcomes. Differentiating between therapeutic indications and clinical risk assessment is vital for balancing possible benefits (e.g. reduced blood transfusions and avoidance of volume overload, dilutional coagulopathy, pathogen transmission, transfusion‐related acute lung injury and the time‐consuming process of thawing blood group‐specific FFP) and drawbacks (e.g. inducing life‐threatening thrombotic events such as acute myocardial infarction, pulmonary embolism or cerebral venous thromboembolism) associated with treating an acute bleeding complication or preventing development of the bleeding complication by using exogenous coagulation factors (Warren 2009). More compelling evidence is needed in this area, and increasing use of PCC should be accompanied by an ongoing systematic assessment of its potential benefits and adverse events. The aim of this review is to assess the evidence suggesting that PCC is beneficial or harmful for vitamin K antagonist‐treated patients undergoing acute surgery compared with placebo, FFP or other haemostatic agents.
Objectives
We assessed the benefits and harms of PCC compared with fresh frozen plasma in the acute medical and surgical setting involving vitamin K antagonist‐treated bleeding and non‐bleeding patients. We investigated various outcomes and predefined subgroups and performed sensitivity analysis. We examined risks of bias and applied trial sequential analyses (TSA) (Wetterslev 2008) to examine the level of evidence, and we prepared a 'Risk of bias’ (ROB) table to test the quality of the evidence (Guyatt 2008).
Methods
Criteria for considering studies for this review
Types of studies
We included randomized controlled trials (RCTs), irrespective of publication status, date of publication, blinding status, outcomes published or language. We contacted investigators and study authors to request relevant data. We included unpublished trials only if trial data and methodological descriptions were provided in written form or could be retrieved from study authors. We included cluster‐randomized trials because the number of trials eligible for inclusion was expected to be low, but we excluded cross‐over trials and observational studies. We included in this review trials that applied interventional invasive procedures such as endoscopy, interventional radiology and vascular procedures (i.e. embolization, stenting of aneurisms) and less extensive procedures (i.e. nasal balloon tamponade) to extend the general applicability of the evidence.
Types of participants
We included participants who were treated with vitamin K antagonists (VKAs) and were undergoing acute surgery or surgical intervention related to bleeding complications arising from VKA treatment. We also included VKA‐treated participants with supranormal INR values undergoing acute surgery or surgical intervention as the result of co‐morbidity. We excluded trials that examined neonates and participants with hereditary bleeding disorders and/or significant liver dysfunction.
Types of interventions
We included trials on PCC versus FFP. We included trials using any dose of PCC over any duration of administration and/or co‐interventions. Furthermore, we included trials that compared PCC with other haemostatic agents (e.g. vitamin K, recombinant factor VIIa, other plasma derivatives). We undertook separate subgroup analyses of trials in which PCC was compared with other active interventions or was combined with co‐interventions.
-
PCC versus any comparator.
-
PCC versus placebo or no treatment.
-
PCC versus FFP.
-
PCC versus other haemostatic agents (vitamin K, recombinant factor VIIa, other plasma derivatives, factor‐substitution products).
-
PCC in combination with other haemostatic agents versus placebo, no treatment or usual treatment with or without haemostatic agents.
Types of outcome measures
Primary outcomes
-
Overall mortality: We used longest follow‐up data from each trial regardless of the period of follow‐up.
-
Overall 28‐day mortality: We included data provided as 30‐day mortality in the same analysis.
Secondary outcomes
-
The number of new occurrences of blood transfusion (e.g. avoidance of transfusion) and quantity of blood products transfused.
-
Complications probably related to the intervention (e.g. thrombotic episodes (pulmonary embolism, myocardial infarction, disseminated intravascular coagulation), heparin‐induced thrombocytopenia, major immunological and allergic reactions (TRALI), cardiopulmonary overload (TACO), infection, sepsis (e.g. transmission of viral infections)).
-
Complications during the inpatient stay not specific to the trial intervention (e.g. pneumonia, congestive cardiac failure, respiratory failure, renal failure).
-
Number of days in hospital.
-
Mean length of stay in intensive care unit (ICU).
Search methods for identification of studies
Electronic searches
We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (2013, Issue 5; see Appendix 1); MEDLINE (Ovid SP, 1950 to 1 May 2013; see Appendix 2); EMBASE (Ovid SP, 1980 to 1 May 2013; see Appendix 3); International Web of Science (1964 to 1 May 2013; see Appendix 4); Latin American Caribbean Health Sciences Literature (LILACS via BIREME, 1982 to 1 May 2013; see Appendix 5); the Chinese Biomedical Literature Database; advanced Google; and the Cumulative Index to Nursing & Allied Health Literature (CINAHL, 1980 to 1 May 2013; see Appendix 6).
We applied a systematic and sensitive search strategy to identify relevant RCTs with no language or date restrictions. We adapted our MEDLINE search strategy for searching all other databases.
We reran the search in October 2014. We will add new studies of potential interest to a list of Studies awaiting classification, and we will incorporate them into formal review findings at the time of the review update.
Searching other resources
We handsearched the reference lists of reviews, randomized and non‐randomized studies and editorials to look for additional studies. We contacted the main authors of studies to ask about missed, unreported or ongoing studies. We contacted pharmaceutical companies for information on unpublished trials. We searched for ongoing clinical trials and unpublished studies on the following Internet sites (search date to 21 October 2014).
Data collection and analysis
Selection of studies
We assessed reports identified through the described searches and excluded obviously irrelevant reports. Two review authors independently examined these studies for eligibility (MJ, JL). We performed this process without blinding of study authors, institution, journal of publication or results. We resolved disagreements by discussion, and when no agreement was reached, we consulted a third person (AA). We provide a detailed description of the search and assessment.
Data extraction and management
Using a data extraction sheet (Appendix 7), we evaluated each study, entered the data into Review Manager 5 (RevMan 5.3) and checked them for accuracy. If data in the identified reports were unclear, we contacted study authors to invite them to provide further details. Two review authors (MJ, JL) independently extracted data in accordance with Appendix 7. We resolved disagreements by discussion, and if no agreement was reached, we consulted a third person (AW).
Assessment of risk of bias in included studies
Two review authors (MJ, JL) independently assessed risk of bias using the criteria outlined in Higgins 2011. We resolved disagreements by discussion, and if no agreement was reached, we consulted a third person (AW). We addressed each question of validity systematically as described here.
1. Random sequence generation
Assessment of randomization: sufficiency of the method in producing two comparable groups before intervention.
Grade: ’low risk’: a truly random process (e.g. random computer number generator, coin tossing, throwing dice); ’high risk’: any non‐random process (e.g. date of birth, date of admission by hospital or clinic record number or by availability of the intervention); or ’unclear risk’: insufficient information.
2. Allocation concealment
Allocation method prevented investigators or participants from foreseeing assignment.
Grade: ’low risk’: central allocation or sealed opaque envelopes; ’high risk’: using open allocation schedule or other unconcealed procedure; or ’unclear risk’: insufficient information.
3. Blinding
Assessment of appropriate blinding of the team of investigators and participants: person responsible for participant care, participants and outcome assessors.
Grade: ’low risk’: blinding considered adequate if participants and personnel were kept unaware of intervention allocations after inclusion of participants into the study, and if the method of blinding involved a placebo indistinguishable from the intervention, as mortality is an objective outcome; ’high risk’: not double‐blinded, categorized as an open‐label study, or without use of a placebo indistinguishable from the intervention; ’unclear risk’: blinding not described.
4. Incomplete outcome data
Completeness of outcome data, including attritions and exclusions.
Grade: ’low risk’: numbers and reasons for drop‐outs and withdrawals in the intervention groups described, or no drop‐outs or withdrawals was specified; ’high risk’: no description of drop‐outs and withdrawals provided; ’unclear risk’: report gave the impression of no drop‐outs or withdrawals, but this was not specifically stated.
5. Selective reporting
The possibility of selective outcome reporting.
Grade: ’low risk’: reported outcomes prespecified in an available study protocol, or, if this is not available, published report includes all expected outcomes; ’high risk’: not all prespecified outcomes reported, reported using non‐prespecified subscales, reported incompletely or report fails to include a key outcome that would have been expected for such a study; ’unclear risk’: insufficient information.
6. Other bias
Assessment of any possible sources of bias not addressed in domains 1 to 5.
Grade: ’low risk’: report appears to be free of such biases; ’high risk’: at least one important bias is present that is related to study design, early stopping because of some data‐dependent process, extreme baseline imbalance, academic bias, claimed fraudulence or other problems; or ’unclear risk’: insufficient information, or evidence that an identified problem will introduce bias.
With reference to domains 1 to 6 above, we planned to assess the likely magnitude and direction of the bias, and whether it was likely to impact the findings. Impact was explored through sensitivity analyses. Please see Sensitivity analysis.
Measures of treatment effect
Dichotomous data
We calculated risk ratios (RRs) with 95% confidence intervals (CIs) for dichotomous data (binary outcomes). These included the following.
Primary outcomes
-
Overall mortality: We used longest follow‐up data from each trial regardless of the period of follow‐up.
Secondary outcomes
-
Complications probably related to the intervention (e.g. thrombotic episodes (pulmonary embolism, myocardial infarction, disseminated intravascular coagulation), major immunological and allergic reactions (TRALI), cardiopulmonary overload (TACO), infection, sepsis (e.g. transmission of viral infection)).
-
Complications during inpatient stay not specific to trial intervention (e.g. pneumonia, congestive cardiac failure, respiratory failure, renal failure).
Continuous data
We used mean differences (MDs) if data were continuous and were measured in the same way between trials. We used standardized mean differences (SMDs) to combine trials that measured the same outcome using different scales. These included the following.
-
Number of new occurrences of blood transfusions (e.g. avoidance of transfusion) and quantity of blood products transfused.
-
Number of days in hospital.
-
Mean length of stay in ICU.
Unit of analysis issues
Cross‐over trials
We excluded cross‐over trials from our meta‐analyses because of the potential risk for 'carry‐over' of treatment effect in the context of bleeding.
Studies with multiple intervention groups
In studies designed with multiple intervention groups, we combined groups to create a single pair‐wise comparison (Higgins 2011). In trials with two or more PCC groups receiving different doses, we combined data, when possible, for primary and secondary outcomes.
Dealing with missing data
We contacted the authors of trials with missing data to retrieve the relevant information. For all included studies, we noted levels of attrition and any exclusions. We conducted a sensitivity analysis to explore the impact of included studies with high levels of missing data. In cases of missing data, we chose ’complete‐case analysis’ for our primary outcome, which excludes from the analysis all participants with missing outcomes. Selective outcome reporting occurs when non‐significant results are selectively withheld from publication (Chan 2004); it is defined as selection, on the basis of results, of a subset of the original variables recorded for inclusion in publication of trials (Hutton 2000). The most important types of selective outcome reporting are selective omission of outcomes from reports; selective choice of data for an outcome; selective reporting of different analyses using the same data; selective reporting of subsets of the data; and selective under‐reporting of data (Higgins 2011). Statistical methods devised to detect within‐study selective reporting are still in their infancy. We tried to explore for selective outcome reporting by comparing publications with their protocols, if the latter were available.
Assessment of heterogeneity
We explored heterogeneity by using the I² statistic and the Chi² test. An I² statistic above 50% represents substantial heterogeneity (Higgins 2003). In cases of I² > 0 (mortality outcome), we tried to determine the cause of heterogeneity by performing relevant subgroup analyses. We used the Chi² test to provide an indication of heterogeneity between studies, with P value ≤ 0.1 considered significant.
Assessment of reporting biases
Publication bias arises when dissemination of research findings is influenced by the nature and direction of results (Higgins 2011). We planned to explore the level of publication bias related to the trials included in this review by generating a funnel plot. As only a small number of studies were eligible for inclusion in this review, this step was not performed as suggested by the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).
Funding bias is related to possible publication delay or discouragement of undesired results in trials sponsored by the industry (Higgins 2011). To explore the role of funding, we planned to conduct a sensitivity analysis based on our primary endpoint. However, because of lack of data, this approach was not applied.
Data synthesis
We used Review Manager 5 software (RevMan 5.3) to perform meta‐analyses on pre‐stated outcomes from the included trials. If we performed the meta‐analyses and found I² = 0, we reported only results from the fixed‐effect model; in cases of I² > 0, we reported only results from the random‐effects model unless one or two trials contributed more than 60% of the total evidence provided, in which case the random‐effects model may have been biased. We believed that little value was derived by using a fixed‐effect model in cases of substantial heterogeneity, which we expected would be due to the various factors leading to massive bleeding. We pooled studies only in cases of low clinical heterogeneity. When meta‐analysis is used to combine results from several studies with binary outcomes (i.e. event or no event), adverse side effects may be rare but serious, and hence important (Sutton 2002). Most meta‐analytical software does not include trials with ’zero events’ in both arms (intervention vs control) when calculating a risk ratio (RR). Exempting these trials from calculation of an RR and 95% confidence interval (CI) may lead to overestimation of a treatment effect. The Cochrane Collaboration recommends application of the Peto odds ratio (OR), which is the best method of estimating odds ratios when many trials are found with no events in one or both arms (Higgins 2011). However, the Peto method generally is less useful when trials are small, or when treatment effects are large. We planned to conduct a sensitivity analysis by applying the Peto OR if this appeared to be a valid option. However, the trials included did not fulfil criteria for the Peto OR (Effects of interventions).
Trial sequential analysis
Meta‐analyses may result in type 1 errors caused by sparse data and repeated significance testing following updates with new trials (Brok 2009; Thorlund 2009; Wetterslev 2008; Wetterslev 2009). Systematic errors from trials with high risk of bias, outcome reporting bias, publication bias, early stopping for benefit and small trial bias may result in spurious P values.
In a single trial, interim analysis increases the risk of type 1 errors. To avoid type 1 errors, group sequential monitoring boundaries (Lan 1983) are applied to decide whether a trial could be terminated early because of a sufficiently small P value, that is, the cumulative Z curve crosses the monitoring boundary. Sequential monitoring boundaries can be applied to meta‐analyses as well and are called 'trial sequential monitoring boundaries'. In 'trial sequential analysis' (TSA), the addition of each trial in a cumulative meta‐analysis is regarded as an interim meta‐analysis, which helps to show whether additional trials are needed (Wetterslev 2008). It seems appropriate to adjust new meta‐analyses for multiple testing on accumulating data to control the overall type 1 error risk in cumulative meta‐analysis (Pogue 1997; Pogue 1998; Thorlund 2009; Wetterslev 2008).
The idea in TSA is that if the cumulative Z curve crosses the boundary, a sufficient level of evidence is reached, and no further trials may be needed. However, evidence is insufficient to allow a conclusion if the Z curve does not cross the boundary or does not surpass the required information size. To construct the trial sequential monitoring boundaries (TSMB), the required information size is needed and will be calculated as the least number of participants needed in a well‐powered single trial (Brok 2008; Pogue 1998; Wetterslev 2008). We adjusted the required information size for heterogeneity with the diversity adjustment factor (Wetterslev 2009). We applied TSA, as it prevents an increase in the risk of type 1 errors (< 5%) due to potential multiple updating and testing on accumulating data, whenever new trial results are included in a cumulative meta‐analysis (Pogue 1997; Pogue 1998). This provided important information for estimation of the level of evidence on the experimental intervention (Pogue 1997; Pogue 1998; Thorlund 2009) and the need for additional trials and their required sample size (Wetterslev 2008; Wetterslev 2009). We applied TSMB according to an information size suggested by Wetterslev 2008 and Wetterslev 2009 and an a priori 20% relative risk reduction (RRR) of serious adverse events using a control event proportion suggested by the pooled estimate of the event proportion in included trial control groups. We conducted TSA at least on primary outcomes (Brok 2009; Pogue 1997; Pogue 1998; Thorlund 2009; Wetterslev 2008), and on secondary outcomes if the accrued information size was an acceptable fraction of the estimated required information size to allow meaningful analyses (> 20%). If the actual accrued information size was too small, we provided the required information size given the actual diversity (Wetterslev 2009)
Subgroup analysis and investigation of heterogeneity
We planned the following subgroup analyses.
-
Benefits and harms of prothrombin complex concentrate (PCC) versus fresh frozen plasma (FFP) without other haemostatic agents.
-
Benefits and harms of PCC versus FFP and cryoprecipitate.
-
Benefits and harms of PCC versus other haemostatic agents (rVIIa, antifibrinolytics, desmopressin, plasma derivatives or other factor‐substitution products).
-
Benefits and harms of PCC versus FFP in combination with other haemostatic agents.
-
Benefits and harms of PCC in combination with other haemostatic agents versus 'standard treatment'.
-
Benefits and harms of PCC in trials investigating the emergency surgery population (defined as those in whom surgery was performed within 24 hours after meeting the indication for surgery) requiring acute surgical intervention due to bleeding complications versus trials investigating the emergency surgery population with no bleeding complications requiring acute surgical intervention due to co‐morbidity.
-
Benefits and harms of PCC in trials investigating the trauma population versus trials investigating the non‐trauma population.
-
Benefits and harms of PCC in trials investigating the neurosurgical population versus trials investigating the non‐neurosurgical population.
-
Benefits and harms of PCC in trials investigating the cardiac surgery population versus trials investigating the non‐cardiac surgery population.
-
Benefits and harms of PCC in trials investigating the paediatric population (age younger than 18 years, neonates not included) versus trials investigating the adult population.
-
Benefits and harms of PCC when the pooled intervention effect in trials with a dose regimen higher than the median dose of administered PCC was compared with that in trials with a dose regimen equal to or smaller than the median dose. This comparison was performed to detect possible dependency of the estimate of intervention effect on dose regimen. In cases of considerable between‐trial heterogeneity, we applied meta‐regression.
-
Efficacy and safety of three‐factor PCC versus four‐factor PCC.
If analyses of various subgroups with binary data were significant, we planned to perform a test of interaction by applying the fixed inverse variance method incorporated in RevMan 5.3. Alternatively, we planned to apply meta‐regression if a fixed‐effect model was not considered sensible because between‐study variability was considerable. We considered P value < 0.05 as indicating significant interaction between the PCC effect on mortality and subgroup category (Higgins 2011, Chapters 9.6.1 and 9.7). We also considered applying Q‐partitioning for interaction and subgroup differences if appropriate (RevMan 5.3).
We did not perform subgroups analysis because data were insufficient. See Differences between protocol and review.
Sensitivity analysis
We planned to conduct the following sensitivity analyses.
-
Comparison of estimates of the pooled intervention effect in trials with low risk of bias versus estimates from trials with high risk of bias (i.e. trials having at least one inadequate risk of bias component).
-
Comparison of estimates of the pooled intervention effect in trials based on different components of risk of bias (random sequence generation, allocation concealment, blinding, completeness of outcome data, selective reporting and 'other bias').
-
Comparison of estimates of the pooled intervention effect in trials with high levels of missing data. In cases of missing data, we applied ’complete‐case analysis’ for primary and secondary outcomes, thereby excluding from the analysis all participants with missing outcome data.
-
Examination of the role of funding bias by excluding trials that are sponsored exclusively by pharmaceutical and medical devices companies.
-
Comparison of estimates of pooled intervention effects performed by excluding data from studies published only as abstracts.
-
Examination of the importance of thromboembolic events when participants with high prior risk of thrombotic events were compared with others.
We calculated RRs with 95% CIs and decided to apply 'complete‐case analysis', if possible, for sensitivity and subgroup analyses based on our primary outcome measure (mortality).
We did not perform sensitivity analysis because data were insufficient. See Differences between protocol and review.
Summary of findings tables
We used the principles of the GRADE (Grades of Recommendation, Assessment, Development and Evaluation) system (Guyatt 2008; GradePro) to assess the quality of the body of evidence associated with specific outcomes (overall mortality; number of new occurrences of allogenic blood transfusion; numbers of adverse events and complications probably related to the intervention) (summary of findings Table for the main comparison).
Results
Description of studies
See Characteristics of included studies; Characteristics of excluded studies; Characteristics of studies awaiting classification; and Characteristics of ongoing studies.
Results of the search
A systematic search of electronic databases combined with a handsearch of references resulted in identification of 1250 publications (Figure 1). Animal studies, reviews, duplicates and trials with irrelevant endpoints or populations resulted in exclusion of 1240 publications. On the basis of title and abstract, 10 publications were included for a full paper review. We identified six ongoing studies but were unable to retrieve any data from the investigators (Ahonen 2013; Frenzel 2008; Innerhofer 2012; Ranucci 2011; Roy 2013; Steiner 2009). We included four RCTs with a total of 453 participants (Boulis 1999; Demeyere 2010; Majeed 2013; Sarode 2013).
PRISMA. We reran the search in October 2014. We found 1 study of interest (Kerebel 2013). This study was added to a list of ‘Studies awaiting classification' and will be incorporated into formal review findings at the time of the review update .
We reran the search in October 2014 and identified an additional 390 studies and found one study of interest (Kerebel 2013). We will add potential new studies of interest to a list of Studies awaiting classification, and we will incorporate these studies into formal review findings at the time of the review update.
Included studies
See Characteristics of included studies; Appendix 8.
We included four RCTs published from 1999 to 2013 to investigate the potential role of PCC in the VKA‐treated participant with concomitant bleeding and/or need for acute surgical intervention. Sample sizes varied from 21 to 216 adult participants, and trials were categorized as surgical or non‐surgical. All trials assessed the use of weight‐based and INR‐based PCC administration with infusion of FFP. Three trials (Boulis 1999; Majeed 2013; Sarode 2013) investigated the potential of replacing FFP with PCC in the acutely bleeding participant. The fourth study investigated the preemptive replacement of FFP with PCC in the setting of semi urgent cardiac surgery (Demeyere 2010). In all trials but one (Demeyere 2010), investigators co‐administered vitamin K to both intervention and control groups.
Boulis 1999 assessed the effect of PCC administration in decreasing INR in 21 participants with intracranial bleeding potentially requiring acute surgical intervention.
Demeyere 2010 assessed the effect of preemptive PCC administration on numbers of postoperative bleedings, re‐operations and blood transfusions and on plasma concentrations of vitamin K‐dependent coagulation factors as additional outcomes in 40 participants.
Majeed 2013 assessed the effect of preemptive PCC administration in a mixed surgical population of 176 VKA‐treated participants requiring an acute surgical/invasive procedure with haemostatic efficacy during surgery and decreased INR, plasma factor levels and transfusion of red blood cells (RBCs) or whole blood.
Finally, Sarode 2013 assessed the effect of PCC administration on 216 VKA‐treated non‐surgical participants with elevated INR experiencing major bleeding. Additional outcomes included mortality, cause of death, length of hospital stay, length of stay in ICU, number and volume of blood transfusions, type of bleeding, haemostatic efficacy and plasma concentrations of vitamin K‐dependent coagulation factors at various time points (Sarode 2013).
Formulation of PCC (Table 1), dose of PCC and time of PCC administration differed between trials. The mean volume of FFP transfused in the trials by Boulis 1999, Demeyere 2010 and Sarode 2013 varied from 813.5 mL to 2712 mL in the FFP group and from 0 mL to 399 mL in the PCC group.
| PCC | Factor II | Factor VII | Factor IX | Factor X | Protein C | Protein S | Antitrombin | Heparin | Trial |
| Konyne | 38 IU/mL | 4 IU/mL | 25 IU/mL | 38 IU/mL | Not stated | Not stated | Not stated | Not stated | |
| Cofact | 14‐35 IU/mL | 7‐20 IU/mL | 25 IU/mL | 14‐35 IU/mL | 11‐39 IU/mL | 1‐8 IU/mL | ≤ 0.6 IU/mL | Not stated | |
| Beriplex (Kcentra) | 19‐40 IU/mL | 10‐25 IU/mL | 20‐31 IU/mL | 25‐51 IU/mL | 21‐41 IU/mL | 12‐34 IU/mL | 0.2‐1.5 IU/mL | 0.4‐2 IU/mL |
Product information sought (contact with study authors and Internet search) and applied when not listed in references. When "not stated", the information was not retrievable.
PCC = prothrombin complex concentrate.
Excluded studies
We excluded four studies (Aart 2006; Eerenberg 2011; Marlu 2012; Taberner 1976).
Aart 2006 is a dose response study that did not include other interventions for comparison. Taberner 1976 did not report on characteristics of participants included in the study (baseline characteristics, morbidity, reasons for anticoagulation), process of randomization, allocation, concealment or bias. Also, they did not report on any of our predefined outcomes. We were not able to ensure the inclusion of relevant participants, hence the study was excluded. Eerenberg 2011 and Marlu 2012 were ex vivo studies that included healthy individuals without bleeding or pending surgery. Researchers reported no relevant outcome measures.
(See Characteristics of excluded studies.)
Ongoing studies
We identified six ongoing trials (see Characteristics of ongoing studies).
-
Ahonen 2013: 40 participants randomly assigned to PCC + fibrinogen concentrate or fresh frozen plasma (FFP) in the treatment of postpartum haemorrhage. Primary outcome was blood loss.
-
Frenzel 2008: 200 participants under VKA therapy with the need for urgent surgery or invasive procedures randomly assigned to octaplex or FFP. Primary efficacy endpoint was correction of INR to < 1.5.
-
Innerhofer 2012: 200 participants experiencing trauma‐induced coagulopathy randomly assigned to coagulation factor concentrates or FFP. Primary outcome measure was multiple organ failure.
-
Ranucci 2011: 120 participants undergoing cardiac surgery randomly assigned to placebo (saline), fibrinogen concentrate or PCC. Primary outcome was avoidance of allogeneic blood product transfusion.
-
Roy 2013: 400 VKA‐treated participants with minor craniocerebral trauma admitted to hospital randomly assigned to receive "standard therapy" or preventive reversal of coagulopathy with PCC. Primary outcome measure was percentage of intracranial bleeding diagnosed on computed tomography (CT) scan.
-
Steiner 2009: 74 VKA‐treated participants with CT‐confirmed intracerebral bleeding randomly assigned to receive PCC or FFP. Primary outcome was INR ≤ 1.2 within three hours after the start of drug infusion.
Risk of bias in included studies
We could classify no trials as having overall ’low risk of bias’. We evaluated the overall quality of trials on the basis of major sources of bias (domains) as previously described. We present the various bias domains in the ’Risk of bias’ graph and in the ’Risk of bias’ summary (Figure 2; Figure 3).
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Allocation
Two trials provided sufficient information regarding generation of allocation sequence (Demeyere 2010; Sarode 2013). One trial provided allocation concealment (Demeyere 2010). Two trials were designed as open‐label (Majeed 2013; Sarode 2013) and provided no allocation concealment.
Blinding
No included trials reported blinding of participants. Two trials did not blind personnel involved (Boulis 1999; Sarode 2013). One trial did not describe blinding of personnel (Majeed 2013). Finally, one trial (Demeyere 2010) blinded the surgeons involved but did not describe blinding of anaesthesiologists who were in charge of study drug administration. Two trials did not blind outcome assessors (Boulis 1999; Demeyere 2010). One trial did not report on blinding of outcome assessors. Finally, one study provided relevant blinding of outcome assessors (Sarode 2013). Attempts to contact study authors to request further information were not successful.
Incomplete outcome data
All four trials excluded participants after randomization. Two trials excluded participants on the basis of parameters that were not in accordance with exclusion criteria (Boulis 1999; Demeyere 2010). Boulis 1999 excluded 38% of the initial population because of withdrawal of consent and omission in data collection. Demeyere 2010 excluded two participants because vitamin K was erroneously administered. One trial did not describe reasons for exclusion of participants (Majeed 2013), and another trial adequately described the reasons for exclusion (Sarode 2013).
Selective reporting
We were able to compare one trial with its trial registration at www.clinicaltrials.gov (Sarode 2013). The time frame of the co‐primary outcome was changed from one hour and four hours post infusion at trial registration to 24 hours post infusion as reported in the publication. Despite contacting the study authors, we were unable to obtain protocol information for the three remaining trials (Boulis 1999; Demeyere 2010; Majeed 2013). Demeyere 2010 concluded that use of PCC caused less bleeding, even when this was a non‐significant finding. With the exception of the latter, none of the included studies were suspected of selective reporting.
Other potential sources of bias
Funding bias
One study stated that investigators had no financial interest and no apparent link to the medical industry (Boulis 1999). The three remaining trials were supported in varying degrees by the manufacturer of the investigated drug (Demeyere 2010; Majeed 2013; Sarode 2013). Support primarily consisted of financial backing, employment of investigators by the manufacturer, access to free drugs or assistance in preparing and processing the scientific data.
Study design and early stopping
One trial reported a sample size calculation and provided clinically relevant outcomes (Sarode 2013). The three remaining trials provided no sample size calculation (Boulis 1999; Majeed 2013) or based the calculation on pragmatic considerations (Demeyere 2010). Primary outcomes for the three latter trials were based on surrogate parameters. One study (Boulis 1999) reported early stopping as the result of shortage of study drugs.
Baseline imbalance
Demeyere 2010 provided baseline parameters (weight, age and surgical data) with exploratory statistical analyses and reported no baseline differences. Sarode 2013 and Majeed 2013 also provided baseline parameters but without statistical explanatory support. Boulis 1999 concluded that no differences in demographic data, including participant weight, age and indications for warfarin anticoagulation, were present.
Effects of interventions
Primary outcome 1. Overall mortality longest follow‐up
We will use the longest follow‐up data from each trial regardless of the period of follow‐up. Three trials reported data on "Overall mortality ‐ longest follow up" (risk ratio (RR) 0.93, 95% confidence interval (CI) 0.37 to 2.33; participants = 421; trials = 4; I2 = 43%; very low quality of evidence). All‐cause mortality did not seem to be associated with the use of PCC (Figure 4; Boulis 1999; Majeed 2013; Sarode 2013). The small number of studies eligible for inclusion precluded subgroup and sensitivity analyses referring to the primary outcome.
Forest plot of comparison: 1 Mortality, outcome: 1.1 Overall mortality ‐ longest follow‐up.
We conducted trial sequential analysis of effects of PCC versus FFP on all‐cause mortality as reported by three trials (Boulis 1999; Majeed 2013; Sarode 2013) with a control event proportion of 8.4%, diversity of 44% and an anticipated intervention effect of 50% relative risk reduction. Despite the anticipated large intervention effect of 50% relative risk reduction, in this analysis, the cumulative Z‐curve does not break through any boundaries for benefit, harm or futility. The analysis is inconclusive because information was sparse.
Primary outcome 2. Overall 28‐day mortality
We intended to provide data as 30‐day mortality. None of the included studies provided data as 28‐day mortality, hence we were not able to further conduct subgroup or sensitivity analyses.
Secondary outcome 1. Number of new occurrences of blood transfusions (e.g. avoidance of transfusion) and quantity of blood products transfused
The number of new occurrences of blood transfusions (RBCs) was reported by two trials (Majeed 2013; Sarode 2013). The use of PCC did not appear to be associated with fewer new occurrences of RBC transfusions (RR 1.08, 95% CI 0.82 to 1.43; participants = 370; trials = 2; I2 = 0%; very low quality of evidence) (Analysis 2.1). We were unable to further investigate the true distribution of RBC units transfused per participant because data were insufficient (Demeyere 2010; Sarode 2013 ‐ we contacted the study authors but received no relevant reply; Table 2). We were able to obtain data regarding the volume of RBC transfused in two trials (Majeed 2013; Sarode 2013). The volume of RBCs transfused did not appear to be associated with the use of PCC (mean difference (MD) ‐1.85, 95% CI ‐85.89 to 82.20; I2 = 13%; very low quality of evidence) (Figure 5).
Forest plot of comparison: 3 Quantity of blood products transfused, outcome: 3.1 Transfusion of RBCs.
| Analysis | Study | Participants (PCC/FFP) | Notes | Findings | |
| Number of new occurrencesof RBC transfusion | 20/20 | 19 participants allocated to FFP group received 50 units of packed RBCs 16 participants in PCC group received 33 units of packed RBCs | True distribution of packed RBCs per participant unknown because of lack of data (study authors contacted but no reply; Included studies). Hence, trial data were not amenable for meta‐analysis | ||
| Transfusion of FFP | 6/11 | 6 participants allocated to PCC group received a mean of 399 mL (SD = 271 mL) FFP 11 participants allocated to FFP group received a mean of 2712 mL (SD = 346 mL) FFP | MD ‐2313 mL; 95% CI ‐2611.04 mL to ‐2014.96 mL; participants = 17; studies = 1 | ||
| 20/20 | None of the participants allocated to PCC group received FFP 20 participants allocated to FFP group received a mean of 1372 mL (SD = 320 mL) FFP | Secondary to study design and based on lack of events in PCC group, this study was not included in further analysis of FFP transfusion (Data synthesis; Results) | |||
| 103/108 | None of the participants allocated to PCC group received FFP 108 participants allocated to FFP group received a mean of 813.5 mL (SD = 187.5 mL) FFP | Secondary to study design and based on lack of events in PCC group, this study was not included in further analysis of FFP transfusion | |||
| Length of stay in hospital (days) | 98/104 | 98 participants in PCC group had mean length of stay in hospital of 4.5 days (SD = 0.55) 104 participants in FFP group had mean length of stay in hospital of 4.2 days (SD = 0.55) | MD = 0.30 days; 95% CI 0.15 days to 0.45 days | ||
| Length of stay in ICU (hours) | 98/104 | 98 participants in PCC group had mean stay of 0 hours in ICU (IQR 0‐44.7) 104 participants in FFP group had mean stay of 0 hours in the ICU (IQR 0‐40.0) | MD = 0.00 hours; 95% CI ‐0.08 hours to 0.08 hours | ||
| Number of complications during inpatient stay not specific to trial intervention | 103/109 | 103 participants in PCC group experienced 31 events 109 participants in FFP group experienced 24 events | RR = 1.37; 95% CI 0.86 to 2.16 | ||
| Amount of bleeding | 20/20 | No statistical differences in 1 hour, 4 hours and 24 hours postoperative chest tube drainage | 1 hour (FFP): mean, mL (SD) 97.0 (± 54.7) 1 hour (PCC): mean, mL (SD) 67.6 (± 46.8) 4 hours (FFP): mean, mL (SD) 163.5 (± 89.8) 4 hours (PCC): mean, mL (SD) 133.5 (± 73.4) 24 hours (FFP): mean, mL (SD) 471.7 (± 294.4) 24 hours (PCC): mean, mL (SD) 439.3 (± 247.3) |
CI = confidence interval.
FFP = fresh frozen plasma.
ICU = intensive care unit.
IQR = interquartile range.
MD = mean difference.
PCC = prothrombin complex concentrate.
RBC = red blood cell.
SD = standard deviation.
We conducted trial sequential analysis (TSA) of PCC versus FFP on quantity of RBCs transfused from two trials (Majeed 2013; Sarode 2013) with diversity of 18% and an anticipated mean difference in intervention effect of 125 mL (Figure 6). The cumulative Z‐curve breaks through the boundary for futility (non‐superiority). Therefore the analysis is able to reject an intervention effect of 125 mL in the light of sparse data and repetitive analyses. The TSA‐adjusted confidence interval for the intervention effect estimated in a random‐effects model of ‐1.83 mL is ‐123 mL to 120 mL.
Trial sequential analysis (TSA) of PCC vs FFP on quantity of RBCs transfused from 2 trials with diversity of 18% and an anticipated mean difference in intervention effect of 125 mL. The cumulative Z‐curve breaks through the boundary for futility (non‐superiority). The analysis therefore led to rejection of an intervention effect of 125 mL based on sparse data and repetitive analyses. The TSA adjusted confidence interval on the intervention effect estimated in a random‐effects model of ‐1.83 mL is ‐123 mL to 120 mL.
Three trials reported the number of new occurrences of FFP transfusion (Boulis 1999; Demeyere 2010; Sarode 2013). In the studies of Demeyere 2010 and Sarode 2013, transfusion of FFP was not protocolized in the intervention group. In the study by Boulis 1999, transfusion of FFP was administered to control and intervention groups. As a result of the apparent study design (FFP vs PCC), an obvious difference in the number of new occurrences of FFP transfusion was reported. Analysing data by applying the fixed‐effect model (Data synthesis) showed an apparent effect of PCC on the number of new occurrences of FFP transfusion (RR 0.06, 95% CI 0.03 to 0.12; participants = 264; trials = 3; I2 = 99%; Analysis 2.2). However, applying the random‐effects model showed no association between administration of PCC and the number of new occurrences of FFP transfusion (RR 0.05, 95% CI 0.00 to 270914.97; participants = 264; trials = 3; I2 = 99%). The high degree of heterogeneity and diversity in these results poses important questions regarding internal validity and indicates that data most likely are not appropriate for meta‐analysis. We planned to conduct a corrected analysis by applying the Peto OR if this was seen as a valid option (Data synthesis). However, only two trials (Demeyere 2010; Sarode 2013) included ’zero event’ in one group (PCC/intervention arm), and Demeyere 2010 was a small study with 20 participants in each group. The Peto method generally is less useful when trials are small, or when treatment effects are large. Hence we chose not to proceed with these exploratory analyses.
The volume of FFP transfused concomitantly among control and intervention groups differed in three trials (Boulis 1999, FFP arm: mean = 2712 mL; standard deviation (SD) = 346 mL; PCC arm: mean = 399 mL; SD = 271 mL; Demeyere 2010, FFP arm: mean = 1372 mL; SD = 320 mL; PCC arm: no transfusion of FFP; Sarode 2013, FFP arm: mean = 813.5 mL; SD = 187.5 mL; PCC arm: no transfusion of FFP. Majeed 2013 did not report on FFP transfusion requirements. We contacted the study authors, but they did not reply. As a result of the occurrence of two 'zero event' trials (Demeyere 2010; Sarode 2013), we were unable to perform meta‐analysis.
One trial reported on amount of bleeding. Demeyere 2010 reported chest tube drainage at one hour, four hours and 24 hours postoperatively, but researchers were not able to make an association between the use of PCC and the amount of bleeding (Table 2).
Secondary outcome 2. Complications probably related to the intervention (e.g. thrombotic episodes (pulmonary embolism, myocardial infarction, disseminated intravascular coagulation), heparin‐induced thrombocytopenia, major immunological and allergic reactions (TRALI), cardiopulmonary overload (TACO), infection and sepsis (e.g. transmission of viral infection))
All trials reported on complications probably related to trial intervention. We could not show an association between use of PCC and complications (adverse events) probably related to trial intervention (RR 0.92, 95% CI 0.78 to 1.09; I2 = 0%; very low quality of evidence) (Figure 7). One trial specifically reported data on complications not related to trial intervention (Sarode 2013). These primarily involved cardiac ischaemic events, respiratory failure and intracranial haemorrhage. No other trials stated information regarding complications not specific to trial intervention. Demeyere 2010 reported adverse events in 7/20 participants allocated to the PCC group and in 9/20 in the FFP group. Types of adverse events and potential correlation of adverse events to trial intervention were not further investigated by study authors (these authors were contacted but failed to reply). Majeed 2013 reported overall numbers of adverse events (PCC = 41/88; FFP = 44/88) and serious adverse events (PCC = 22/88; FFP = 23/88) but did not classify these events as related or not related to trial intervention.
Forest plot of comparison: 4 Adverse events, outcome: 4.1 Number of complications probably related to the intervention.
We conducted trial sequential analysis of PCC versus FFP on adverse events possibly related to the intervention from four trials with a control event proportion of 57%, diversity of 0% and an anticipated intervention effect of 20% relative risk reduction (RRR) (Figure 8). The cumulative Z‐curve breaks through the boundary for futility (non‐superiority). The analysis therefore is able to reject an intervention effect of 20% RRR, in the light of sparse data and repetitive analyses, but an intervention effect less than 20% RRR may be present.
Trial sequential analysis (TSA) of PCC vs FFP on adverse events from 4 trials with a control event proportion of 57%, diversity of 0% and an anticipated intervention effect of 20% relative risk reduction (RRR). The cumulative Z‐curve breaks through the boundary for futility (non‐superiority). The analysis therefore led to rejection of an intervention effect of 20% RRR based on sparse data and repetitive analyses, but the intervention effect may be less than 20% RRR.
As studies eligible for inclusion were few and provided a limited quantity of data along with considerable heterogeneity, we were not able to further assess this subgroup analysis.
Secondary outcome 3. Complications during inpatient stay not specific to the trial intervention (e.g. pneumonia, congestive cardiac failure, respiratory failure, renal failure)
Only one study (Sarode 2013) reported on complications during inpatient stay not specific to the trial intervention. As studies eligible for inclusion were few and provided a limited quantity of data, we were not able to further assess this subgroup analysis.
Secondary outcome 4. Number of days in hospital
Only one study (Sarode 2013) reported on "number of days in hospital". We assessed duration of stay in hospital (Table 2) and found no statistically significant effect. As studies eligible for inclusion were few and provided a limited quantity of data, we were not able to further assess this subgroup analysis.
Secondary outcome 5. Mean length of stay in the ICU
Only one study (Sarode 2013) reported on "number of days in hospital". We assessed duration of ICU stay (Table 2 ) and found no statistically significant effect. As studies eligible for inclusion were few and provided a limited quantity of data, we were not able to further assess this subgroup analysis. No trials reported data on "days free from ventilator".
Finally, none of the trials provided data on quality of life assessment or cost‐benefit analyses.
Discussion
This systematic review included four randomized controlled trials comprising a total of 453 medical and surgical vitamin K antagonist (VKA)‐treated participants undergoing reversal therapy with prothrombin complex concentrate (PCC), fresh frozen plasma (FFP) or both. Mortality may be contested by many as the choice of primary outcome, but it summarizes ultimate harms and benefits simultaneously. Three studies provided data on mortality (Boulis 1999; Majeed 2013; Sarode 2013). None of these studies were powered to detect a difference in mortality. Meta‐analysis showed no statistical benefit of PCC treatment versus FFP (RR 0.93, 95% CI 0.37 to 2.33). We identified use of a medical treatment (PCC) that has been known for many years and that lately has found its way to European and American guidelines for acute reversal of acute life‐threatening bleeding related to elevated international normalized ratio (INR) (Ansell 2008; Baglin 2006). Still, it must be kept in mind that a high level of heterogeneity, bias, differences in study design (87% of the participants included in this review originate from two RCTs designed as non‐inferiority studies) and use of surrogate outcome measures (e.g. INR values and time to INR reversal) contribute to overall low clinical evidence and hence applicability.
However, it is important to stress the fact that use of PCC seems to enable reversal of vitamin K‐induced coagulopathy without the need for transfusion of FFP. Demeyere 2010 showed that the comparator FFP arm required considerably more therapeutic measures to reach the target INR value (20/20 participants) compared with 6/20 participants in the PCC arm. Specifically, the FFP arm required a total of 19.4 L FFP additional to the 8 L intended in the protocol (400 mL × 20 participants). Furthermore, intermittent use of PCC was needed to achieve the target INR value in the FFP arm. In addition, Boulis 1999 showed that introducing the use of PCC reduced the mean volume of FFP transfused from 2712 mL (SD = 346 mL) to 399 mL (SD = 271 mL), and Sarode 2013 showed that total avoidance of FFP was possible for urgent reversal of VKA therapy in major bleeding events, as demonstrated by clinical assessments of bleeding and laboratory measurements of INR and factor levels.
Assessment of safety is not uniform and raises important questions regarding underreporting. Both Demeyere 2010 and Majeed 2013 stated the number of adverse events and classified them as serious or non‐serious. No further elaboration of type of adverse event or possible correlation with medical treatment was investigated. We were unable to associate use of PCC with the number of complications probably related to the intervention (RR 0.92, 95% CI 0.78 to 1.09). Both Boulis 1999 and Sarode 2013 stated the occurrence of "fluid overload" as the most frequent adverse event in the FFP‐treated group. Still, lack of definition and possible correlation with underlying patient co‐morbidity remain to be further investigated. The number of new occurrences of blood transfusions (RBCs) did not seem to be associated with use of PCC (RR 1.08, 95% CI 0.82 to 1.43)
Thus, this review summarizes an area in which the evidence differs widely from clinical practice, and the information provided here should indeed ultimately guide the direction of future research. The quality of evidence for our primary and secondary outcomes was very low. Two of four included studies had high or unclear risk of selection bias and high risk of attrition bias; 50% of included studies had high risk of performance bias and 75% had high risk of detection bias. Most of the included studies had unclear risk of reporting bias. More randomized clinical trials with clinically relevant endpoints and a systematic definition of adverse events are urgently needed.
Summary of main results
Our systematic review comprised a total of 453 participants included in four "high risk of bias" trials performed in a widespread clinical array with a high degree of both clinical and statistical heterogeneity. No statistically significant effect on mortality or number of serious adverse events (e.g. life‐threatening thrombotic events such as acute myocardial infarction, pulmonary embolism or cerebral venous thromboembolism) was observed. Prothrombin complex concentrate appears to enable reversal of vitamin K‐induced coagulopathy without the need for transfusion of fresh frozen plasma and without generating a higher rate of adverse events.
Overall completeness and applicability of evidence
It must be kept in mind that all of the four included trials were small (imprecision of results) and aimed primarily to investigate surrogate outcomes (e.g. time to reverse international normalized ratio (INR), indirectness). Additionally, all trials were characterized by a high degree of heterogeneity in study design and clinical setup. Currently, evidence supporting the use of PCC in bleeding patients is weak. Discussions related to prothrombin complex concentrate treatment of the vitamin K antagonist‐treated patient experiencing acute life‐threatening bleeding or with a need for urgent surgery are often complicated by lack of evidence. A high level of heterogeneity, use of a non‐standardized product (three‐factor PCC vs four‐factor PCC; contents of naturally occurring anticoagulants such as proteins C and S), lack of definition of relevant clinical outcomes (e.g. adverse events) and conflicts of interest often complicate interpretation of results.
Quality of the evidence
Randomized controlled trials (RCTs) are considered the gold standard for evaluating clinical effects and new treatments. Prothrombin complex concentrate has been used in different forms (e.g. three‐factor and four‐factor concentrates) since 1976, and still no well‐designed and powered RCT supports the general applicability of its use. The body of RCTs currently available provide only evidence of low quality. Reasons for downgrading the evidence include risk of bias in included studies, wide confidence intervals and relatively few participants overall.
Potential biases in the review process
The strength of this review lies in the specification that all included studies must have a comparator group. Still, this excludes most of the studies found by a well‐constructed search strategy that minimizes the chance of missing randomized controlled trials not fulfilling inclusion criteria. Authors of this review have no conflicts of interest. We ensured that language would not introduce bias by imposing no limitations on such. With regards to safety outcomes, reporting was not uniform and raises concerns of underreporting. We reran the search strategy in October 2014 and found one study of interest. This study was added to a list of ‘Studies awaiting classification' and will be incorporated into formal review findings at the time of the review update.
Agreements and disagreements with other studies or reviews
The overall conclusions of this review do not contradict the statements and findings of previous reviews and meta‐analysis (Dentali 2011; Lin 2013; Sørensen 2011, Vang 2011). A recent systematic review (Lin 2013) on the use of coagulation factors such as fibrinogen and prothrombin complex concentrate (PCC) concludes that too often surrogate primary endpoints (laboratory parameters) are used instead of preferred patient‐related clinical outcomes (need for allogenic blood transfusion and postoperative chest tube drainage). Also, the authors of this review found the existing body of literature to be characterized by a high degree of bias, underpowered and unable to evaluate safety outcomes, as reporting of adverse events is not uniform and raises concerns of underreporting. Dentali 2011 performed a meta‐analysis of 27 studies evaluating the safety of PCCs for rapid reversal of vitamin K antagonist treatment. Investigators concluded that a low but quantifiable risk of thromboembolism in VKA‐treated participants receiving PCCs for anticoagulation reversal exists, and that these findings should be confirmed in randomized controlled trials. In accordance with these findings, Sørensen 2011 and Vang 2011 (non‐systematic reviews) emphasize the importance of patient risk stratification in relation to possible ongoing bleeding, indications for VKA therapy, risk of thromboembolic events and type of surgery.
PRISMA. We reran the search in October 2014. We found 1 study of interest (Kerebel 2013). This study was added to a list of ‘Studies awaiting classification' and will be incorporated into formal review findings at the time of the review update .
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Forest plot of comparison: 1 Mortality, outcome: 1.1 Overall mortality ‐ longest follow‐up.
Forest plot of comparison: 3 Quantity of blood products transfused, outcome: 3.1 Transfusion of RBCs.
Trial sequential analysis (TSA) of PCC vs FFP on quantity of RBCs transfused from 2 trials with diversity of 18% and an anticipated mean difference in intervention effect of 125 mL. The cumulative Z‐curve breaks through the boundary for futility (non‐superiority). The analysis therefore led to rejection of an intervention effect of 125 mL based on sparse data and repetitive analyses. The TSA adjusted confidence interval on the intervention effect estimated in a random‐effects model of ‐1.83 mL is ‐123 mL to 120 mL.
Forest plot of comparison: 4 Adverse events, outcome: 4.1 Number of complications probably related to the intervention.
Trial sequential analysis (TSA) of PCC vs FFP on adverse events from 4 trials with a control event proportion of 57%, diversity of 0% and an anticipated intervention effect of 20% relative risk reduction (RRR). The cumulative Z‐curve breaks through the boundary for futility (non‐superiority). The analysis therefore led to rejection of an intervention effect of 20% RRR based on sparse data and repetitive analyses, but the intervention effect may be less than 20% RRR.
Comparison 1 Mortality, Outcome 1 Overall mortality ‐ longest follow‐up.
Comparison 2 Number of new occurrences of blood transfusion, Outcome 1 Number of new occurrences of RBC transfusion.
Comparison 2 Number of new occurrences of blood transfusion, Outcome 2 Number of new occurrences of fresh frozen plasma transfusion.
Comparison 3 Quantity of blood products transfused, Outcome 1 Transfusion of RBCs.
Comparison 4 Adverse events, Outcome 1 Number of complications probably related to the intervention.
| Prothrombin complex concentrate versus any comparator for reversal of vitamin K antagonist treatment in bleeding and non‐bleeding patients | ||||||
| Patient or population: patients with vitamin K antagonist‐induced bleeding requiring acute intervention | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | Number of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Control | Prothrombin complex concentrate | |||||
| Overall mortality ‐ longest follow‐up | Study population | RR 0.93 | 421 | ⊕⊝⊝⊝ | Effect of the use of PCC on mortality was uncertain. None of the included studies were powered to detect differences in mortality. Large confidence intervals. Few participants and few events | |
| 84 per 1000 | 78 per 1000 | |||||
| Moderate | ||||||
| 91 per 1000 | 85 per 1000 | |||||
| Number of new occurrences of RBC transfusion | Study population | RR 1.08 | 376 | ⊕⊝⊝⊝ | Number of new occurrences of blood transfusion did not seem to be associated with use of PCC. Large confidence intervals. Few participants and few events | |
| 319 per 1000 | 344 per 1000 | |||||
| Moderate | ||||||
| 300 per 1000 | 324 per 1000 | |||||
| Number of complications probably related to the intervention | Study population | RR 0.92 | 442 | ⊕⊝⊝⊝ | Assessment of safety was not uniform, which raises concerns of underreporting. Large confidence intervals. Few participants and few events | |
| 573 per 1000 | 527 per 1000 | |||||
| Moderate | ||||||
| 563 per 1000 | 518 per 1000 | |||||
| Transfusion of RBCs | Mean transfusion of RBCs in the intervention groups was | 370 | ⊕⊝⊝⊝ | Trial sequential analysis of PCC vs FFP on quantity of RBCs transfused from 2 trials led to rejection of an intervention effect of 125 mL based on sparse data and repetitive analyses. Large confidence intervals. Few participants and few events | ||
| *The basis for the assumed risk (e.g. median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence. The quality of the evidence is downgraded 3 levels based on risk of bias in the included studies and the relatively small overall number of participants. | ||||||
| aOnly poster and abstract published, open‐label study. Industry sponsored. Industry funded, small sample size. Describes several breaches of the study protocol. dImprecision of results (large confidence intervals). Indirectnes of evidence (few participants and few events). CI = confidence interval. FFP = fresh frozen plasma. PCC = prothrombin complex concentrate. RBC = red blood cell. RR = risk ratio. | ||||||
| PCC | Factor II | Factor VII | Factor IX | Factor X | Protein C | Protein S | Antitrombin | Heparin | Trial |
| Konyne | 38 IU/mL | 4 IU/mL | 25 IU/mL | 38 IU/mL | Not stated | Not stated | Not stated | Not stated | |
| Cofact | 14‐35 IU/mL | 7‐20 IU/mL | 25 IU/mL | 14‐35 IU/mL | 11‐39 IU/mL | 1‐8 IU/mL | ≤ 0.6 IU/mL | Not stated | |
| Beriplex (Kcentra) | 19‐40 IU/mL | 10‐25 IU/mL | 20‐31 IU/mL | 25‐51 IU/mL | 21‐41 IU/mL | 12‐34 IU/mL | 0.2‐1.5 IU/mL | 0.4‐2 IU/mL | |
| Product information sought (contact with study authors and Internet search) and applied when not listed in references. When "not stated", the information was not retrievable. PCC = prothrombin complex concentrate. | |||||||||
| Analysis | Study | Participants (PCC/FFP) | Notes | Findings | |
| Number of new occurrencesof RBC transfusion | 20/20 | 19 participants allocated to FFP group received 50 units of packed RBCs 16 participants in PCC group received 33 units of packed RBCs | True distribution of packed RBCs per participant unknown because of lack of data (study authors contacted but no reply; Included studies). Hence, trial data were not amenable for meta‐analysis | ||
| Transfusion of FFP | 6/11 | 6 participants allocated to PCC group received a mean of 399 mL (SD = 271 mL) FFP 11 participants allocated to FFP group received a mean of 2712 mL (SD = 346 mL) FFP | MD ‐2313 mL; 95% CI ‐2611.04 mL to ‐2014.96 mL; participants = 17; studies = 1 | ||
| 20/20 | None of the participants allocated to PCC group received FFP 20 participants allocated to FFP group received a mean of 1372 mL (SD = 320 mL) FFP | Secondary to study design and based on lack of events in PCC group, this study was not included in further analysis of FFP transfusion (Data synthesis; Results) | |||
| 103/108 | None of the participants allocated to PCC group received FFP 108 participants allocated to FFP group received a mean of 813.5 mL (SD = 187.5 mL) FFP | Secondary to study design and based on lack of events in PCC group, this study was not included in further analysis of FFP transfusion | |||
| Length of stay in hospital (days) | 98/104 | 98 participants in PCC group had mean length of stay in hospital of 4.5 days (SD = 0.55) 104 participants in FFP group had mean length of stay in hospital of 4.2 days (SD = 0.55) | MD = 0.30 days; 95% CI 0.15 days to 0.45 days | ||
| Length of stay in ICU (hours) | 98/104 | 98 participants in PCC group had mean stay of 0 hours in ICU (IQR 0‐44.7) 104 participants in FFP group had mean stay of 0 hours in the ICU (IQR 0‐40.0) | MD = 0.00 hours; 95% CI ‐0.08 hours to 0.08 hours | ||
| Number of complications during inpatient stay not specific to trial intervention | 103/109 | 103 participants in PCC group experienced 31 events 109 participants in FFP group experienced 24 events | RR = 1.37; 95% CI 0.86 to 2.16 | ||
| Amount of bleeding | 20/20 | No statistical differences in 1 hour, 4 hours and 24 hours postoperative chest tube drainage | 1 hour (FFP): mean, mL (SD) 97.0 (± 54.7) 1 hour (PCC): mean, mL (SD) 67.6 (± 46.8) 4 hours (FFP): mean, mL (SD) 163.5 (± 89.8) 4 hours (PCC): mean, mL (SD) 133.5 (± 73.4) 24 hours (FFP): mean, mL (SD) 471.7 (± 294.4) 24 hours (PCC): mean, mL (SD) 439.3 (± 247.3) | ||
| CI = confidence interval. FFP = fresh frozen plasma. ICU = intensive care unit. IQR = interquartile range. MD = mean difference. PCC = prothrombin complex concentrate. RBC = red blood cell. SD = standard deviation. | |||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Overall mortality ‐ longest follow‐up Show forest plot | 3 | 421 | Risk Ratio (M‐H, Random, 95% CI) | 0.93 [0.37, 2.33] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Number of new occurrences of RBC transfusion Show forest plot | 2 | 370 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.08 [0.82, 1.43] |
| 2 Number of new occurrences of fresh frozen plasma transfusion Show forest plot | 3 | 264 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.06 [0.03, 0.12] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Transfusion of RBCs Show forest plot | 2 | 370 | Mean Difference (IV, Random, 95% CI) | ‐1.85 [‐85.89, 82.20] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Number of complications probably related to the intervention Show forest plot | 4 | 442 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.92 [0.78, 1.09] |
