Detecting, managing and monitoring haemostasis: viscoelastometric point‑of‑care testing (ROTEM, TEG and Sonoclot systems)

Searches were carried out to identify cost‑effectiveness studies of viscoelastometric point‑of‑care testing. The searches identified 331 records of which 5 studies fulfilled the inclusion criteria. Two were only available as conference abstracts; 3 were conducted in cardiac patients, 1 in patients having liver transplant and 1 in both cardiac and liver transplant patients.

One study was a formal cost‑effectiveness analysis of viscoelastometric devices in cardiac and liver transplant patients. This study was conducted for the Scottish NHS.

The other 4 studies were cost‑minimisation studies performed alongside a retrospective before and after study. All 4 studies compared the volume and costs of blood transfused before and after the introduction of a viscoelastometric device. Three studies evaluated ROTEM and 1 evaluated TEG. All 4 found that costs were reduced as a result of the introduction of a viscoelastometric device. Only 1 of the 4 studies reported a detailed breakdown of cost savings. It showed that, after the introduction of ROTEM, the cumulative average monthly costs of all blood products decreased by 32%).

For both the cardiac and trauma models, the external assessment group adopted the model structure used in the health technology assessment carried out for NHS Scotland in 2008. This was largely based on a cost‑effectiveness study of cell salvage and alternative methods of minimising perioperative allogeneic blood transfusion by Davies et al. (2006). However, the external assessment group used more recent data sources whenever possible to update the input parameters of the model.

The models were based on a decision tree that started with the choice of strategy to be followed, that is, viscoelastometric device (ROTEM, TEG or Sonoclot) or standard laboratory tests. Within each strategy, patients then either did or did not receive a transfusion. Transfusion, when it occurs, may be associated with adverse events or complications. Complications were categorised as being related to surgery and/or transfusion, or related to transfusion alone.

Complications related to surgery and/or transfusions included in the model were: renal dysfunction, myocardial infarction, stroke, thrombosis, excessive bleeding needing reoperation, wound complications and septicaemia. Transfusion‑related complications included transfusion‑associated graft‑versus‑host disease, complications related to the administration of an incorrect blood component, haemolytic transfusion reactions (acute or delayed), post‑transfusion purpura, transfusion‑related acute lung injury and febrile reaction.

In addition, the external assessment group assumed that patients may also experience transfusion‑transmitted infections. Transfusion‑transmitted infections include bacterial contamination, variant Creutzfeldt–Jakob disease, malaria, human T‑cell lymphotropic virus, HIV and hepatitis A, B and C.

The models' time horizons were set to 1 month and 1 year because the benefits of a reduction in red blood cell transfusion were considered to have occurred within this time frame. At 1 month, the models reflect the period of hospitalisation and accordingly capture the impact of complications related to surgery and blood loss, transfusion‑related complications and infection caused by bacterial contamination. It should be noted that, as in Davies et al. (2006), bacterial contamination is the only transfusion‑transmitted infection that was assumed to occur during the hospitalisation period. For other transfusion‑transmitted infections included in the models, a time horizon of 1 year was considered more appropriate, because these infections do not usually appear immediately.

Costs were estimated from the perspective of the NHS and personal social services. Consequences were expressed in life years gained and quality‑adjusted life years (QALYs). QALY weights (utilities) were assigned to adverse events to express their consequences. Discounting was not necessary since the longest time horizon was set at 1 year.

For the cardiac model, the baseline risk of having a transfusion was estimated based on the number of red blood cell transfusions in the standard laboratory tests group in 4 of the cardiac surgery trials included in the effectiveness review. Since the effectiveness review did not find evidence of a difference in the relative risk of red blood cell transfusion between studies that assessed ROTEM and those that assessed TEG, the external assessment group applied the summary relative risk for red blood cell transfusion estimated for all studies for the ROTEM and TEG models. Limited data suggested that the accuracy of Sonoclot in predicting clinical outcomes may be similar to that of TEG. The external assessment group therefore assumed that this summary relative risk could be applied in the Sonoclot model. A beta and a normal distribution were assigned for the probabilistic sensitivity analyses.

For the trauma model, the baseline risk of red blood cell transfusion for the standard laboratory tests group was also estimated using data from those studies that reported data on the proportion of patients who received red blood cell transfusion. A random effects model was used to estimate the mean proportion of patients who received red blood cell transfusion. As there were no data comparing the proportion of transfused patients in a trauma population who received viscoelastometric testing with those who received standard laboratory tests, the external assessment group applied the same relative risk as in the cardiac surgery population.

Reoperation to investigate bleeding is the only complication among those included in the model that was evaluated by the RCTs included in the effectiveness review. Data on the other complications were limited so the external assessment group assumed that there was no difference in the direct risk of having a complication between those tested with viscoelastometric devices and those tested with standard laboratory tests. The same assumption was made in Davies et al. (2006). The risk of complications in each testing strategy was influenced indirectly by the different red blood cell transfusion rates associated with each strategy. The probability of experiencing septicaemia was obtained from 1 study. However, the population in this study was not representative of the population in the assessment because it only included patients who received 4 or more units of red blood cell within 1 day of surgery (that is, patients with massive bleed). The external assessment group judged this estimate to be too high and reduced the estimate by an arbitrary factor of 0.5.

For the trauma model, 2 complications were included: acute respiratory distress syndrome and multiple organ failure. Estimates for the incidence of acute respiratory distress syndrome were obtained from a study of 14,070 trauma patients conducted in the USA, which reported an overall incidence of 4.6%. The same study was used to calculate the proportion of patients with acute respiratory distress syndrome among those who received a transfusion as 15.5%. For multiple organ failure, no studies were found that either provided estimates or allowed direct calculation of incidence for those transfused. However, based on the overall incidence of multiple organ failure being 3 to 5 times higher than that of acute respiratory distress syndrome, the external assessment group considered that a multiple organ failure incidence rate of 30% was a realistic assumption.

The trials included in the clinical effectiveness review did not report data on transfusion‑related complications. Therefore, data on the probabilities of experiencing transfusion‑related complications were based on a UK Serious Hazards of Transfusion report, which collects and analyses anonymised information on adverse events and reactions in blood transfusions from all relevant healthcare organisations in the UK. The observations from the report were corrected for participation in the UK Serious Hazards of Transfusion survey (98%). The external assessment group assumed that the total number of transfused patients per year is around 800,000.

For the trauma model, the probability of transfusion‑related complications was assumed to be the same as that for the cardiac surgery patients. The external assessment group considered this likely to be an underestimation given that people with trauma on average receive more units of blood than cardiac surgery patients, which increases the exposure to various donors.

The probabilities of experiencing transfusion‑transmitted infections were also taken from the UK Serious Hazards of Transfusion report, using the same method of calculation as for transfusion‑related complications. These were also reported as the risk per patient transfused.

For the trauma population, the probability of transfusion‑transmitted infections was assumed to be the same as that for the cardiac surgery population. The external assessment group acknowledged that this is likely to be an underestimation, because patients with trauma receive on average more units of blood than cardiac surgery patients, which increases the exposure to various donors.

For the cardiac model, the estimated risk of mortality in the standard laboratory tests group at 1 month was estimated based on the number of deaths reported in 1 study. This study was based on a large sample (n=8,598) of a population that matched the target population for the assessment. It reported a 1‑month mortality of 0.4% for non‑transfused patients and 4.3% for transfused patients. Using the transfusion percentage applied in the current model (59.2%), this gave an overall 1‑month mortality of 2.7%.

Even though mortality may vary by complication, it was assumed that the mortality of all transfused patients (essentially the sum of mortalities due to each complication and no complication) was fixed at 4.3%. Therefore, in order to obtain a 4.3% mortality in the transfused group, the external assessment group used a calibration procedure, meaning that when reliable estimates were available, a specific mortality estimate was applied to each complication. For the rest, and for no complications, the mortality value was calculated so that the total mortality added up to 4.3%. This mortality value was calculated to be 4.28%. For the transfusion‑transmitted infections (except bacterial contamination), 1‑month mortality was assumed to be zero because it was assumed that these infections appeared after the hospitalisation period. Mortality for various transfusion‑related complications and bacterial contamination were derived from the UK serious hazards of transfusion survey, unless they were lower than the calibrated mortality value. Those complications with mortality lower than 4.3% were included in the calibration procedure.

In order to estimate the mortality associated with the use of viscoelastometric testing, the external assessment group assumed that any mortality benefit from viscoelastometric testing resulted from fewer patients receiving a transfusion. This meant that the 1‑month mortality for each group (not transfused, transfused without complications, and transfused with complications) in the viscoelastometric group was assumed to be the same as in the standard laboratory tests group.

At 1 year, the mortality in the standard laboratory tests group was estimated using data from a study which reported 1‑year mortality of 1.2% for non‑transfused patients and 7.8% for transfused patients. For the non‑transfused patients, 0.4% mortality at 1 month and 1.2% mortality at 1 year gave a mortality of 0.8% for between 1 and 12 months. Similarly, for the transfused patients, mortality for between 1 and 12 months was calculated as 3.66%. The 1‑year mortality for each subgroup of patients in the viscoelastometric group was assumed to be the same as in the standard laboratory tests group.

For the trauma population, the external assessment group used a random effects model to estimate mortality at 1 month based on the studies included in the effectiveness review. In the standard laboratory tests group, the mean 1‑month mortality was 15.7% (95% CI 11.7% to 20.1%). The external assessment group assigned 1‑month mortality to transfused and non‑transfused patients, such that the overall mortality would be equal to 15.7%. One study was retrieved that showed that mortality was 3.3 times higher among patients who received a transfusion. Therefore, the goal was to estimate mortality such that the weighted average of these gave an overall mortality of 15.7%, the mean mortality in the standard laboratory tests group derived from the systematic review. From this it follows that mortality was 9.1% in patients who did not receive a transfusion and 29.8% in those who did.

Mortality for the 2 trauma and/or transfusion‑related complications acute respiratory distress syndrome and multiple organ failure were estimated from other sources. The probability of mortality was estimated from a trial in acute respiratory distress syndrome patients that reported a mortality of 83 of 385 (21.6%). Data from 2 studies were pooled to estimate the mortality in patients with multiple organ failure, giving an overall mortality of 26.2%.

One‑month mortality rates for transfusion‑related complications and transfusion‑transmitted infections were derived when possible from the Serious Hazards of Transfusion survey, and, as in the cardiac surgery population, it was assumed that all infections apart from bacterial contamination would only appear after 1 month, implying zero mortality in the first month. As in the cardiac population, the 1‑month mortality for each subgroup of patients in the viscoelastometric group was assumed to be the same as in the standard laboratory tests group, implying that any mortality benefit in the viscoelastometric group was due to fewer patients being transfused.

Few data were available for mortality between 1 and 12 months after trauma. One study was identified, which reported 3% mortality for this period, but no information was identified on how this mortality was distributed over transfused and non‑transfused patients. The external assessment group therefore applied the same ratio as for 1‑month mortality (3.3). This gave a mortality of 1.7% in non‑transfused patients and 5.7% in transfused patients. These values were assumed to apply to both the standard laboratory tests and the viscoelastometric group.

Health benefits were expressed in terms of life years and QALYs gained at 1 month and 1 year. For the calculation of the life years, patients were assumed to die in the middle of the period where death occurred. Life years were then valued with different utilities depending on the health state of the patient. Except for stroke, the external assessment group used utility values from the 1996 Health Survey for England.

For the trauma model, the external assessment group identified a study that collected EQ‑5D utilities 12 to 18 months after trauma. This study included patients with severe trauma and reported a mean utility of 0.69 (standard error 0.016) 12 to 18 months after the trauma. No studies reporting utilities for the period of hospitalisation and shortly afterwards were identified. The external assessment group therefore assumed the same utility for the period of hospitalisation as for the cardiac population during hospitalisation.

For patients with acute respiratory distress syndrome, the external assessment group used the results of a prospective cohort study that measured quality‑adjusted survival in 200 patients in the first year after acute respiratory distress syndrome. This study reported utilities of 0.60 (standard error 0.01) and 0.64 (standard error 0.01) at 6 months and 1 year after onset of acute respiratory distress syndrome respectively. The external assessment group applied a value of 0.60 to the period of 1 month, and 0.64 to the period between months 1 and 12. Similar data were unavailable for patients with multiple organ failure, so the external assessment group applied the same utilities as for patients with acute respiratory distress syndrome based on the assumption that both complications are similar in severity. For patients with transfusion‑related complications, the external assessment group assumed that after discharge, as in the cardiac population, the utility would be equivalent to patients without complications.

For the cardiac model, all the viscoelastometric devices dominated (that is, were more effective and less costly than) standard laboratory tests. The external assessment group assumed the same treatment effects for each viscoelastometric testing device (QALY=0.8773). The cost of Sonoclot was lower than that of ROTEM and TEG, and the device was associated with greater cost savings (£132) than either TEG (£79) or ROTEM (£43).

For the trauma model, all the viscoelastometric technologies dominated standard laboratory tests. The effectiveness of the devices was the same (QALY=0.5713). The cost of Sonoclot was lower than that of ROTEM or TEG and so this device was associated with greater cost savings (£818) than TEG (£721) or ROTEM (£688).

The results of other outputs from both the cardiac and trauma models showed that, compared with standard laboratory tests, the use of viscoelastometric devices is associated with lower mortality, a reduced probability of experiencing complications, and lower transfusion and hospitalisation costs. The probability of experiencing transfusion‑transmitted infections was very low (almost zero) in both groups but lower in the viscoelastometric group.

The impact of statistical uncertainties regarding the models input parameters was explored through probabilistic sensitivity analysis.

For the cardiac model, the scatter plot of the probabilistic sensitivity analysis outcomes in the cost‑effectiveness plane was not very informative because the model only assumed a difference in costs between the technologies. The probabilistic sensitivity analysis confirmed that using standard laboratory tests is the strategy with the lowest probability of being cost effective. If the maximum acceptable incremental cost‑effectiveness ratio (ICER) was £30,000 per QALY gained, the probability of cost effectiveness for each of the 3 viscoelastometric technologies compared with standard laboratory tests was 0.79 for ROTEM, 0.82 for TEG and 0.87 for Sonoclot. When the maximum acceptable ICER was higher than £30,000 per QALY gained, the cost‑effectiveness probabilities converged to around 0.80 for all technologies.

Probabilistic sensitivity analysis results for the trauma model were similar to the cardiac model. The analysis confirmed that using standard laboratory tests was the strategy with the lowest probability of being cost effective. A comparison of ROTEM with standard laboratory tests found a cost‑effectiveness probability equal to 0.96 for ROTEM for a ceiling ratio equal to £0. As the ceiling ratio increased, the cost‑effectiveness acceptability curve for ROTEM converged to 0.87. A similar pattern was observed for TEG and Sonoclot.

For the cardiac model, all scenario analyses suggested that ROTEM remained cost saving. The only exception was the number of tests run on each device per year. If the number of tests run on each device were reduced to 200, ROTEM no longer dominated standard laboratory tests, and ICER was £16,487 per QALY gained. The external assessment group estimated, using iterative analysis, that if all other parameters in the model remained unchanged, the costs of ROTEM and standard laboratory tests would be equal if 326 tests were run on ROTEM each year. At this level the ICER would be £0 per QALY gained. If the number of tests per year were reduced to 152, the ICER would be around £30,000 per QALY gained.

Additional scenario analyses for cardiac surgery suggested that when viscoelastometric testing is carried out in conjunction with standard laboratory testing, TEG was more effective and less costly (-£1) than standard laboratory testing alone and that the ICER for ROTEM and standard laboratory testing alone was £7,487 per QALY gained. When the number of tests and type of assays used were varied, viscoelastometric testing dominated standard laboratory testing alone.

For the trauma model, all scenario analyses suggested that ROTEM remained cost saving. The iterative analysis performed to estimate the number of tests per year such that ROTEM would still be cost saving suggested a break‑even value of 81 tests per year; at this level the ICER was £0 per QALY gained. When the number of tests per year was reduced to 65, the ICER was approximately £30,000 per QALY gained.

Additional scenario analyses for trauma surgery suggested that when viscoelastometric testing is carried out in conjunction with standard laboratory testing, ROTEM and TEG dominated (-£558 and -£591 respectively) standard laboratory testing alone. When the number of tests and type of assays used were varied, viscoelastometric testing dominated standard laboratory testing alone.

For the trauma model, threshold analysis on the combined effect of a reduction in the percentage of patients transfused and the blood volumes transfused (assuming that equal volumes of blood were transfused in the viscoelastometric testing and standard laboratory test groups) showed that, at a relative risk of transfusion of 0.9822 or more, ROTEM was no longer cost saving (with an ICER of £0 per QALY gained). When the relative risk of transfusion increased to 0.9874, the ICER of ROTEM compared with standard laboratory tests was £30,000 per QALY gained.

Reducing baseline transfusion risk in the standard laboratory test group (assuming that equal volumes of blood were transfused in the viscoelastometric testing and standard laboratory test groups) showed that ROTEM was no longer cost saving at a transfusion rate of 5%, and that the ICER was £30,000 per QALY gained for a transfusion rate of 4%. This compares with a transfusion rate of 32% used in the base‑case analysis. When the analysis was repeated with an increased relative risk of red blood cell transfusion (from 0.88 to 0.95), the ICER was above £30,000 per QALY gained for a transfusion rate of 8% or less. After reducing the probability of complications related to trauma and/or transfusion, transfusion‑related complications and transfusion‑related infection to zero, ROTEM remained cost saving with a reduction in costs of £372.