5.1 The company identified 2 studies that incorporated a cost-effectiveness analysis. The external assessment centre (EAC) judged that the company's search strategy was highly sensitive and well-constructed, and that the selection criteria reflected the NICE scope. The EAC carried out its own economic search and found no additional studies.
5.2 Both of the identified studies were published as conference abstracts and compared HumiGard with standard care in the UK. The company provided unpublished, academic-in-confidence draft manuscripts relating to both abstracts.
5.3 The abstract by Jenks et al. (2015) reported on a cost-utility analysis using a decision analytic model of HumiGard compared with standard care open or laparoscopic colorectal surgery. This showed that HumiGard dominated standard care in both open and laparoscopic surgery (that is, it was both less costly and more effective than standard care). The full manuscript by Jenks et al. provided further detail on the study and was available to the EAC. Effectiveness data for open surgery were derived from Frey et al. (2012; section 3.9). Data on the probability of complications related to hypothermia were taken from a published retrospective study (Billeter et al. 2014) and linked to the data from Frey et al. (2012). The effectiveness data for laparoscopic surgery were taken from a retrospective cohort study reported in a conference abstract (Noor et al. 2015). The costs of myocardial infarction, stroke, sepsis and pneumonia were taken from NHS reference costs 2013/14. The cost of surgical site infections was derived from the NICE quality standard. The results presented in the full manuscript matched those reported in the abstract.
5.4 The study by Mason et al. 2016 (section 3.7) also reported a cost-benefit analysis of HumiGard compared with standard care in patients having laparoscopic colorectal surgery. The EAC was unable to replicate the cost analysis from this study.
5.5 The company presented a de novo economic model adapted from Jenks et al. estimating mean cost savings per patient in open and laparoscopic colorectal surgery. The model assumed a 70:30 split for the use of HumiGard in laparoscopic and open surgery respectively. It comprised 2 decision trees incorporating complications associated with hypothermia and related NHS costs for each kind of surgery. The model runs over 1 year; horizons up to 5 years were reported in scenario analyses, but because these extend post-myocardial infarction and stroke costs they affect only open surgery. The model was based on 3 studies: Noor et al. 2015 (laparoscopic surgery: incidence of surgical site infections and pneumonia), Frey et al. 2012 (open surgery: proportion of patients with hypothermia at the end of surgery) and Billeter et al. 2014 (open surgery: incidence of myocardial infarction, stroke, sepsis, pneumonia, surgical site infection and mortality).
5.6 The company's scenario analyses included exploring the use of alternative sources of clinical effectiveness, a univariate deterministic sensitivity analysis and a probabilistic analysis of the base-case results. For open surgery, it used 3 alternative sources for the proportion of patients experiencing complications (Kurz et al. 1996, Flores-Maldonado et al. 2001, Anannamcharoen et al. 2012). For laparoscopic surgery, the company presented 2 scenario analyses that used data on the proportion of patients with hypothermia linked with complications associated with open surgery (Billeter et al. 2014). The first of these used data from Mason et al. (2016) whereas the second used data from Sammour et al. (2010). The analyses showed that the costs for treating stroke (£2,715 to £13,858) and surgical site infections (£2,100 to £10,500) had the largest effects on the results.
5.7 The company's base case showed that, overall, HumiGard costs £419 per patient compared with £724 per patient for standard care. The company therefore estimated that using HumiGard would save £305 per patient. Most cost savings (69%) come from fewer surgical site infections after laparoscopic surgery (with cost savings of £20 per patient in open surgery and £428 per patient in laparoscopic surgery).
5.8 Sensitivity analyses showed that HumiGard becomes cost incurring when the absolute difference in infection risk is 0.3% (for example, 4.7% versus 5%). For open surgery, using data from Frey et al. (2012), HumiGard was associated with a modest additional cost (using complication data from Billeter et al. 2014 or Flores-Maldonado et al. 2001).
5.9 The company's probabilistic sensitivity analysis found that HumiGard was cost saving in 97.4% of iterations and the average probabilistic cost savings were £302 per patient. The company noted that the results of its probabilistic sensitivity analysis have a skewed distribution and stated that this is because of the distribution of costs of complications within the model (which have a gamma distribution bounded by 0, but no upper limit).
5.10 The EAC re-ran the company's base case and univariate sensitivity analyses for open and laparoscopic surgery separately, and conducted additional analyses using its preferred estimates. The main changes to the company's model were:
including updated NHS reference costs for pneumonia, acute myocardial infarction and sepsis
annuitising the capital cost of HumiGard
re-estimating the costs of 'post-myocardial infarction' to reflect current drug prices
using alternative costs of treating stroke and surgical site infections
using a 5-year time horizon and including data on hypothermia from the randomised control trial in laparoscopic surgery linked to data on complications from the retrospective cohort study (laparoscopic surgery only).
5.11 The EAC re-ran univariate sensitivity analyses for open and laparoscopic surgery, including updated costs for adverse events and a discount rate for HumiGard of 3.5% over 5 years. In addition to this for laparoscopic surgery, the EAC took hypothermia data from Sammour et al. (2010) and risk of complications data from Billeter et al. (2014). The EAC considered that because stroke and myocardial infarction have long-term resource implications, a longer time horizon was preferable. However, the model incorporates this by simply adding in additional costs to later years, so the EAC also conducted analyses using a 1-year time horizon. Additional EAC sensitivity analyses included an alternative estimate for the cost of treating surgical site infections (£5,164, based on Jenks et al. 2015) and laparoscopic surgery complication data from Noor et al. (2015).
5.12 For open surgery, the results of the EAC's analysis suggest that HumiGard is cost saving compared with standard care, with an average saving per patient of £209. This is a larger cost saving than that identified in the company's model because of the longer (5-year) time horizon. The probability that HumiGard is cost saving was 98% in the sensitivity analysis. The results for a 1-year time horizon were broadly similar to those reported by the company (an average cost saving of £28 per patient).
5.13 For laparoscopic surgery, the EAC concluded that savings were lower than in the company model (an average of £77 per patient) because the EAC used data from Sammour et al. (2010) rather than Mason et al. (2016). The probabilistic analysis found that HumiGard was cost saving in 67.5% of iterations. Using a 1-year time horizon, HumiGard was associated with a small additional cost of £11 per patient.
5.14 The committee was uncertain about assumptions and parameters in the cost modelling which could not be addressed by the evidence presented. The committee noted that the effect of hypothermia on the risk of stroke during abdominal surgery, the incidence of surgical site infection and the cost of a surgical site infection to the NHS were parameters associated with most uncertainty. The EAC was asked to make further changes to the model to better inform the economic analysis (sections 5.15 to 5.19).
5.15 The committee was advised by clinical experts that the risk of stroke during abdominal surgery is very low. In the context of elective colorectal surgery, the experts estimated it to be less than 1%. Hospital Episode Statistics data were presented to the committee on perioperative stroke rates for England. The data represented selected abdominal procedures that were done in April 2014 and were followed by a primary diagnosis of a stroke at any time during the 2014/15 financial year. The relevant procedures were selected following expert advice. The stroke rates were 0.4% for laparoscopic surgery and 0.6% for open surgery.
5.16 The EAC reviewed the NICE guideline on hypothermia to identify additional data on the associated complications. The guideline cited a study by Frank et al. (1997), as well as 2 studies (Kurz 1996 and Flores-Maldonado 2001) used in the company's model (section 5.6). Nevertheless, following the review, the EAC re-affirmed its view that Billeter et al. (2014) was most relevant to the decision problem.
5.17 The EAC used 2 sources (Sammour et al. 2010 and Mason et al. 2016) of clinical-effectiveness data to better characterise the remaining uncertainties in further cost analyses for laparoscopic surgery. The EAC used data in a personal communication from Mason et al. (2016) to calculate adjusted risks for hypothermia and surgical site infections, taking into account the population characteristics in each study arm. The EAC also used data in a personal communication from Sammour et al. (2010) to assess hypothermia risk with and without HumiGard.
5.18 The EAC used a range of additional analyses to assess how different stroke rates, surgical site infection costs and sources of effectiveness data affect HumiGard's potential cost savings.
5.19 For open surgery (using data from Frey et al. 2012 data on hypothermia risk and Billeter et al. 2014 data on risk of complications), HumiGard appears to be associated with a cost saving for scenarios when the difference in risk of stroke between hypothermic and normothermic patients is greater than 0.75% to 1.25% (depending on the cost of surgical site infections). At a stroke risk difference below this range, HumiGard is associated with a modest increase in mean cost per patient.
5.20 For laparoscopic surgery (using data from Billeter et al. 2014 and Sammour et al. 2010), HumiGard is cost saving only if the difference in stroke risk between hypothermic and normothermic patients is greater than 1.75% to 2.25% (depending on the cost of surgical site infections). Additional analyses using the data from Mason et al. (2016; and the updated predicted risk data calculated by the EAC) suggest that HumiGard is cost saving regardless of the cost of surgical site infections and stroke risk when using a range of complications data from Billeter et al. (2014), but cost saving or cost neutral when using only direct data on surgical site infection complications. However, the EAC was unable to fully appraise these models because of incomplete information from Mason et al. (2016).
5.21 The committee was informed by the clinical experts that the 5.5% stroke risk extrapolated from Billeter et al. (2014) in the company's cost model was an overestimate of the risk in current UK NHS practice, and that this is more likely to be less than 1%. The committee concluded that this distinction is likely to be very influential in the outcome of cost modelling. The committee was informed by the EAC that reducing the stroke risk to 0% in the cost model would make the use of HumiGard cost incurring. The committee concluded that the use of HumiGard was unlikely to reduce stroke rates for patients having abdominal surgery in the NHS.
5.22 The committee was informed that the NHS costs associated with surgical site infections were uncertain and that published estimates vary. The committee noted that the average cost used in the EAC cost analysis was reflective of current practice. Expert advice stated that surgical site infection costs vary considerably in colorectal surgery.