Advice

Evidence review

Evidence review

Clinical and technical evidence

Regulatory bodies

A search of the Medicines and Healthcare Products Regulatory Agency website revealed no manufacturer Field Safety Notices or Medical Device Alerts for this device.

Sixteen reports, relating to 14 separate adverse events, were identified from a search of the US Food and Drug Administration (FDA) database: Manufacturer and User Device Facility Experience (MAUDE). These reports were dated between November 2010 and March 2015. When investigated by the manufacturer, many of these reported events were either found to be non‑system faults, irreproducible faults, or cases where the system in question was not returned to the manufacturer for investigation. Malfunctions confirmed by the manufacturer included:

  • a defective lower controller board, which resulted in failure of the coolant to reduce in temperature

  • a clogged cold well mesh, which resulted in the system failing to cool the patient

  • leaking coolant, which resulted in the system failing to cool the patient

  • a damaged temperature sensor and wire harness, which resulted in the system failing to cool.

One further event, dated 4 March 2015, does not yet appear to have been investigated by the manufacturer. This event reported a small hole in the catheter, resulting in the system alarm activating. The intravascular solution bag was found to be empty and the catheter had to be removed.

Clinical evidence

Ten studies reporting cooling efficiency outcomes following cardiac arrest were identified. Only 1 of these explicitly stated that the Thermogard XP device was used (Oh et al. 2015). Two others stated that its immediate predecessor, the CoolGard 3000, was used (Holzer et al. 2006; Wolff et al. 2009). The Thermogard XP has increased cooling power output compared with the CoolGard 3000 (190 watts compared to 115 watts). These 3 studies were assessed in full in this briefing.

Although the other 7 studies stated that they used the CoolGard system, it was not clear which version was used. Because the earlier CoolGard 2050 system had a different operation and user interface to both the CoolGard 3000 and Thermogard XP, these other studies are included in summary form only (table 7).

The study by Oh et al. (2015) used the Thermogard XP. This retrospective cohort study analysed South Korean registry data from 24 hospitals collected between 2007 and 2012 to compare the neurological outcomes, efficacy and adverse events of various surface cooling techniques (with and without an automatic temperature feedback system), compared with Thermogard XP, following cardiac arrest. Results were reported for the whole cohort and further matching was done to adjust for the baseline characteristics in the Thermogard XP and surface cooling groups. The outcomes were then re‑evaluated in the matched cohort.

Cooling efficacy was reported as induction time to target temperature in minutes. In the Thermogard XP group, target temperature was achieved in an average of 209.4±15.4 minutes, compared with 235.3±18.0 minutes for surface cooling devices (odds ratio 1.13, 95% confidence interval [CI] 0.79 to 1.62, p=0.51). There was no significant difference in neurological outcomes or hospital mortality between the surface cooling and Thermogard XP groups (p=0.31 and p=0.44, respectively). The rates of some adverse events were statistically significantly higher in the surface cooling group compared with the Thermogard XP cooling group including overcooling, rebound hyperthermia, rewarming‑related hypoglycaemia and rewarming‑related hypotension. These complications were not associated with surface cooling using hydrogel pads (which employ automatic temperature feedback regulation). An overview and summary of results from Oh et al. (2015) are included in appendix tables 1 and 2.

Holzer et al. (2006) investigated the effectiveness and safety of the CoolGard 3000 in 1038 people consecutively admitted following cardiac arrest by collating retrospective clinical data in a bespoke registry format. The hypothermia group consisted of 97 patients who had the CoolGard 3000 to achieve a target temperature of 33°C. patients who had standard post‑resuscitation therapy, comprising analgesia and sedation, served as controls (941 patients). Patients who had initiation cooling through cold infusion as well as the CoolGard 3000 were excluded from analysis of cooling rate. This resulted in 56 of 97 CoolGard 3000 patients contributing to the analysis of cooling rate, which was reported as 1.2°C per hour, with an interquartile range of 0.7 to 1.5°C per hour. The primary reported outcome of survival at 30 days was 69% (67/97) in the CoolGard 3000 group and 50% (466/941) in the standard care group (odds ratio 2.28, 95% CI: 1.19 to 3.23, p=0.008). When neurological outcome was considered, the proportion of patients with good neurological recovery (defined as alert and with sufficient cerebral function to live independently and work part‑time) was 53% (51/97) in the CoolGard 3000 group and 34% (320/941) in the standard care group (odds ratio 2.15, 95% CI: 1.57 to 4.17, p<0.001). The only statistically significant difference in adverse events between the 2 groups was transient bradycardia, with 15% (9/62) reported in the CoolGard 3000 group and 3% (2/104) in the standard care group (p=0.025). The authors concluded that treatment with the CoolGard 3000 after resuscitation following cardiac arrest reduced mortality and improved favourable neurological recovery at 30 days after cardiac arrest, or discharge (if sooner), compared with controls from a retrospective resuscitation database. An overview and summary of results from Holzer et al. (2006) are included in appendix tables 3 and 4.

Wolff et al. (2009) conducted a prospective case series of 49 patients with documented out‑of‑hospital or in‑hospital cardiac arrest in a single tertiary hospital in Germany, with the hypothesis that the clinical benefit of mild therapeutic hypothermia (in the range of 32 to 34°C) is greater when more rapidly achieved. All of the patients were cooled using the CoolGard 3000, with a target temperature of 33°C. The median time to target temperature was 410 minutes (interquartile range 271 to 544 minutes). Based on the neurological outcome at discharge, the patient group was split into good outcomes (no/mild cerebral disability; cerebral performance categories 1 and 2) and poor outcomes (severe disability, coma, brain death; cerebral performance categories 3–5). The median time to target temperature for good outcome (28 patients) was 334 minutes (interquartile range 250 to 498 minutes) compared with 450 minutes (interquartile range 322 to 674 minutes) in the poor outcome group (21 patients; p=0.71). Mild therapeutic hypothermia was maintained adequately for 24 hours in all patients. The authors concluded that achieving mild therapeutic hypothermia early improves neurological outcomes and thus measures to speed up the initiation of cooling therapy after cardiac arrest appear warranted. An overview and summary of results from Wolff et al. (2009) are included in appendix tables 5 and 6.

The 7 additional studies investigated the cooling efficiency of unspecified models of the CoolGard system. Three studies compared CoolGard with surface cooling systems that have automatic temperature feedback control (de Waard et al. 2014; Pittl et al. 2013; Tømte et al. 2011), 2 compared it with surface cooling systems without automatic temperature feedback control (Knapik et al. 2011, Flemming et al. 2006), and 1 compared it with both system types (Gillies et al. 2010). This briefing also includes a single non‑comparative study (Pichon et al. 2007).

In the studies which compared CoolGard with surface cooling systems that have automatic temperature feedback control, CoolGard tended to have a similar cooling rate but better temperature control in the maintenance phase. As with the findings from Oh et al. (2015) regarding the Thermogard XP compared with hydrogel pads (a surface cooling system with automatic temperature feedback control), these 3 studies showed no significant differences in survival to hospital discharge and neurological outcomes. The study by de Waard et al. (2014) reported a measured cooling rate of 0.65°C per hour for CoolGard compared with the 1.2°C per hour reported for the CoolGard 3000 by Holzer et al. (2006). de Waard et al. also reported an improved Glasgow Coma Scale of 15 at discharge in the CoolGard group (interquartile range 3 to 15) compared with 10 in the surface cooling group (interquartile range 4 to 13; p=0.008.)

In the studies comparing CoolGard with surface cooling systems without automatic temperature feedback control, CoolGard tended to demonstrate both improved cooling rate and better temperature control.

In the final CoolGard study (Pichon et al. 2007), the authors reported an average time to the 33°C target temperature of 187 minutes (interquartile range 30 to 600 minutes), with hypothermia maintained in 91% of patients.

A summary of the results from these 7 additional studies is included in table 7.

Recent and ongoing studies

Two relevant ongoing or in‑development trials of the Thermogard XP (or its predecessor devices) were identified in the preparation of this briefing:

  • The COOL‑ARREST pilot study, for witnessed out‑of‑hospital cardiac arrest (ClinicalTrials.gov identifier NCT01818388)

  • Finding the Optimal Cooling Temperature After Out‑of‑Hospital Cardiac Arrest (FROSTI) for 3 different levels of hypothermia (32, 33 and 34°C), in comatose survivors from out‑of‑hospital cardiac arrest (ClinicalTrials.gov identifier NCT02035839).

Costs and resource consequences

The UK supplier, Delta Surgical, reported that as of April 2015, 56 hospitals across the UK were using the technology. Therapeutic hypothermia is accepted as routine clinical practice in the intensive care setting in the NHS according to NICE interventional procedure guidance on therapeutic hypothermia following cardiac arrest.

No published evidence on the resource consequences of Thermogard XP was identified.

One paper was identified which addressed key nursing aspects of implementing the CoolGard 3000 and surface cooling methods (Thermowrap and Arctic Sun, plus ice‑water soaked towels) in critical care. Våga et al. (2008) surveyed intensive care nurses to subjectively evaluate ease of use, visual patient monitoring, workload, hygiene and noise level on a 4‑point scale (1=worst, 4=best possible). There were some statistically significant differences in the results. For workload, all 3 commercial cooling devices scored significantly better than ice‑water soaked towels (p<0.05). Only the CoolGard 3000 system scored significantly better than ice‑water soaked towels for visual patient monitoring (p<0.001).

Strengths and limitations of the evidence

Although only 1 paper was found within the scope of this briefing that explicitly identified the Thermogard XP device for cooling following cardiac arrest (Oh et al. 2015), there is a large evidence base for its predecessor technology the CoolGard 3000. It was judged reasonable to assume that evidence for the CoolGard 3000 would be generalisable to the Thermogard XP, since both systems have the same user interface, use the same disposable accessories, and are operated the same. The only difference between the systems is the increased cooling power output of the Thermogard XP.

The retrospective and multicentre nature of the Oh et al. (2015) registry introduced the possibility of selection bias and reporting bias in the study. With regard to cooling efficiency, the target temperature was not reported. Because local clinical cooling protocols were likely to vary across the 24 hospitals recruited in the registry, the influence of target temperature on induction time to cooling could not be quantified. In addition, the authors highlighted that a number of confounding factors including cardiac interventions, haemodynamic status, time of endovascular catheter insertion following cardiac arrest and complications relating to vascular access were not recorded in the registry data fields. A lack of data on these confounders means that the reported cooling and adverse event rates should be interpreted with caution. One strength of this study was the analysis carried out to adjust for differences in the baseline characteristics of each cohort. However, after the matching process was implemented, the final sample size was relatively small and comparisons had low statistical power, which was illustrated by the wide confidence intervals for the odds ratios in a number of the reported outcomes. It was also notable that this study was set in South Korea and therefore differences in the care pathways and population may limit generalisability to the NHS setting.

Similar to the study by Oh et al. (2015), Holzer et al. (2006) adopted a retrospective data analysis approach. In this instance, the authors compiled a bespoke registry from historical clinical data, which introduces the potential for selection bias and reporting bias. However, because these data are taken from a single institution in Austria, the study may be less susceptible to multicentre variations in clinical practice. Holzer et al. do not report cooling efficacy as a predefined outcome measure, but the authors present their cooling protocol in sufficient detail to obtain comparative data for the purposes of this briefing. Potential confounders, including variability in baseline characteristics, are robustly addressed in the data analysis and statistical treatment. However, the study was not blinded for the assessment of neurological outcome, which the authors state could have led to information bias. The primary end point of survival was objective and therefore robust. Disclosures by these authors make clear that the study was conducted with financial support from the manufacturer, although the company was not involved in key decisions around study design, data analysis and final manuscript content.

Wolff et al. (2009) was a non‑comparative study with a sample size of 49 patients. Consecutive enrolment of eligible patients in this prospective case series minimised selection bias, but the post‑hoc nature of stratification by neurological outcomes at discharge may be viewed as a weakness in the study design. The authors acknowledged the heterogeneous patient characteristics, including cardiac arrest occurring both in and out of hospital, with ventricular fibrillation/tachycardia and asystole. Although this may introduce potential confounders into the study, it is likely to be a realistic reflection of daily practice. A multivariate analysis was also done to adjust for differences in the baseline characteristics between good and poor outcomes. The authors reported that 11 of the 49 patients failed to reach the target temperature of 33°C, despite maximum cooling rates being applied. This may be explained by the inclusion of patients with potentially concurrent infection or neurogenic fever occurring as a result of severe anoxic brain injury. Each of these conditions would typically raise core body temperature, counteracting the cooling effect of the therapeutic hypothermia intervention.