Let’s start by considering what our pretest probability should be regarding whether hypothermia is beneficial after cardiac arrest.
Every enzyme in our body has been evolutionarily designed to function at our normal body temperature. Therefore, hypothermia will affect every chemical pathway.
As such, hypothermia is an incredibly blunt tool. It’s hard to think of any other medical intervention with such an infinite range of consequences. There is no tissue, nor cell, nor extracellular space which will escape being altered somehow by hypothermia.
This makes it wholly impossible to predict the net effect of hypothermia. Hypothermia will slow down inflammation, but it will also slow down anti-inflammatory pathways. Hypothermia will reduce intracranial pressure, but it may also reduce blood pressure and cerebral perfusion pressure. We could play this game all day and get nowhere.
On the whole, though, it seems unrealistic to expect hypothermia to be greatly beneficial. The body is an enormously complex machine that is optimized to function at a certain temperature. Changing the temperature – and thereby affecting every metabolic process – and somehow expecting this to have a huge benefit… well, that seems like a stretch.
This is somewhat reminiscent of many historical trends in neurocritical care. For periods of time, there has been enthusiasm for hypoventilation, hyperoxygenation, and routine hypernatremia. In every case, it was eventually determined that simply maintaining normal homeostasis is best (normocapnia, normothermia, normonatremia). So the neurointensivist is best off being humble – rather than pushing things to extremes, it’s better to merely defend normal homeostasis.
hypothermia begins: the HACA and Bernard trials
Two similar studies in the early 2000’s kicked off clinical enthusiasm for therapeutic hypothermia after cardiac arrest. The Bernard trial was a small, quasi-randomized trial with substantial methodological limitations (explored further here by Justin Morgenstern). The HACA trial was a larger RCT that really got the party started.
The HACA trial found that hypothermia caused a 14% absolute reduction in mortality after six months. 14%!! That is an insanely massive absolute mortality reduction. That’s greater than the benefit of any medical therapy in the history of critical care RCTs! (discussed further here) A 14% absolute mortality benefit from a therapy with a dodgy mechanism of action is difficult to believe.
Both of these trials suffered from many of the same limitations:
- They were unblinded. Many deaths following cardiac arrest are the result of withdrawal of life supportive therapy, due to a perceived poor neurological prognosis. Exposure of some patients to a new and exciting therapy could have biased clinicians towards waiting longer prior to neuroprognostication, which would bias results in favor of hypothermia.
- Performing therapeutic hypothermia requires more clinical attention from nurses and physicians. It’s possible that randomization into the hypothermia group simply caused patients to receive more intensive management, which improved their outcomes.
- Many patients in the control group developed fever. Thus, it’s possible that hypothermia was effective merely because it prevented fever (rather than offering any benefit from the hypothermia).
hypothermia continues: mid 2000’s – 2013
Based on these studies, hypothermia became the law of the land. Hypothermia was recommended by all the guidelines. Every hospital had a hypothermia protocol. Hypothermia was so hot.
I remember being frustrated by this boondoggle as a fellow. We were under lots of pressure to maintain patients on hypothermia, even when the hypothermia seemed to be contributing to problematic bradycardia and cardiogenic shock. I remember reading the HACA trial and wondering whether the benefit was simply due to fever prevention. Unfortunately, without more evidence, we were stuck following the available data – that hypothermia improved outcomes.
TTM1 trial – 2013
After years of freezing patients without knowing whether it was actually helping, the TTM trial was downright dreamy. The trial asked whether there was benefit from hypothermia (33C) when compared to enforced normothermia (36C). The trial was modern, methodologically robust, and large (involving 950 patients – compared to 220 patients in HACA and Bernard et al. combined).
The patient outcomes in both arms of the TTM trial were essentially the same. There were a few subtle signs of superior hemodynamics in the 36 C group, but overall there were really no significant differences.
…and the hypothermia continues: 2013-2021
So where did that leave us? The biggest and best study showed that our darling therapy was really just an extremely fancy and expensive antipyretic.
I switched my practice towards using TTM36 for comatose post-arrest patients. The main reasons for this were the following: (more on this from a 2015 post here)
- 33C may cause hemodynamic instability in some patients.
- 33C is associated with an increased risk of infection.
- 33C may cause patients to wake up later. This prolongs time on the ventilator and increases ICU complications.
- 33C causes bradycardia, which can aggravate torsade de pointes in some patients.
- Uniform management of all patients with a single protocol (36C) rather than two different protocols (33C vs 36C) streamlines management, and avoids perpetual debates about what temperature to use.
This has seemed to work well. At first we had some increased difficulty managing shivering, but with the use of a proactive multimodal shivering package this was easily manageable. With the use of 36C and light sedation using short-acting sedatives (e.g., propofol monotherapy), patients woke up earlier and were more easily rehabilitated.
However, enthusiasm for the magic of hypothermia persisted. Many centers continued to use 33C. The guidelines recommended using either 33C or 36C.
Persistent use of 33C largely reflected status quo bias – wherein we become accustomed to a status quo and tend to assume that it’s the best. Alternatively, imagine if the TTM trial had been published before HACA and Bernard et al. It’s difficult to imagine that HACA and Bernard et al. would have incited much enthusiasm for hypothermia, had it already been known that hypothermia was equivalent to normothermia.
Towards the end of this period, the HYPERION trial came out, which barely demonstrated an improvement from hypothermia among patients with cardiac arrest from a nonshockable rhythm (10% vs. 6% with good neurological outcome, p=0.047). The combination of a fragility index of one with and an outcome based on neurological scoring (which is notoriously wobbly) precludes reaching any solid conclusion.
In related news, the POLAR and EUROTHERM studies demonstrated no benefit from hypothermia when applied to patients with traumatic brain injury (with some signs of harm in EUROTHERM). Consensus guidelines warned against the use of hypothermia for patients with traumatic brain injury. Meta-analysis of RCTs using hypothermia for applications other than cardiac arrest suggested serious harms:
TTM2 – 2021
The TTM trial group has finally came back to the rescue, with a second attempt to inject some reality into the situation. In the paper explaining their study design, the authors remind us that hypothermia has only been demonstrated to work in two studies with high risk of bias. This is a low level of evidence, on par with a small phase II clinical trial. We would never approve a new drug based on this level of evidence.
The TTM2 trial was larger (containing nearly 2,000 patients) and more aggressive at challenging the status quo (comparing hypothermia at 33C to merely fever control). Patients in the control arm were not automatically placed on a temperature control device, but instead this was initiated only if they developed a temperature above 37.8 C (with the device set to target a temperature of 37.5C for 40 hours).
To limit bias involved in decisions regarding the withdrawal of life-sustaining therapy, neuroprognostication was performed after 96 hours by blinded clinicians in a standardized fashion (with this information provided to the primary team). Families were also blinded. The primary team involved in final decisions regarding management was not blinded, so some bias could have occurred. Nonetheless, the implementation of a standardized, blinded, uniform approach for neuroprognostication represents a major improvement compared to earlier studies (e.g., HACA).
The primary results were neutral. Hypothermia had no effect on mortality or neurological endpoints.
There were numerous signs of iatrogenic harm due to hypothermia:
- Patients in the hypothermia group had a higher risk of arrhythmia causing hemodynamic instability (24% vs. 17%, p<0.001).
- Patients in the hypothermia group required paralytics more often (66% vs. 45%, p<0.001).
- Patients in the hypothermia group had a longer median length of mechanical ventilation (3.8 days vs. 2.9 days).
- Patients in the hypothermia group experienced more than twice as many unexpected severe adverse events (3.7% vs. 1.4%, p=0.003).
where do we go from here?
It is increasingly clear that hypothermia is merely an unnecessarily aggressive approach to avoiding fever. With two robust studies showing no benefit from hypothermia (TTM1 & TTM2), it’s no longer possible to justify the cost and iatrogenic harms of hypothermia (outside of a clinical trial).
However, we must not throw out the baby with the bathwater. Temperature control may be critically important for post-arrest patients. As such, a wholly nihilistic approach to temperature management could cause substantial harm. As we transition away from hypothermia, we must do so cautiously. For now, the most prudent strategy is strict enforcement of normothermia using the protocol from TTM2. That means placing continuous temperature probes in post-arrest patients (e.g., bladder or esophageal), paying attention to the temperature, and aggressively treating any temperatures >37.8 (e.g., with adaptive cooling systems that will achieve definitive temperature control). The upcoming TTM3 trial will hopefully provide further guidance.
True believers in hypothermia will continue to try to poke holes in TTM2 (good luck with that – the study is a methodological fortress). Or they will propose that hypothermia would have worked if it was started earlier (actually, as shown below, the rate of cooling was very fast – just as fast as the HACA and HYPERION studies that are traditionally lauded as being “positive” trials). Or they will say that their patient population is different (actually, results in TTM2 are robust across every subgroup analysis). The bottom line is that the theoretical and evidentiary basis for post-arrest hypothermia in adult humans was never there to begin with. It just doesn’t work.✌️
- IBCC chapter on post cardiac arrest care (updated based on the TTM2 trial, with a new algorithm for temperature & shivering control).
- The Bottom Line review of TTM2 by Celia Bradford.
- Paul Young's tweet thread on TTM2.