Neurocritical Care

29. Do-Not-Resuscitate Orders and Withdrawal of Life Support

A 78-year-old woman is admitted to the NICU with a destructive ICH (large ganglionic hemorrhage involving the diencephalon, Figure 29.1) On examination, she has midsize fixed pupils but with preserved corneal reflexes and a good cough response and she overbreathes the ventilator. She has flexion withdrawal of both arms with nail bed compression in the fingers and triple flexion responses of the legs.

The family arrives and wants “everything done.” The family is very clear about her: She is a fighter and in the past was able to overcome desperate situations in which physicians had given up any hope for recovery. She has told the family “do not let me go so easily.” Recently, an ICD has been placed, and the family interpretation of that procedure is that this also shows she wanted to live. Therefore the family specifically requests to give her all the time she needs to recover and to resuscitate her if that were needed. The patient’s condition is unchanged 3 days later.

What do you do now?


FIGURE 29.1 CT scan showing a massive right intraparenchymal hematoma with intraventricular extension, hydrocephalus (A), and brain tissue displacement causing complete effacement of the basilar cisterns (B).

End-of-life care in the intensive care has become a shared decision-making process in the United States and such an approach is considered more satisfactory than decisions based on physician’s authority alone. When asked in surveys, most families appreciate a physician’s openness and directness. Families would want to know what to expect and what are the limitations of aggressive interventions. In patients with very poor prognosis it is important to review the chances of a successful resuscitation effort.

Where did a do-not-resuscitate (DNR) order originate? In 1974, the American Medical Association stated in an article devoted to standards that “cardiorespiratory resuscitation was not indicated in certain situations.” Other opinions were voiced, including the 1983 President’s Commission on “deciding to forego life sustaining support.” This led to a major change in practice, in which doctors now had the opportunity to discuss any intervention before the event. Most major medical organizations supported the view that do-not-resuscitate orders can be discussed as part of care. How to communicate this to family members or even whether to discuss the actual procedure of cardiopulmonary resuscitation has remained an underdeveloped field of medicine, and training of this part of end-of-life care in residencies is generally not common place.

What do the data say about the success of cardiopulmonary resuscitation in the intensive care unit? In most large series of resuscitated critically ill patients with diverse diagnoses not more than 15% survive to discharge. Advanced age and comorbidity (i.e., cancer) reduce the odds even more. Patients with acute deteriorating neurologic disease complicated by cardiac arrest and cardiopulmonary resuscitation have a very dismal outcome, if they survive at all. Some empirical guidelines in patients with acute neurologic disease are clearly warranted, but none exist. In an earlier statement there appeared to be consensus among stroke physicians that DNR orders are appropriate if 2 of the 3 following criteria are met: 1) severe deficit, persistent or deteriorating and with impaired consciousness; 2) life-threatening brain damage with brainstem compression involving multiple brainstem levels; 3) significant comorbidity, including pneumonia, pulmonary emboli, sepsis, recent myocardial infarction, and life-threatening arrhythmias. Criteria for other critical neurologic conditions have not been developed, but most physicians attending in the NICU would discuss DNR orders—if not already made clear by family members—or an advance directive, if there is permanent and severe primary brain and brainstem injury.

A DNR order clearly specifies no cardiopulmonary resuscitation (no chest compressions, no pharmacologic or electrical cardioversion). Do not intubate orders (no endotracheal intubation or invasive mechanical ventilation) typically accompany the DNR order, but exceptions occur. Orders limiting aggressive care may also prohibit use of noninvasive (BiPAP) mechanical ventilation, intravenous drugs or infusions for cardiac arrhythmias with preserved circulation, cardiac pacemakers, or chest tubes, among other supportive devices indicative of aggressive care. However, it is important to keep in mind that a DNR order per se should not affect the level of care provided to the patient except obviously in the case of a cardiac arrest. Other restrictions of medical treatment or de-escalation of care should be specified separately from the DNR order.

These distinctions are crucial to avoid unintended problems with a DNR order. Some studies have found that DNR orders may negatively influence triage to the ICU. Some patients may feel that DNR order may impact aggressiveness of care. Multiple studies have found that certain cultures will see a DNR order as equivalent to withholding treatment. It should not. In some situations DNR could be the first step toward de-escalation of care. However, in itself DNR merely defines the limits of care.

In the United States surrogates are able to make decisions, and they could be guided by advance directives. A living will directs proxy to withhold or withdraw treatment at the end of life. A living will is usually formulated in broad terms (often containing a sentence such as, “if I have terminal disease I do not want to be resuscitated”), and obviously rarely includes specifics on acute neurologic disease. Decision makers for the patients therefore will have to interpret such a will. Nonetheless the mere fact a living will exists indicates that the patient has anticipated that a difficult medical situation may occur in the future. It expresses a wish by the patient to assist family members in making such decisions.

So what should we do in this situation? Providing factual information is the first course of action, and this requires a formal family conference (Table 29.1). Sitting down and having a conversation in a separate room is far more appropriate than a cursory discussion at the bedside. Physicians may need to use visual aids (showing the large destructive hemorrhage), establish trust under stressful circumstances, and may need multiple conversations which should include having the family summarize the assessment of the situation. Physician should respect cultural and religious beliefs, but these may be an impediment to rational medical care. In many families considerable time may be needed to grasp the finality of the condition.

TABLE 29.1 The Family Conference (10 Steps)

1. Sit down in a quiet place (separate room)

2. Identify yourself

3. Summarize recent developments

4. Proceed with a summary of the clinical course

5. Summarize the big picture and treatment goals

6. Estimate and describe disability

7. Discuss tracheostomy and gastrostomy

8. Discuss palliative care

9. Discuss code status

10. Answer questions

Explanations should relay specific information (Table 29.2). If the patient has worsened, the family should understand why the patient has worsened. In some situations a third party may be helpful in the discussions, and a medical ethics committee may be able to resolve differences if there is an emerging conflict between the patient’s family and the treating physician. In fact, withholding or withdrawing treatment is a common reason for consulting the medical ethics committee. Ethics consultants may be able to spend additional time with the family—lack of time is often a limiting factor in the communications in the ICU—explaining the issues at hand with equanimity and compassion. Yet, it remains to be seen whether ethics committees can defuse conflict once a very antagonistic family–physician relationship has developed.

TABLE 29.2 Information to Convey to Family Members When Discussing Do-Not-Resuscitate


Chest compression, defibrillation, intubation, mechanical ventilation, invasive catheters, medications; may mean 20–30 minutes with poor brain perfusion


2/3 survive resuscitation; 1/10 survive to discharge


Care otherwise not different
Cardioversion still option if needed
Aggressive ICU care continues (e.g., hemodialysis still optional)

How did we approach this quandary? The family in this particular patient example was told that it was very likely she was going to remain comatose and cardiopulmonary resuscitation could bring her heartbeat back, but she would not be able to recover important function needed to understand her situation. After daily conversations with the family about the patient’s condition, it became clear all of them understood the gravity of the brain injury. In these discussions long-term care using a tracheostomy and gastrostomy was brought up. The family eventually decided that long-term care was not in her best interest and a do-not-resuscitate order was placed. Several days later the family decided to withdraw the ventilator and to provide palliative care only.

The case illustrated here represents an extreme in the spectrum of severity of acute brain damage, but it is a common clinical scenario in NICUs. It was clear to all of us upon arrival that she could not recover, even with the most aggressive supportive treatment. Yet, several family conferences were necessary before the family accepted that their loved one would not regain consciousness. By the time they requested that life support measures be withdrawn, they were at peace with their decision and appreciative of the time we had spent with them and the care the patient had received.

Families expect an estimate of outcome. What we say as neurologists about prognosis and the way we deliver this message may greatly influence decisions on subsequent level of care. This responsibility must be accepted with full understanding of its weight.

When communicating a poor prognosis leads to limitations in the level of care or withdrawal of life support, a self-fulfilling prophecy may occur. This philosophy has received considerable attention in the literature over the last decade. Some have even provocatively claimed that the prognosis of a poor outcome might be the single factor most strongly associated with mortality in patients with intracerebral hemorrhage. Our prognostic abilities are imperfect, and we should be aware of this possibility. We should also be mindful of exaggerations or trivialization in the discussion of this important topic.

The point of no return, a condition incompatible with survival or meaningful recovery, is expected with severe brainstem injury, but defining the boundary of good, not too bad, and poor outcome has proven far more difficult than it seems. Useful indicators of poor prognosis—often loss of pontomesencephalic reflexes and coma—have been described, but almost invariably the studies supporting their predictive value have not accounted for the potential influence of withdrawal of life support or restriction of aggressive care measures. In other words, an intracerebral hematoma volume exceeding 60 cc has been consistently associated with high mortality and poor recovery, but death in these patients is often preceded by withdrawal of life support, and prognostic studies have rarely accounted for this caveat or provided more neurologic detail. Finding a solution to this limitation of studies on prognosis is very difficult because it would require analyzing a population of patients treated aggressively even after clinicians feel that such care has become frankly futile. It also would require very detailed analysis of the neurological condition. For example, a comatose patient with a destructive hemorrhage and persistent loss of several brainstem reflexes and extensor posturing is not expected to improve substantially, but a patient with spared brainstem reflexes might, and in such cases withdrawal of life support may be sometimes too premature.

The other important issue to keep in mind is that cognitive and physical incapacity rather than death is the outcome most feared by the great majority of patients and families. The key question we are asked by families is not whether the patient can survive the acute brain insult but whether survival can be followed by meaningful functional recovery (admittedly the word “meaningful” in this setting carries a certain arbitrariness). Studies evaluating the possible occurrence of the self-fulfilling prophecy using mortality as the main endpoint fail to address this point. In addition, even the most detailed statistical analysis may be insufficient to account for the effect of the combination of catastrophic brain disease and previous chronic illnesses in an elderly and previously debilitated patient. In all honesty we may not have the answer in all medical conditions, but it would be concerning—and create for physicians a truly unworkable environment—if the concept of self-fulfilling prophesy were to become a pretext to keep patients alive at whatever effort and cost.

Withdrawal of life support measures consists of extubation and discontinuation of any administered drugs. Central access and arterial catheters are removed, and the monitor is turned off. In our institution, we may start a morphine infusion of 0.1 mg/kg/hr and titrate to comfort by increasing infusion with small increments every 15 min until the patient is “comfortable.” We may institute a lorazepam infusion of 0.05 mg/kg/hr and titrate upward slowly until symptoms of agitation or restlessness are controlled. In deeply comatose patients none of this is indicated unless breathing after extubation becomes markedly labored. Most patients with brainstem injury die within hours after life support is withdrawn. Transfer to a palliative care room—in the event the patient remains stable—is desirable.


· Do-not-resuscitate orders may be warranted in catastrophic neurologic injuries.

· The appropriateness of cardiopulmonary resuscitation (full code) depends on the probability of good outcome and absence of life-shortening comorbidity.

· Repeated conversations with family members are extremely important, and time should be reserved to do so in an appropriate way.

· End of life care should include discussions with surrogates using a shared decision model.

· Decisions should include a broad picture of validated predictors of outcome, previous functional status, coexistent illnesses and patient’s preferences.

· Difficult situations may be resolved with the assistance from a hospital ethical committee.

Further Reading

Alexandrov AV, Pullicino PM, Meslin EM, Norris JW. Agreement on disease-specific criteria for do-not-resuscitate orders in acute stroke. Members of the Canadian and Western New York Stroke Consortiums. Stroke 1996; 27:232–237.

Becker KJ, Baxter AB, Cohen WA et al. Withdrawal of support in intracerebral hemorrhage may lead to self-fulfilling prophecies. Neurology 2001; 56:766–772.

Bernat JL. Ethical aspects of determining and communicating prognosis in critical care. Neurocrit Care 2004; 1:107–117.

Burns JP, Edwards J, Johnson J et al. Do-not-resuscitate order after 25 years. Crit Care Med 2003; 31:1543–1550.

Curtis JR, Tonelli MR Shared Decision-making in the ICU. Am J Respir Crit Care Med 2011;183:840–841.

Rabinstein AA. Ethical dilemmas in the neurologic ICU: withdrawing life-support measures after devastating brain injury. Continuum 2009; 15:13–25.

Rabinstein AA, McClelland RL, Wijdicks EFM et al. Cardiopulmonary resuscitation in critically ill neurologic-neurosurgical patients. Mayo Clin Proc 2004; 79:1391–1395.

Tian J, Kaufman DA, Zarich S et al. Outcomes of critically ill patients who received cardiopulmonary resuscitation. Am J Resp Crit Care Med 2010; 182:501–506.

White DB, Braddock CH III, Bereknyei S, Curtis JR. Toward shared decision making at the end of life in intensive care units: opportunities for improvement. Arch Intern Med 2007; 167:461–467.

Wijdicks EFN, Rabinstein AA. Absolutely no hope? Some ambiguity of futility of care in devastating acute stroke. Crit Care Med 2004; 32:2332–2342.

Wijdicks EFN, Rabinstein AA. The family conference: end-of-life guidelines at work for comatose patients. Neurology 2007; 68:1092–1094.