Ira R. Byock, American Journal Hospice & Palliative Care, March/April 1995.

Patient Refusal of Nutrition and Hydration: Walking the Ever-Finer Line

Ira R. Byock, M.D. 
American Journal Hospice & Palliative Care, pp. 8-13, March/April 1995. 

In the midst of an increasingly heated debate over physician-assisted suicide (PAS) another option available to patients who are determined to end their lives is receiving serious attention — the conscious refusal of nutrition and hydration. Patient refusal of nutrition and hydration (PRNH) is hardly new, indeed, virtually all hospice clinicians remember people who came to a point in their illness when they could be described as having “lost their will to live” and who recognized that continued eating and drinking was having an undesired, life-prolonging effect. In the hospice context, death that follows the decision to refrain from food or drink is not usually considered a suicide, however, by choosing to do so these patients were conscious that their death would likely be hastened. 

The general impression among hospice clinicians that starvation and dehydration do not contribute to suffering among the dying and might actually contribute to a comfortable passage from life. In contrast the general impression among the public and non-hospice medical professionals is that starvation and dehydration are terrible ways to die. Scientific support for either viewpoint has been scanty, yet modern medical practice has reflected an aversion to allowing a person to “starve to death.” 

Indeed, during the era in which most hospice providers have trained and practiced, a patient unable to eat has been routinely treated with a feeding tube; the option of declining such intervention never having been offered or fully considered. The symbolic importance of offering food and fluids is well-recognized. While it has been utilized by people throughout human history, in public discussion and debates regarding physician-assisted suicide, hospice providers have wisely avoided suggesting PRNH as an alternative. There has been concern that in the political arena such a suggestion might appear as a self-serving way to deny hospice providers’ “ultimate responsibility” to the suffering patient. 

But the situation may now be changing. Several recent articles are serving to dispel fears of suffering and are making it more acceptable to speak more openly about this inherent ability of patients to influence the timing of their demise. Late in 1993 an article entitled, Patient Refusal of Hydration an Nutrition: An Alternative to Physician-Assisted Suicide or Voluntary Active Euthanasia, by Bernat, et. al. in the Archives of Internal Medicine reviewed the salient clinical literature and discussed the ethical implications of this option.[Bernat] The authors include PRNH within the ethical category of “voluntary passive euthanasia” since it involves not only the refusal of oral food and fluid but the associated refusal of non-oral (enteral or parenteral) alimentation and hydration. They assert that the critical moral and legal distinction to be made regarding a life-ending decision is not whether it involves an act of commission or omission on the part of the physician, but whether or not it constitutes the refusal of a medical therapy by a competent patient. Patient refusal of nutrition and hydration meets this criteria and, thus, can be considered among the commonly accepted practices of patient-initiated refusal (or withdrawal) of mechanical ventilation, renal dialysis, or antibiotic use. 

A more extensive review of the scientific literature relevant to starvation and dehydration appears in an article by Sullivan entitled, Accepting Death without Artificial Nutrition or Hydration.[Sullivan] Published studies of healthy volunteers report that total fasting causes hunger for less than 24 hours. Ketonemia occurs and is associated with relief of hunger and an accompanying mild euphoria. When ketonemia is prevented by small feedings hunger persists, explaining the obsession with food commonly observed during semi-starvation occurring in times of famine or war. Animal studies also suggest that ketonemia may have a mild systemic analgesic effect. Experimentally induced dehydration in normal volunteers may report thirst, yet this sensation is consistently relieved by ad lib sips of fluid in cumulative volumes insufficient to restore physiologic fluid balance. One study of healthy subjects suggests there is a decrease in the severity of experienced thirst associated with older age. 

Recently, two important clinical studies have been published which investigate the effects of fasting and dehydration in the patient population relevant to hospice. Burge surveyed patients at two inpatient palliative care units in Canada. Visual analog scales (100-mm) were used to assess seven symptoms: thirst, dry mouth, bad taste, nausea, pleasure in drinking, fatigue and pain. [Burge] Reported symptoms were studied in relation to potential confounding variables. Thirst was considered to be the principal outcome of interest. The mean symptom rating for thirst was 53.8 mm. In multiple linear regressions no statistically significant association between thirst and fluid intake, serum sodium, urea or osmolality was found. The presence of oral disease yielded the most significant association between thirst and examined variables. 

In the October 26 issue of JAMA, McCann, et. al. report a prospective study of patients in a ten-bed “comfort care unit” located within a long-term care facility. [McCann] The physical care described closely resembles benchmark hospice care. “Food was offered and if necessary fed to patients but was never forced. All patients received meticulous mouth care that included combinations of cleaning, various swabs, ice chips, hard candy, and lubricants. Narcotics were used for most of the patients to treat symptoms of pain and shortness of breath when present. The dose of narcotics was titrated to provide pain relief while avoiding sedation. When the window of providing pain relief and causing sedation was small, the patients’ wishes were weighed regarding the discomfort of pain vs. the discomfort of sedation in determining subsequent doses and intervals of narcotic administration.” 

Of 32 patients assessed by this group during a 12 month period, 63% denied hunger entirely, while 34% reported hunger during only the initial part (first quarter) of their course in the unit. Thirst or dry mouth was reported by 66% of patients; 28% transiently and 38% throughout the terminal phase of their illness. Thirty-four percent denied either symptom. The authors found that in all patients reporting either hunger or thirst, these symptoms were consistently and completely relieved by oral care or the ingestion of small amounts of food and fluid. While patients could eat or drink ad lib, the amount of food or fluid ingested — and documented to relieve associated symptoms — was consistently less than that required to correct dehydration or to meet obligate fluid and energy requirements. 

Perhaps the most persuasive of recent articles is that entitled, A Conversation with My Mother.[Eddy] It is a narrative written by Dr. David Eddy regarding the progressive illness and dying of his mother. Initially published in the Journal of the American Medical Association, it was subsequently reprinted in the New York Times, eliciting substantial discussion and notably favorable public response. Mrs. Eddy was suffering from progressive debilitation, chronic depression, anemia, recent surgery and recurrent rectal prolapse. She expressed a desire to die and, in the course of relentless decline, asked her son for help. Dr. Eddy sought to provide his mother with the means to end her life peacefully. However, prior to obtaining a lethal prescription, she developed pneumonia and was hospitalized. Antibiotics were begun (we are not told why), but quickly withdrawn at the patient’s request. When she began to improve despite the lack of life-prolonging intervention, Mrs. Eddy asked her son about the option of refusing food and fluids. (It was her idea.) He assured her that without nutrition and, especially without adequate fluid, the end would come quickly. She was elated and, following the celebration of her 85th birthday and with the support of her primary physician, she stopped eating and drinking. (Her last morsel was chocolate.) She died, peacefully, six days later. The description of her last few days is compelling. “Over the next four days, my mother greeted her visitors with the first smiles she had shown for months. She energetically reminisced about the great times she had had and about things she was proud of… She also found a calming self-acceptance in describing things of which she was not proud. She slept between visits but woke up brightly whenever we touched her to share more memories and say a few more things she wanted us to know. On the fifth day it was more difficult to wake her. When we would take her hand she would open her eyes and smile, but she was too drowsy and weak to talk very much. On the sixth day, we could not wake her. Her face was relaxed in her natural smile, she was breathing unevenly, but peacefully. We held her hands for another two hours, until she died.” 

Sullivan’s aforementioned review of the literature also begins with a clinical narrative case presentation. The case is that of a 78 year old woman with recurrent, complete small bowel obstruction, following treatment of endometrial carcinoma. Although no evidence of recurrent or metastatic disease could be found, the patient steadfastly refused further surgery. While accepting a nasogastric tube for decompression, she declined any nutritional support and, after an additional 14 days, also refused intravenous fluids. The patient judged her symptoms, which included nausea and ascites, to be intolerable. Sullivan states, “She repeatedly requested that her life be ended by injection of a lethal dose of morphine. This request was respectfully declined by her physician, who did offer to relieve any pain or discomfort.” Eventually, morphine was administered at the request of the patient “to relieve boredom and help with sleep.” The patient lived a surprisingly long time; a total of 42 days of complete fasting and 29 days with minimal fluids. At no time did she report pain. Her sensorium remained clear and affect normal until the last day of life. “Her friends visiting during this period uniformly confirmed that she was a woman who knew her own mind and had the courage of her convictions.” 


Critical examination of this latter case gives rise to a number of questions. Seasoned hospice clinicians will wonder whether sophisticated palliative care would have uncovered other options available to this patient. However, two points of relevance to the current discussion stand out: a) neither starvation nor dehydration seemed to contribute to this patient’s physical distress, and b) the patient lived for an unexpectedly long time. In addressing the prolonged course of his patient, Sullivan emphasizes that she was not known to have an underlying terminal illness or any ongoing catabolic process. Additionally, she began her fast with apparently normal nutritional stores (she is described as having been “always mildly obese”) and her refusal of fluid followed two weeks of intravenous hydration. 

This patient’s condition contrasts with the large majority of hospice patients who are likely to undergo a shift from adipose to protein metabolism much earlier in the course of fasting. This shift coincides with a decrease in metabolic free water and renal function and a rapid progression to uremia, electrolyte imbalance, somnolence and demise. A more prolonged course might be expected for a person in the circumstance of a massive stroke or post-hypoxic syndrome whose advanced directive prohibits non-oral nutritional support or hydration. 

In reviewing the literature on thirst associated with negative fluid balance derived from studies of healthy volunteers as well as studies involving those with far advanced illness, I have gained an impression that the word “thirst” may well refer to two distinct sensations. In the patients whom hospices serve, “thirst” most often refers to a sensation of dryness in the mouth and throat rather than an experienced need to ingest a volume of fluid which it normally conveys. This explains why the “thirst” of persons with far-advanced illness is consistently relieved by amounts of fluid insufficient to expand volume or osmotic receptors. 

Risks and Potential Benefits of Starvation and Dehydration in Far-advanced Illness

While the currently available research is not exhaustive, hospice clinicians now possess respectable scientific data to supplement anecdotal experience in forming their own opinions, and informing patients about nutrition and hydration. This information has relevance beyond responding to a hospice patient who is expressing suicidal intent. When diminished gag reflex, altered sensorium or obstruction makes oral feeding impractical or risky, thoughtful clinicians and their patients often struggle with decisions about whether to place nasogastric or PEG feeding tubes. 

Medical decision-making rest upon the pillars of informed consent and the concept of proportionality. The publications cited collectively expand the information base on which clinicians can make recommendations and patients (and families) can make informed choices. Proportionality is commonly explained as a weighing of the risks versus the potential benefits of a proposed intervention. In the context of the present discussion, the principle of proportionality can be understood more fully as requiring the weighing of the risks and potential benefits of each available intervention against the risks and potential benefits of non-intervention. In the process of informing and obtaining consent from patients — and, perhaps especially, in discussions with the legal surrogate(s) of an incapacitated patient — it must be remembered that the decision to prevent malnutrition or dehydration is a de facto decision to have the person die of something else. Stated another way, the euvolemic, nutritionally supported patient may live longer, but is probably more likely to die from acute systemic infection, from acute respiratory failure, from acute cerebral or myocardial ischemia or from abrupt blood loss. 

These clinical reviews, case reports and new research data lend credence to the clinical impression that, among the terminally ill, the risks of uncorrected malnutrition and dehydration are few. (In this circumstance, death is not properly regarded as a risk, since it is a principal expected outcome.) Considered together, they allow us to state that, at least within the context of adequate palliative care, the refusal of food and fluids does not contribute to suffering among the terminally ill. The literature is consistent on two points: a) rarely does fasting cause any discomfort beyond occasional and transient hunger, and b) symptoms referable to dehydration are few — mostly dry oral and pharyngeal mucous membranes — and are readily relieved by simple measures. 

Yet, while it may now be appropriate to present the risks of starvation and dehydration within a hospice setting as being minimal, this is logically distinct from being able to assert that there is tangible benefit from either condition. From a patient’s perspective, I suspect that the critical, intangible benefit of PRNH will be an improved sense of confidence that death will occur peacefully. As noted, this notion is supported (and probably fostered) by the commonly expressed observations of hospice clinicians that such deaths are generally preceded by a gentle, deepening somnolence with the person often described as having “slipped away”. What is known of the physiology of far-advanced, progressive disease and of combined nutritional and fluid deprivation suggests a synergism that, indeed, might explain a peaceful pattern of demise. This may occur through gradual organ dysfunction (especially renal and hepatic) with progressive metabolic derangements or via primary circulatory failure. In either case, a clouding of sensorium can be expected with an arrhythmia likely being the terminal event. While currently available studies do not refute this clinical impression, neither are they sufficient to substantiate it. Thus, while the heightened probability of a gentle passing by PRNH is intriguing, at present, it remains speculative. 

Another potential benefit merits discussion. At least for some persons, starvation does correlate with reported euphoria. While there is a tendency for medical professionals to dismiss euphoria as an aberration of consciousness — a non-distressing “symptom” — it clearly has positive impact on a person’s subjective quality of life. Quality of life, it should be remembered, is a wholly subjective construct. It has emerged as one of the “gold standards” of outcome measures for hospice care. The question of whether euphoria is somehow “valid” or “reality-based” may prove irrelevant to its implications for hospice practice. Further prospective physiologic studies with clinical correlates of affect and meaningful quality of life assessments should be actively encouraged. 

Ethical and clinically effective response to requests for assistance in suicide 
The practice of a physician or nurse responding to a patient’s request for assistance in suicide by discussing the option of refusing to eat or drink appears to stretch the “fine line” of ethical practice to the point of disappearing. It will seem chilling to some, and self-absorbed hair-splitting to others. 

Perhaps all patients with essentially normal mental status who are aware that they are dying engage in some degree of suicidal ideation. The request from a patient to their doctor or nurse for assistance with suicide has many meanings, at a minimum it must be heard as a call for help. It may represent an urgent plea for symptoms to be better controlled or for the patient to be listened to as a person — for their suffering to be heard. In practice serious requests for assistance with suicide commonly prove to be an opening to deeply important clinical work. Thus, while suicidal ideation is common, suicide among hospice patients is rare. 

In a perfect world resources and expertise would be unlimited and this type of therapeutic response — combining effective symptom control with expert counseling and an authentic healing presence — would suffice. However, in the imperfect world in which we practice, not every patient achieves an acceptable quality of life, nor is every patient willing “to work through” his or her suffering. Some people insist that they have “had enough” and continue to plead for assistance with suicide. The information reviewed has pragmatic application to a response that remains ethically consistent, effective and genuinely caring. 

Ethically, while a clinician may decline to actively assist a person’s suicide, we must never abandon our patient. Even a patient who is intent on suicide continues to deserve our care. At a minimum the patient requires continued attention to control of symptoms. Further, the patient remains entitled to accurate medical information about the options available to them. Undeniably, one such option is the refusal of further food and drink. But it is, indeed, a fine line. Unlike physician-assisted suicide, refusing to eat or drink is a purely personal act. While it may require information, the decision obviates the need for physicians, nurses or other agents of society to participate. After adequate discussion, and in the context of continued caring, at some point the patient’s choice becomes “none of our business”. 

Seeking a clear philosophical distinction, between the scenario of one patient swallowing a prescribed lethal potion and the scenario another refusing to swallow anything at all, may be an exercise in splitting ever-finer ethical hairs. Yet, the interpersonal meaning of these two acts could not be more different. In every circumstance — legal or not — acceding to a patient’s request for a lethal prescription entails a complicity on the part of the clinician. In its meaning the act is a collusion in the patient’s belief that their situation is hopeless and that their existence is beyond conceivable value. 

In my own practice, while I steadfastly refuse to write a prescription with lethal intent or otherwise help the patient commit suicide, I can share with the patient information that he or she already has the ability to exert control over the timing death. Virtually any patient with far-advanced illness can be assured of dying — comfortably, without any additional physical distress — within one or two weeks simply by refusing to eat or drink. This is less time than would be legally imposed by waiting periods of assisted suicide initiatives. The discussion and subsequent decision are wholly ethical and legal, requiring no mandated psychiatric evaluations, attorneys, court decisions or legislation. On the level of meaning there is wisdom in not discussing the option of PRNH too soon. The impetus for the discussion should arise (as it did for Mrs. Eddy) from the person dying. 

Clinically, for a number of people at the very end of life, the decision to refuse food and fluid may not arise from depression or emotional denial as much as from a felt sense of “being done”. Most such persons I have encountered one way or another expressed a sense that eating or drinking were no longer relevant to their situation. They were far along in a process of withdrawal, having turned their attention inward or “beyond”. Even here the option of PRNH has important advantages over complying with a patient’s request to be killed, for it allows the clinician’s attention to remain focused on relief of suffering — physical, psychosocial and spiritual. It requires — of frees — the clinician to remain vigilant for treatable depression and to remain, in humility, open to the possibility of unexpected opportunities for the person to again discover value in the life that is waning. 

The powerful symbolism of nurturing associated with feeding and the notion of suffering associated with starvation and dehydration are deeply rooted. This is perhaps especially true for those of our patients who lived through world wars and the Great Depression of this century. While the topic of PRNH may no longer be taboo, it must be approached with extraordinary sensitivity. However, discussion of what we know can help reduce anxieties of what we fear. The more we know and the more confidently we know it, the better able we are to make sound decisions. My experience is that the information presented has been well received and resulted in a noticeable allaying of fears. This has been the case in private discussions, as well as in public and professional forums. The recognition that patients with far-advanced illness have always had control over the timing of their demise can enable the focus of discussion and intervention to remain fixed on the goals of comfort and quality of life. 

The debate over physician-assisted suicide will continue. The option of PRNH deserves fuller attention and discussion. Years ago Cicely Saunders defined a core principle for hospice, “You matter because you are you. You matter to the last moment of your life, and we will do all we can not only to help you die peacefully, but also to live until you die.” In our statements and our actions hospice clinicians would do well to reflect on those words.


Bernat, JL, Gert, B, Mogielnicki, RP, Patient Refusal of Hydration an Nutrition: An Alternative to Physician-Assisted Suicide or Voluntary Active Euthanasia, Arch Intern Med Vol 153, Dec 27, 1993 pp2723-2728 

Sullivan, RJ. Accepting Death without Artificial Nutrition or Hydration J Gen Intern Med, Vol. 8, April, 1993, pp 220-224. 

Burge, F, Dehydration Symptoms of Palliative Care Cancer Patients. Journal of Pain and Symptom Management, 1993;8(7) pp454-464. 

McCann, RM, Hall, WJ, Groth-Juncker, A, Comfort Care for Terminally Ill Patients: The Appropriate Une of Nutrition and Hydration, JAMA October 26, 1994 272(16) pp1263-1266 

Eddy, D. A Conversation with My Mother, JAMA July 20, 1994 Vol. 272, No. 3 pp179-181 

Patient Refusal of Nutrition and Hydration: 
Walking the Ever-Finer Line 
American Journal Hospice & Pal Care, pp. 8-13, March/April 1995. 
Ira R. Byock, M.D.