1. Is there a chance that medical intervention will be effective in achieving the patient's treatment goal?

In general, medical decisions should be based on broad goals of care and the patient should—with the physician's support—play a key role in determining what those goals shall be. Therefore, the treatment under consideration has to be evaluated with regard to its ability to achieve the patient's treatment goal. Whether the treatment goal should be a cure, prolonging life or comfort care needs to be negotiated between the patient, family and the provider, and might change over the course of the disease. Offering treatment that is ineffective with respect to a certain treatment goal violates the rules of good clinical practice. However, in practice, predicting the outcome of an intervention can be difficult, because general success rates in defined study populations do not allow for prediction with certainty about the effectiveness in the specific medical situation of an individual patient. Furthermore, effectiveness as a criterion provides an incentive for setting a statistical threshold, as most treatments are effective only in a certain percentage of cases. Schneiderman et al suggested that physicians should not offer a treatment if empirical data shows that the treatment had a <1 in 100 chance of benefiting the patient.14 Apart from the difficulties in breaking down the statistics to individual patients and their individual risk profiles, this threshold might not be accepted by a substantial portion of patients who think that a 1% chance of success is still better than having no chance at all.15

Nevertheless, it is still an important first step to separate the rare cases (in which the treatment will be ineffective) from those cases in which success might be against the odds, but not entirely impossible. In the former case, an intervention can be ineffective either because a treatment does not work physiologically for the stated purpose, or has already proven to be ineffective in the individual patient.16 An example is a patient's request for continuing a chemotherapy regimen during which tumour growth was excessive, even though continuing the chemotherapy would neither prolong the patient's life nor relieve tumour-related symptoms (physiological futility). If there is no chance that the requested intervention will achieve the patient's treatment goal, the physician should not offer the intervention. Instead, the physician should initiate a discussion with the patient and the family about alternative and more realistic treatment goals. This requires that the patient receives full and open information about his or her medical situation and the prognosis according to different treatment options. It is often observed that, over time, patients tend to adjust their unrealistic treatment goals.17

2. How does the physician evaluate the benefit-harm ratio of the treatment?

If there is a chance that the requested treatment will achieve the patient's treatment goal, then the ethical obligations of beneficence and non-maleficence require that the physician initially evaluates the benefit-harm ratio independently of the patient's wishes. This evaluation demands not only medical expertise, but also value judgements about acceptable benefits and harms. Based on this evaluation, the physician can then enter into further discussions with the patient about treatment goals and the benefits and harms of different interventions from the patient's perspective.

While in theory the benefit-harm ratio is a continuous variable ranging from a clear net benefit to a clear net harm, it seems helpful for our purposes to distinguish three different prototypical situations that have important implications for further steps in evaluating the patient's request for a certain treatment (see figure 1, question 2):

• (A) Benefit > harm: If the benefits outweigh the potential harm, the physician should offer the treatment requested by the patient, even if the net benefit is small. In this case, the physician's beneficence-based obligations converge with the patient's preferences. An example could be a therapy with moderate side effects that might prolong the patient's life for another 2 weeks.

• (B) Benefit = harm: There are cases in which benefits and harms seem to be in balance, either because the expected benefit is not very likely to be realised or because the harms counterbalance the benefits.

• (C) Harm > benefit: The harms exceed the expected benefits of the intervention. This would be the case, for example, if a patient with progressive cancer disease under third-line chemotherapy wanted to switch to another chemotherapy regimen with considerable toxicity or risk of immunosuppression, in which the new regimen is very unlikely to change the course of the disease (qualitative futility).

In the last case (C) there is a conflict (and in the second case at least a tension) between the physician's obligation to respect patient autonomy on the one hand, and the physician's obligations of beneficence and non-maleficence on the other hand. How can the physician respond to this conflict? First, identifying the source of the dissent will help the physician to understand why the patient prefers active treatment.

3. Does the patient assess his or her medical situation in a realistic way?

The source of the dissent between physician and patient about what is the best course of treatment can either be located on a factual level (understanding of the medical situation) or on a normative level (balancing benefits and burdens). If the patient has a realistic understanding of the medical situation, the physician should enter into a deliberation with the patient about appropriate treatment goals and the benefits, risks and costs of the different treatment options (see question 4).

There are two main reasons why patients may not assess their medical situation in a realistic way: insufficient information or denial. Patients are often not well informed about their situation and tend to see their prognosis in overly optimistic terms.18 Reasons for this misunderstanding certainly can be that physicians overestimate survival, fail to explicitly explain the prognosis or palliative goal of the treatment, or do not involve patients in decisions on limiting treatment.3 ,19 ,20 The physician should, therefore, explain the situation, provide prognostic information and psychological support to enable the patient to develop a more realistic understanding of his or her situation.

In some cases, however, the unrealistic assessment of the situation is a consequence of a denial reaction by the patient.3 To date, the definitions of denial lack consensus as to whether denial is unconscious or conscious, whether it is a state or a trait, and whether it is a pathological mechanism of ineffective self-defence or an adaptive strategy to protect oneself against overwhelming events and feelings.21 Still, all concepts imply that a coping strategy of denial leads to an inaccurate assessment with respect to feasible treatment goals. Certainly, many patients become better at coping with their disease—oftentimes with the support of psychological or palliative care consults.22

However, the more the harms of an intervention outweigh its benefits, the higher are the demands on patients' informed understanding of their situation. We therefore suggest two different strategies depending on the physician's benefit-harm evaluation, if the patient is in persistent denial despite counselling and psychological support.

• (A) If the patient runs the risk of having his or her denial cause more harm than good, the physician has good reasons to give the beneficence-based obligations priority over the patient's wishes, which will justify a unilateral decision not to offer the requested treatment. It is one of the physician's primary obligations to protect patients from burdensome and potentially harmful decisions that are not based on a realistic understanding of the situation—and which the patients might regret afterwards when they realise that it was based on false hopes. We are aware that the denial argument may be abused to ignore the patient's wishes because they deviate from the physician's recommendation. It is therefore important for a psychologist or psycho-oncologist, who is experienced in assessing coping behaviours and defence mechanisms, to evaluate the denial. This evaluation is often based on whether the patient can or cannot address his or her need for the denial of the informed understanding.23 In addition, it can be helpful to involve the family in order to learn about the patient's hopes and values. With this information, the physician might succeed in developing goals with the patient in the near future that can realistically be achieved.

• (B) If the patient is in denial, and harms and benefits are just about balanced— according to the physician's judgement—the beneficence-based obligations to withhold the requested treatment are not as strong. Therefore, the physician has more discretion whether he complies with the patient's wishes for active treatment or not. In this situation, it seems ethically legitimate to take the cost of care into consideration (elaborated in question 5).

4. Does the patient prefer treatment after evaluating the benefit-harm ratio?

If the patient can assess his or her situation in a realistic way the physician should explain their own benefit-harm assessment (see question 2). The physician should empower the patient to not simply follow unexamined preferences, but instead to make a well-informed decision that is coherent with the patient's values and preferences. Whether the low odds of the benefits justify a trial of treatment, or whether an effect counts as a genuine benefit, involves value judgements. Studies show that a person's assessment of treatment options change as the person's health deteriorates: patients with advanced diseases are much more likely to opt for burdensome treatment with a minimal chance of benefit than healthy people would, including medical and nursing professionals.24–26 Engaging in a joint deliberation about the worthiness of health-related values and outcomes (a deliberative model of the physician–patient relationship27) shall enhance patient autonomy: If the patient still prefers the requested intervention after this deliberation, the physician has good reasons to comply with the patient's wishes. However, if the intervention requires a lot of resources (either financial or personal), the physician should inform the patient about the resource consumption of different treatment options, so that resource consumption can become part of the patient's deliberation.

5. Is resource consumption relevant to the decision?

All healthcare systems face the challenge of increasing demands for healthcare within limited financial budgets. Assuming that setting limits is inevitable, an explicit, rule-based rationing process is preferable to implicit, case-based rationing at the bedside.28 Still, even the best consensus-based rationing guidelines will not be able to completely eliminate bedside rationing, since guidelines will never cover all medical areas and will have to be interpreted and applied to each individual case. Actually, some discretion appears to be desirable, for it allows the physician to do justice to the particularities of individual cases.29

It therefore seems ethically mandated to consider the required resources in medical decision making—to protect the resources of the individual, third-party payers or the community from ill-founded spending. While it is uncontroversial that physicians should always minimise costs while realising the goal of care, it is debated on what grounds patients and physicians should refrain from costly treatments that might result in only a minimal benefit for the patient.30–32 Here, we delineate situations in which autonomy-based reasons prevail (A, below) from those in which cost considerations could play a decisive role in the decision (B, below).

• (A) If a patient with a realistic understanding of the medical situation requests a treatment that—in the physician's judgement—has a questionable or nil net benefit, the physician should inform the patient that everybody in the healthcare system bears some responsibility for cost-sensitive decisions. The physician should point out to the patient that the requested treatment does provide only a small or no net benefit in this situation, and a lot of resources could be saved if the patient would not insist on the treatment. As a result, competent patients are invited to factor costs into their deliberations, however, they should ultimately be free to opt for the expensive treatment strategy. Hence, in this case, sound autonomy-based reasons take precedence over cost considerations at the bedside.

• (B) If patients lack a realistic understanding of their medical situation, and the benefit-harm ratio of the treatment is neither clearly positive nor clearly negative, however, the required resources could play a decisive role. As long as the treatment is inexpensive, the physician may offer it, even if it provides a questionable benefit. However, if the treatment draws heavily on the resource pool (financial, material or staff), cost considerations may justify a unilateral decision to withhold the requested treatment since resource consumption is counterbalanced against neither a clear net benefit nor by sound autonomy-based reasons on behalf of the patient.

It is controversial whether clinicians should inform their patients when they withhold interventions based on cost considerations, and we also acknowledge that healthcare systems differ when speaking of the cost of care, especially with respect to different cultures. For example, physicians in Europe report that they did ration health services on the basis of some small expected benefit, low chances of success and low quality of life.33 Problematically, it is often done in a covert and unpredictable manner. A more explicit process, like the one proposed here, contributes to the transparency of the underlying reasons of a decision against treatment, and to the patients' awareness about the costs of the requested care. In the long run, physicians should establish a ‘culture’ of responsibly discussing cost issues with their patients.34 However, implementing this last step of considerations requires a broader societal consensus about the physician's role in making resource-sensitive decisions at the bedside.