Analysis

Ethics and effectiveness: rationing healthcare by thresholds of minimum effectiveness

BMJ 2011; 342 doi: http://dx.doi.org/10.1136/bmj.d54 (Published 17 January 2011) Cite this as: BMJ 2011;342:d54
  1. Alena M Buyx, assistant professor/director12,
  2. Daniel R Friedrich, research associate1,
  3. Bettina Schöne-Seifert, professor and chair of bioethics1
  1. 1Institute for Ethics, History and Philosophy of Medicine, University Hospital Münster, Germany
  2. 2Nuffield Council on Bioethics, London WC1B 3JS, UK
  1. Correspondence to: A Buyx abuyx{at}nuffieldbioethics.org
  • Accepted 10 November 2010

Alena Buyx, Daniel Friedrich, and Bettina Schöne-Seifert argue that basing rationing decisions on clinical effectiveness rather than cost effectiveness would be fairer and could make it more acceptable

Scarcity of healthcare resources calls for fair, acceptable, and ethically justified ways of allocating and rationing care. Particularly in financially difficult times, this is a formidable challenge. So far, no developed country has managed to introduce criteria for fair rationing that have remained undisputed. This is not surprising. Rationing healthcare by definition goes beyond eliminating waste1 and thus requires difficult moral choices.2 Whoever gets afflicted, it almost always hurts, and cuts must therefore have sound ethical justification. Cost effectiveness in particular—one of the main criteria used by the National Institute for Health and Clinical Excellence (NICE) in funding decisions—has long been attacked for ethical reasons.3

One simple criterion for rationing that has been paid little attention is minimum effectiveness thresholds. Although public debates show increasing reservations about expending enormous effort for very small clinical benefit,4 medical interventions with only minimal effects seem to be part of today’s regular medical practice. For example, some of the newly approved monoclonal antibody drugs for end stage cancer at best prolong patients’ lives for a few weeks or months without any substantial improvement in their health related quality of life. Reasons for the prevalence of such treatments are manifold: in many Western countries, doctors prescribe the drugs because they are (often wrongly) afraid of legal liability or dread to admit that “there is nothing more we can do”; patients dwell on unrealistic hopes; and relatives have trouble facing their loved one’s end.5 6

We think it is irrational and wrong to ration clearly effective treatments while offering others that promise very low individual benefits. Before other cuts are considered, medical interventions with only minimal effectiveness should be excluded from publicly funded healthcare.

Minimum effectiveness thresholds: a proposal

Ultimately, all therapeutic and most other medical treatments aim either to prolong patient survival or to improve their wellbeing. Of course, individual patients may make different subjective evaluations of the clinical benefits, which they can express in their informed consent or dissent to treatment. Nevertheless, healthcare systems rightly assume that most patients welcome extra life years and positive effects on various aspects of health related quality of life. Thus the medical interventions offered within a health service should, at least in theory, be determined on the evidence that they produce such objective clinical benefits. This is captured by measuring a treatment’s direct clinical effectiveness—that is, its positive causal effects on survival or measures of health related quality of life (reduction of pain, improvement of mobility, etc). However, although most health technology assessments measure effectiveness, many studies choose other outcome measures, such as changes in tumour size, detection rates, or physiological effects. We believe that these measures are relevant only in so far as they affect patient survival or quality of life and, therefore, medical interventions should primarily be evaluated by their effects on these two outcomes.

Our suggestion is to cut any intervention from publicly funded healthcare that does not reach a predefined threshold of direct clinical effectiveness and which is in this sense minimally effective. This “razor” should apply not only to drugs but also to last resort surgery, behavioural interventions, and medical technology such as positron emission tomography/computed tomography.

Determining thresholds

An ethically appropriate threshold of minimum clinical effectiveness could be, for example, increasing an aspect of health related quality of life by at least 10% or prolonging patient survival by at least three months compared with established treatment or placebo. If three months of extra survival in end stage cancer treatment is accepted as a relevant threshold, drugs such as erlotinib, which produces mean additional survival of six weeks compared with placebo, would not be offered in public healthcare systems (box 1).

Box 1: Erlotinib: an example of potentially minimally effective treatment

Erlotinib is an anticancer drug that blocks epidermal growth factor receptors found on the surface of some tumour cells. Erlotinib was approved for two indications by the European Medicines Agency in 2007 (with gemcitabine for metastatic pancreatic cancer and as second line treatment for non-small cell lung cancer),7 but its inclusion in public healthcare systems has remained controversial.

Data show that in metastatic pancreatic cancer, mean overall survival was 8.1 months among patients given erlotinib plus gemcitabine compared with 6.7 months among those given placebo plus gemcitabine (95% confidence interval 0.17 to 2.66; (table)).8 Compared with placebo, erlotinib produced a mean life extension of less than 6 weeks (median overall survival was worse). Since the upper limit of the confidence interval is 2.66 months, maximum life prolongation is reasonably assumed to be below three months. In patients with locally advanced pancreatic cancer, there was no significant overall life extension (mean survival 10.7 v 10.5; 95% confidence interval −2.43 to 2.69).

Comparison of mean and median survival of patients with pancreatic cancer treated with erlotinib plus gemcitabine or placebo plus gemcitabine8

View this table:

If we assume three months’ life extension as a threshold of minimum effectiveness, a treatment producing maximum life extension of less than three months and mean extension less than six weeks is minimally effective. Erlotinib is very expensive. Cutting the treatment from German healthcare in 2008 alone—the year after it was approved—would have generated savings of around €57m (£48m; $74m).9

Setting concrete thresholds is of course a matter of some discretion. Therefore, they need to be set transparently by democratically legitimised institutions. In any case, they are backed by acceptable ethical considerations. The first one is that patient sacrifices resulting from this approach would always be relatively small. Small losses are ethically more acceptable than large losses. To ensure that the benefits forgone are small we suggest cutting only treatments that have a low median benefit and narrow confidence intervals. The confidence interval, at least in studies with high statistical power, is a surrogate for whether patient responses are spread narrowly or widely. In cases with a wide confidence interval, although the median effectiveness might be low, a few patients might respond very well, and cutting the treatment would deprive these patients of substantial benefits.

The second consideration is that rationing by minimal effectiveness is blind to the costs of specific treatments. In contrast to cost effectiveness analysis, our thresholds would not depend on calculating costs and identifying cost per value. Rather, as long as the general effect of rationing by minimal effectiveness was to contain costs—and we believe this is very likely—minimum effectiveness thresholds would apply to any treatment, be it expensive or not.

Fairness

Although rationing, if pursued transparently, can never be painless, use of our criterion would arguably distribute losses in the fairest possible way. By applying the effectiveness “razor” systematically to expensive and cheap treatments alike, small losses are spread over many people and losses of clinical effects are more equitably distributed, thus avoiding the danger that some unlucky minority is falling victim to unfair cost containment. By contrast, other rationing criteria such as personal responsibility, age, or cost effectiveness might require large sacrifices from individual patients.

Rationing on the basis of age, income, or personal responsibility all rely on a particular trait or behaviour of a patient. Because the choice of these traits and behaviours is often regarded as arbitrary, some critics have equated rationing by age or behaviour with discriminatory practices such as sexism or racism.10 11 In addition, some of the above groups might rightly be considered to be singled out for “victim blaming.”12 In contrast, minimum effectiveness thresholds rely exclusively on treatment aspects—namely, the magnitude of clinical effects on survival or quality of life. Therefore, we expect far fewer cases of individual disadvantaging. Of course, even our criterion would not reverse or compensate for health inequalities resulting from social determinants or bad luck in the natural genetic lottery—but at least it would not aggravate them.

A fundamental ethical problem of most cost effectiveness analysis is the need to multiply two different variables in calculating quality adjusted life years (QALYs)—namely, patient survival and health related quality of life. However, individual patients might make very different trade-offs between quality and length of life. Some may value a slightly longer survival highly even though their quality of life is low, while others may find this alternative very unattractive. Minimum effectiveness can be measured using these two variables independently. If, for example, a treatment does not meet the threshold of minimum patient survival but does meet the quality of life thresholds, it would not be cut.

Public acceptability

Another advantage of our approach is that, unlike cost effectiveness, costs do not figure in the process of evaluating individual treatments. The argument is not that a treatment is too expensive and, by extension, that the life it prolongs or enhances is not worth spending this money on. It is rather that a treatment is not effective enough—that is, the effects are too small—to be worth any public money. This is likely to make it more acceptable to the public.

Furthermore, rationing by minimum effectiveness does not involve an aggregation of benefits. In cost effectiveness analysis, effects are added up without regard to the number of people involved and the sums of these effects are then weighed against each other. Our approach, while of course relying on statistical data to predict clinical effectiveness, does not “pit” patient groups of different sizes against each other, so thus sidesteps the ethical problems of ignoring individual rights and interests.

Finally, excluding interventions from public funding based on their high cost might involve substantial sacrifices for individual patients—for example, by withholding expensive yet potentially effective life saving interventions such as rescue helicopters. Our suggestion would limit patient sacrifices to those of a small magnitude, which is more tolerable for all parties.

Implementation

Effectiveness data are becoming available for most new treatments as well as for many old and established treatments. These data should be used to determine direct clinical effectiveness as defined above. The additional step of defining what is minimal in order to ration is clearly a value laden moral judgment and should be declared as such. Whether to choose a survival of at least three or five months as the relevant threshold of minimum effectiveness is ultimately arbitrary and depends on public understanding of its tolerability for affected patients. In pluralistic societies people will choose different thresholds when starting to think about whether some clinical benefits are too small to be ethically appropriate. In our view, thinking about which clinical benefits are ethically appropriate in national healthcare systems already amounts to an important step in the fair allocation of scarce healthcare resources. A representative survey of several rationing criteria recently conducted in Germany indicates a remarkable public tolerance of the suggestion to ration publicly funded healthcare by thresholds of minimum effectiveness 13—a stark contrast to the widespread opposition to alternative criteria such as age or cost. A majority of those surveyed thought that the threshold for life extension should be six months or longer (figure)—a higher threshold than we have suggested here.

Figure1

Response to survey question: “At least how many months should a therapy extend patients’ lives on average to be included in the publicly funded healthcare package?” (n=783). Data from large multicentre study on priority setting in German statutory health insurance (FOR655)14

We suggest entering into an open and reasoned debate about the magnitude of clinical benefits we expect our healthcare systems to deliver and collecting data on what the public deems to be ethically appropriate in this respect. Such data should then inform the process of implementing thresholds of clinical effectiveness. For pragmatic reasons, this process could start with medical innovations, effectiveness data for which are required for licensing, but should ideally soon include standard interventions. Medical organisations such as the royal colleges could be asked to draw up lists of interventions suspected of low clinical effectiveness, and these would be assessed by bodies such as NICE, Germany’s Institute for Quality and Efficiency in Health Care, or the US Institute for Clinical and Economic Review.

Challenges

Even if rationing means the loss of only small benefits some doctors might insist on offering any therapeutic effect to patients or stress the role of small effects in medical progress; some patients will always clutch at straws, no matter how flimsy; some citizens might insist on the provision of such straws, at least as long as they are cheap. However, we believe that medical progress should be furthered by testing any hopeful improvement in clinical studies of high power and by expecting more than minimal advances. Minimally effective straws that are cheap can be paid for out of patients’ pockets or by private insurance. There are already established examples of this: the effectiveness of homoeopathic treatments has remained dubious, and most healthcare systems have excluded them from public funding despite the fact that they are cheap, and in the face of continued patient demand (in Germany, spending on homoeopathic treatments was over €258m in 2009, most of this out of pocket).15

Another obvious challenge lies in the practicalities of determining the many different thresholds necessary for judging different medical interventions. This is a formidable problem, but it exists for most health technology assessment.

Finally, there remains the major challenge of proving the financial effect of our suggestion—the savings have to be large enough to make the implementation of minimum effectiveness thresholds worth while. Systematic data on the prevalence of potentially minimally effective interventions are not yet available, but examples discussed by us and others suggest a substantial effect.16

Conclusions

Setting minimum effectiveness thresholds for rationing healthcare has several important advantages over other approaches and should be introduced before any other rationing criteria are considered. Thresholds of minimum effectiveness lend themselves particularly well to evaluating new treatments considered for inclusion in national healthcare systems such as the NHS, US Medicare, or German statutory insurance. However, we believe that, in principle, minimum effectiveness standards could be determined for almost all medical treatments (possibly excluding mental healthcare and palliative care), as well as for public health interventions.

Since effectiveness analysis can be based on available data from research, rationing could be done in a transparent, systematic, and publicly accountable way. If carefully explained, minimum effectiveness thresholds might also be one of the few forms of rationing acceptable to medical professionals as well as to the wider public.

Minimum effectiveness thresholds are of course no silver bullet, and reaching consensus on concrete thresholds will be difficult. Meeting these important challenges could best be considered as a common task for everyone affected. It is high time indeed.

Notes

Cite this as: BMJ 2011;342:d54

Footnotes

  • We thank Jo Wolff, David Hunter, Harald Schmidt, Bob Truog, Dan Brock, Dan Wikler, Nir Eyal, Josef Lentsch, Johann Ach, and the BMJ editors for helpful comments. This article represents the authors’ views, not those of Münster University or the Nuffield Council on Bioethics.

  • Contributors and sources: This article is a result of an ethics project, led by BS-S, within an ongoing multicentre research programme on priority setting and rationing in the German statutory health insurance funded by the German Research Foundation (DFG FOR 655). BS-S is a former member of the German Ethics Council and has published widely on fair rationing and other bioethical topics. AMB is a former scholar in the Harvard Program in Ethics and Health and has published on rationing and other topics in bioethics. AMB prepared the manuscript in collaboration with BS-S and DRF. AMB is guarantor.

  • Competing interest: All authors have completed the unified competing interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare financial support for the submitted work from the German Research Foundation; no financial relationships with commercial entities that might have an interest in the submitted work; and no non-financial interests that may be relevant to the submitted work.

  • Provenance and peer review: Not commissioned; externally peer reviewed.

References