Planning palliative or terminal care: the dilemma of doctors’ prognoses in terminally ill cancer patients

S. Tanneberger*, I. Malavasi, P. Mariano, F. Pannuti and E. Strocchi

Associazione Nazionale Tumori (ANT), Istituto di Ricerca, Via Curiel 7, 40134 Bologna, Italy (E-mail: tanneberger@antnet.it)

In the April 2001 issue of Annals of Oncology, Marco Maltoni and Dino Amadori reviewed the situation in palliative medicine and medical oncology and presented statements on palliative and terminal care of the British National Council for Hospice and Palliative Care Services [1]. Terminal care was defined as "the management of patients during the last few days or weeks or even months of life." This vague definition makes visible the difficulties we have in estimating the duration of life for patients with end-of-life cancer. These difficulties can become an ethical dilemma when a physician, on referring a patient to a hospice programme, is obliged to predict accurately the patient’s prognosis. Moreover, these difficulties can create uncertainty that affect day-to-day patient management in palliative care programmes including physician concern regarding the use of high-dose morphine, enteral or nutritional support and/or providing parenteral fluids during the terminal phase of illness. Certainly guidelines are useful, such as the recently published French ‘standards, options and recommendations’ for nutritional support in palliative or terminal care of adult patients with progressive cancer. However, the recommendation to renounce artificial nutrition when life expectancy is below 3 months remains difficult to realize, since assessment of life expectancy can be erroneous [2].

Available data concerning the accuracy of the clinical estimation of a patient’s life expectancy and the associated ethical dilemmas are scarce. Christakis et al. [3] reported on 468 patients seen by 343 randomly selected doctors (that is, less than two patients per doctor). The doctors were not trained in end-of-life care: fifty-two per cent were specialized in internal medicine and only a minority in oncology (17%) or geriatrics (8%). Only 20% of predictions were accurate, 63% were overly optimistic and 17% overly pessimistic. Even Vigan et al. [4] noted a tendency to overestimate survival. Twenty-five per cent of 225 patients in this study had their survival correctly predicted by the treating oncologist. Twenty-three per cent had their survival underestimated and 52% had their life expectancy overestimated. In this study the qualifications of the ‘treating oncologists’ were not recorded. Similar estimations, consistent with the results given above, were found in 200 patients assessed in parallel by oncologists (25.7%), GPs (21.7%) and nurses (21.5%) in a Spanish study [5].

As an alternative, various prognostic indices, scales and score models were proposed. These are certainly applicable for patients treated in specialized institutions. Nevertheless, for the majority of patients, subjective prognostic predictions of the care-giving staff will remain daily practice. In order to evaluate whether errors in the prognoses of terminally ill patients decrease as the qualifications of the treating doctor increase, we initiated in 1999 a study on the accuracy of clinical estimation of duration of life for patients with end-of-life cancer. In contrast to the studies mentioned above, the participating doctors had a high degree of experience in end-of-life care.

Altogether eight doctors participated in the study completing a questionnaire at the first consultation. In the questionnaire they had to classify patients into one of five survival categories (0–2, 2–4, 4–9, 9–17 or >17 weeks) and to indicate one or more of 13 easily assessable criteria for decision making. Doctors had a median duration of medical practice of 14 years and were all board certified. Half were specialized in oncology and all were working full-time in home care for end-of-life cancer patients for a medium duration of 9.5 years. Two hundred and sixty-nine patients (median age 74 years; mean age 72 years, range 34–96 years) were included in the study. Median survival of patients was 9.6 weeks [95% confidence interval (CI) 8–10.7 weeks]. Main groups by tumor site were gastrointestinal (90 patients), urogenital (61 patients), lung (50 patients) and breast (22 patients). Table 1 shows that in 33% of patients the prediction was correct. Thirty-nine per cent survived less than estimated and 28% survived longer. The prognostic criteria, mainly used by the treating doctor, were localization of tumor (59%), localization of the metastasis (58%), anorexia (47%), tumor dimension (42%) and severe weakness (45%). Other criteria, such as fever (3%), stomatitis (7%) or disorientation (9%), were used much less or only site specifically (shortness of breath 26% in lung pathology).


View this table:
[in this window]
[in a new window]
 
Table 1.  Overall comparison of clinical estimation of survival and actual survival in 269 patients with advanced cancer
 
Data show that long-term experience can improve the accuracy of the clinical estimation of life expectancy. However, data also underline the difficulties of prognosis in advanced cancer even for doctors well trained in end-of-life care. Accuracy is cancer site related. The groups of patients by site in our study are relatively low, but it seems justified to conclude that correct predictions for breast (46%) and gastrointestinal cancer (37%) are more frequent than for urogenital (27%) and lung (18%) cancer. This could be associated with the risk of non-predictable, life-terminating infections (more frequent in lung and urogenitial cancer). Remarkably, tumor-related criteria were used much more frequently for decision making than actual symptoms, such as disorientation (9%), stomatitis (6%) or fever (3%). Obviously, doctors were rating higher their clinical experience than the evaluation of actual symptoms.

We conclude the following from this study:

Clinical estimation of duration of life for patients with end-of-life cancer needs experience and training. Educational programmes in palliative care should place more emphasis on this than has previously been the case.
Prognosis should be based more on proven indices and less on intuition. However, there is no doubt that daily clinical practice limits the use of highly sophisticated computer-based score models.
Even maximal accuracy of prognosis will not exclude the risk of errors in the large majority of patients. This limits the classification of patients into care categories that are too strictly defined.
Health care systems should avoid models for care with standards and budgets based on prognostic estimates and the medical community should avoid claim by disciplines of certain categories of patients defined by their prognoses.
What is needed is a network of assistance for incurable cancer patients with single parts defined by patients’ needs and not by their predicted life expectancy. Separating palliative and terminal care is artificial and often in conflict with the needs of the patient.

S. Tanneberger, I. Malavasi, P. Mariano, F. Pannuti & E. Strocchi

Associazione Nazionale Tumori (ANT), Istituto di Ricerca, Via Curiel 7, 40134 Bologna, Italy (E-mail: tanneberger@antnet.it)

References

1. Maltoni M, Amadori D. Palliative medicine and medical oncology. Ann Oncol 2001; 12: 443–451.[Abstract]

2. Bachmann P, Marti-Massoud C, Blanc-Vincen MP et al. Standards, options and recommendations: nutritional support in palliative or terminal care of adult patients with progressive cancer. Bull Cancer 2001; 88: 985–1006.[ISI][Medline]

3. Christakis NA, Lamont EB. Extent and determinants of error in doctor’s prognoses in terminally ill patients: prospective cohort study. Br Med J 2000; 320: 469–472.[Abstract/Free Full Text]

4. Vigan A, Dorgan M, Bruera E et al. The relative accuracy of the clinical estimation of the duration of life for patients with end of life cancer. Cancer 1999; 86: 170–176.[ISI][Medline]

5. Llobera J, Esteva M, Rifa J et al. Terminal cancer. Duration and prediction of survival time. Eur J Cancer 2000; 36: 2036–2043.[ISI][Medline]





This Article
Full Text (PDF)
E-letters: Submit a response
Alert me when this article is cited
Alert me when E-letters are posted
Alert me if a correction is posted
Services
Email this article to a friend
Similar articles in this journal
Similar articles in ISI Web of Science
Similar articles in PubMed
Alert me to new issues of the journal
Add to My Personal Archive
Download to citation manager
Search for citing articles in:
ISI Web of Science (2)
Disclaimer
Request Permissions
Google Scholar
Articles by Tanneberger, S.
Articles by Strocchi, E.
PubMed
PubMed Citation
Articles by Tanneberger, S.
Articles by Strocchi, E.