A cancer survival scale based on readily available clinical and laboratory variables reliably predicted whether patients in palliative care had days, weeks, or months to live, British investigators found.
Models designed to predict survival of less than 14 days, 14 to 55 days, or 56 days or longer showed good correlation with actual survival for the three prognostic categories, reflected in an area under the curve range of 0.79 to 0.86, researchers reported online in BMJ.
The models performed as well as or better than physician and nurse estimates of survival.
"The [Prognosis in Palliative care Study] PiPS-A score [based only on clinical data] can be calculated for any patient with advanced cancer who is no longer receiving disease-modifying treatment, and it is at least as good as, but not significantly better than, a clinician's estimate of survival," Patrick C. Stone, MD, of St. George's University of London, and coauthors wrote in conclusion.
Prognostic information is important to terminally ill patients and their physicians. However, clinician predictions of survival tend to be inaccurate and overly optimistic, the authors noted in their introduction.
Several studies have identified clinical and laboratory variables that predict survival in advanced cancer, and prognostic scoring systems based on the variables have been developed. A limitation of all the instruments, however, is that none has been "benchmarked" against clinicians' survival estimates, the authors continued.
So they sought to develop a prognostic scale that "did not rely on clinicians' estimates of survival but was at least as accurate as their best predictions."
To develop the prognostic system, the investigators recruited participants from 18 palliative care services in England. Data were collected on all participating patients from March 2006 to August 2009, and all patients were followed for at least thee months after enrollment in the study.
The investigators obtained clinical data from patients' medical records, and participants' mental status was assessed by a 10-item scale. For each patient, physicians and nurses involved in the care were asked to predict the patient's survival, using one of four categories: days (<13), weeks (two to less than eight), months (two to less than 12), and years (12 months or more).
Each patient's illness severity was determined using the Charlson Comorbidity Index, and the researchers asked permission to obtain a blood specimen from all patients judged competent to make decisions.
The final analysis included 1,018 patients, three-fourths of whom were competent to make their own decisions. The study group had a median survival of 34 days.
Multivariate analysis revealed 11 core variables that independently predicted two-week and two-month survival.
Four variables (dyspnea, dysphagia, bone metastases, and alanine transaminase) predicted only two-week survival. Eight variables predicted only two-month survival (primary breast cancer, male genitourinary cancer, fatigue, weight loss, lymphocyte count, neutrophil count, alkaline phosphatase, and albumin).
Prognostic models were developed for patients who did not have laboratory data (PiPS-A) and those who did (PiPS-B). Within each model, investigators developed separate models to predict survival of 14 days or more (PiPS-A14, PiPS-B14) and survival of two months (56 days) or longer (PiPS-A56, PiPS-B56)
Within each model, patients' estimated survival was designated as days, weeks, or months/years. The median survival across the three categories was 5, 33, and 92 days for the PiPS-A models and 7, 32, and 100.5 days for the PiPS-B models.
The investigators combined the 14-day and 56-day models within PiPS A and B to address the question of whether a patient would survive for more than two weeks but less than two months.
The PiPS-A models agreed with actual patient survival 59.6% of the time, compared with 56.3% for physician estimates, 55.2% for nurses', and 57.5% for the two professions combined. The results showed that PiPS-A models performed as well as clinician estimates.
Combining the PiPS-B models resulted in agreement with actual survival 61.5% of the time, which proved to be significantly better than estimates of physicians (52.6%, P=0.0135) or nurses (52.3%, P=0.012). The combined-professional estimate was in agreement with actual survival 53.7% of the time, which did not differ significantly from PIPS-B.
Limitations of the study included selective recruiting, since not all evaluable patients were studied, and inability to assess the tool in an independent cohort.
Prognosis "needs to be restored as a core clinical skill, to optimize the patient's treatment and planning." Paul Glare, MD, of Memorial Sloan-Kettering Cancer Center in New York, wrote in an accompanying editorial.
"A new science of prognosis is emerging in palliative care," Glare wrote. "There are two components to the skill of prognosis -- formulating the prediction and communicating it to the patient."
Although helpful, however, prognostic tools should not be applied arbitrarily, Glare continued. As an example, a predicted survival of days might be accurate for natural death at home but would probably be inaccurate for a patient admitted to a hospital for aggressive life-sustaining treatment in an ICU.
Moreover, even the most evidence-based prognostic tool will often be inaccurate in estimating a patient's survival, Glare wrote. New and more reliable prognostic factors are needed. Until prognostic strategies' inherent inaccuracy can be corrected, most physicians will resist prognostication, and those who do not will be accused of "playing God."
"Patients' preferences for prognostic information vary during the course of the illness, so communicating the prediction to the patient is as important as forecasting it," Glare concluded.
No comments:
Post a Comment