BMJ: Predictive model forecasts survival in those with advanced cancer
A prognostic model designed for use in patients with advanced cancer bested clinicians’ estimates of survival and performed in an array of patients and circumstances, according to a study published online Aug. 25 in British Medical Journal.

Patients with advanced cancers, along with their caregivers and clinicians, often request an estimate of expected survival time. “Realistic survival estimates can inform decisions about the appropriateness of medical interventions and the timing of referral to specialist palliative care services or admission to a hospice,” wrote Bridget Gwilliam, PhD, clinical research fellow, division of population, health sciences and education at St. George’s University of London, and colleagues.

Current methods for predicting patient survival rely on physician estimates, which tend to be inaccurate and overly optimistic, she suggested. Thus, the researchers sought to develop a prognostic model based on previously identified clinical and laboratory variables that predict survival in patients with advanced cancer and benchmark the tool against clinicians’ estimates.

The authors devised a multicenter trial, recruiting 1,018 patients with locally advanced or metastatic incurable cancer from 18 palliative care services across England between March 2006 and August 2009.

Gwilliam et al collected multiple variables with good a priori evidence of prognostic utility, including a symptom checklist of pain, breathlessness at rest, loss of appetite, difficulty swallowing and dry mouth. They recorded patients’ Eastern Cooperative Oncology Group performance status, clinical observations and used an abbreviated mental test score to assess cognitive status.

They requested clinicians and nurses to estimate survival based on days (more than 14 days), weeks (from two weeks to less than eight weeks), months (from two months to less than 12 months) and years (12 months or more). Finally, they recorded demographic, disease-related and treatment-related variables, co-morbidities and laboratory variables (when available).

Next, they constructed four models to predict survival. They analyzed outcomes at two weeks and two months for patients with the full dataset [Prognosis in Palliative care Study (PiPS-A) model] and the dataset for patients referred to patients with laboratory results (PiPS-B).

Gwilliam and colleagues reported that the area under the curve exceeded 0.79 for all four models. The PiPS-A models performed at least as well as clinicians’ predictions of survival in terms of days (less than 14), weeks (14 to 56 days) and months/years (more than 56 days). PiPS-A was correct 59.6 percent of the time, and clinicians were accurate 57.5 percent.

PiPS-B models, which include blood results, also were significantly better than physicians (61.5 percent vs. 52.6 percent) and nurses (61.5 percent vs. 52.3 percent). However, they were not significantly better than a multi-professional estimate in which a physician and nurse independently produced the same survival estimate. This multi-professorial estimate was correct 53.7 percent of the time vs. 61.5 percent for the model.

Although the differences in prognostic accuracy are not large, the models provide a clinician-independent score that can be reproduced and compared across settings, the researchers said.

“These models are able to identify reliably those patients with expected prognoses of days, weeks or months/years and can be used in either competent or incompetent patients and in circumstances when blood results are available and when additional investigations would be inappropriate,” wrote Gwilliam.

“In a clinical context, we believe that PiPS estimates would usually be used to inform and augment clinicians’ own subjective estimates (rather than to replace them),” they continued.

The researchers emphasized the generalizability of the model, but acknowledged that limited access to evaluable patients because of gatekeeping by clinical staff could have distorted the findings.

In addition, the models should be externally validated in a separate cohort to confirm their accuracy before integration into clinical practice. Finally, the researchers called for future studies to evaluate the clinical utility of the approach.