The 'missing data'
Gwilliam et al are to be congratulated for the excellent study in
attempting to devise a prediction tool for use in palliative care. (1).
There are several factors that would affect the validity and clinical
utility of the proposed prognostic score.
The high refusal rate of 57% is understandable but it is not known
whether the patients who refused had a different survival rate.
The PiPS-A predictions being correct on 59.6% of occasions (and multi
-professional predictions being correct on 57.5%) is very close to 50%,
which as a non-statistician, I would regard as the predictive power of a
toss of a coin!.
Eligible patients for the study were no longer receiving active
treatment for cancer, and no further disease modifying treatment was
planned. About 33% of patients in the study had no distant metastasis and
apparently 12% (100 minus 67.2+ 20.5) patients had localised disease with
no nodal spread (table 1). The reasons why this subset of patients did not
have active treatment would be helpful. Were the long-term survivors in
the study are from this group?
It would also be helpful to know the proportion of study patients
treated in teaching hospitals and whether the model retained the
'predictive power' in patients who had multiple prior treatments including
It would also interesting to see whether the proposed models of
multiple variables performed better than a simple model using selected
variables such as performance status, serum albumin and resting pulse
There are various other factors which significantly affect the long
term (weeks to months) prognosis of patients such as median age of the
study sample, co-morbidity, stage at diagnosis, time from diagnosis to
referral to palliative care, disease free interval prior to development of
metastasis and number of prior cancer treatments. Were these collected as
part of the study?
Finally, One can argue that this tool, if validated, would be very
useful in research setting. But, on the other hand, some patients who are
not receiving any active treatment and whose prognosis might be predicted
by this tool to be in days might find the information disconcerting.
Patients quite often remember specific survival data and tend to forget
caveats. We also quite often see patients in the clinic who pride
themselves in beating the odds. Hence the clinical utility of such a tool
needs to be investigated. Were the patients in the study asked for their
views and wishes regarding a prediction tool?
1. Gwilliam B, Keeley V, Todd C, Gittins M, Roberts C, Kelly L, et
al. Development of Prognosis in Palliative care Study (PiPS) predictor
models to improve prognostication in advanced cancer: prospective cohort
2. Chow E, Harth T, Hruby G, Finkelstein J, Wu J, Danjoux C.How
accurate are physicians' clinical predictions of survival and the
available prognostic tools in estimating survival times in terminally ill
cancer patients? A systematic review. Clin Oncol (R Coll Radiol).
3. Jouven X, Escolano S, Celermajer D, Empana JP, Bingham A, Hermine
O, Desnos M, Perier MC, Marijon E, Ducimeti?re P.Heart rate and risk of
cancer death in healthy men. PLoS One. 2011;6(8):e21310. Epub 2011 Aug 3.
4. Gupta D, Lis CG. Pretreatment serum albumin as a predictor of
cancer survival: a systematic review of the epidemiological literature.
Nutr J. 2010 Dec 22;9:69.
5. Goldwasser P, Feldman J. Association of serum albumin and
mortality risk.J Clin Epidemiol. 1997 Jun;50(6):693-703.
Competing interests: No competing interests