

S90
A B S T R A C T S
decline, which is a major threat for older persons in general
with increased risk for hospital and nursing home admission
and mortality. Functional decline is becoming an important
health care focus for older patients in general during cancer
treatment and more specific for older patients receiving
chemotherapy.
Objectives:
This study aims to evaluate the evolution
of functional status (FS) 2 to 3 months after initiation of
chemotherapy, to identify predictors associated with func-
tional decline during chemotherapy treatment and to
investigate the prognostic value of decline for overall survival
(OS).
Methods:
Patients
70years with a malignant tumor were
included when chemotherapy was initiated. All patients
underwent a geriatric assessment (GA) including FS measured
by Activities of Daily Living (ADL) and Instrumental Activities
of Daily Living (IADL). FS of patients was followed by repeating
ADL and IADL to define functional decline.
Results:
From 10/2009 till 07/2011, 439 patients were
included. At follow-up, ADL and IADL data were available
for 387 patients. Functional decline for ADL and IADL was
observed in 19.9% and 41.3% of the patients respectively. In
multivariable logistic regression analysis, baseline predictors
for ADL decline are abnormal nutritional status (OR: 2.02)
and IADL dependency (OR: 1.76). Time-point of assessment
(disease progression/relapse vs new diagnosis) (OR: 0.59) is
the only determinant of decline in IADL. Functional decline
in ADL is strongly prognostic for OS (logrank p-value <0.0001;
Wilcoxon p-value <0.0001) with HR 2.34 and functional decline
in IADL is also prognostic for OS but less prominent with HR
1.25.
Conclusion:
Functional decline occurs in about a third of
older patients with cancer receiving chemotherapy and can
be predicted by GA components. It strongly predicts survival,
the most prominent for ADL. These predictors can be used to
identify older persons with cancer receiving chemotherapy
eligible for interventions to prevent functional decline.
Disclosure of interest:
None declared
Keywords:
Cancer, functional decline, geriatric assessment,
older person, overall survival
P092
SCREENING FOR MULTIDIMENSIONAL HEALTH PROBLEMS
IN OLDER PATIENTS WITH CANCER: EFFECT OF VARYING
GOLD STANDARD DEFINITIONS ON THE DIAGNOSTIC
PERFORMANCE OF THE G8 AND MODIFIED G8 SCREENING
TOOLS
C. M. Martinez-Tapia
1,
*, M. Laurent
1,2
, E. Paillaud
1,2
,
J.-L. Lagrange
3
, M. Allain
4
, A. Chahwakilian
5
,
P. Boudou-Rouquette
6
, S. Bastuji-Garin
1,4,7
, E. Audureau
1,7
and
the ELCAPA Study Group
1
EA 7376 CEpiA (Clinical Epidemiology and Ageing Unit),
Université Paris Est,
2
Internal Medicine and Geriatric Department,
3
Department of Medical Oncology,
4
Clinical Research Unit, Henri-
Mondor Hospital, Assistance Publique-Hôpitaux de Paris, Créteil,
5
Oncogeriatrics, Geriatric Department, Broca Hospital, Hôpitaux
Universitaires Paris Centre, Assistance Publique-Hôpitaux de
Paris,
6
Department of Medical Oncology, Cochin hospital, Hôpitaux
Universitaires Paris Centre, Assistance Publique-Hôpitaux de
Paris, Paris,
7
Public Health Department, Henri-Mondor Hospital,
Assistance Publique-Hôpitaux de Paris, Créteil, France
Introduction:
Multidimensional geriatric assessment (GA)
is recommended by the SIOG for detecting health problems in
older patients with cancer and tailoring treatment decisions
accordingly, but is time- and resource-consuming. Screening
tools have been developed to help identifying patients
warranting a complete GA, but there is no unequivocal
definition of what constitutes this population and what the
reference gold standard should be. A pragmatic definition
based on
1 abnormal test at the GA has been widely used,
but this approach is hampered by a lack of standardization in
GA components across studies. This definition also does not
capture important aspects of the reality of clinical practice in
geriatric oncology, such as actual treatment decisions based
on GA findings, expert-based clinical classifications and/or
broader approaches to frailty. No previous study has examined
the variability of diagnostic performance of screening tools
under multiple clinically relevant reference definitions.
Objectives:
To measure and compare the effect of
using varying gold standard definitions on the diagnostic
performance of two screening tools specifically developed for
older patients with cancer, the G8 and modified G8
1
.
Methods:
We used a prospective cohort of
70 year-old
patients with cancer referred to geriatricians for GA (ELCAPA
cohort). Areas under the Receiver Operating Characteristic
(AUROC) curves were calculated to compare the diagnostic
performance of both tools against the following reference
standards: a) detection of
1 or b) 2 impaired components of
the GA; c) prescription of
1 clinically significant intervention
by the geriatrician; d) identification of a vulnerable profile
as defined by a latent class approach or e) by expert-based
classifications from Balducci and f) Droz.
Results:
1136 patients were included for the present
analysis (median age, 80 years; 52% men; 44% metastastic
cancer; 48% ECOG-PS 0-1). AUROC were equal or higher than
0.80 for both tools and all definitions tested. Comparing the
two instruments, AUROC were significantly higher in favor of
the modified G8 to predict 4 out of the 6 definitions tested:
GA
1 impairment (modified G8: 0.93 [95% CI 0.91–0.95] vs.
original G8: 0.90 [0.87–0.92]; p=0.0029), GA
2 impairments
(0.90 [0.88–0.92] vs. 0.87 [0.88–0.92]; p=0.0006),
1 significant
intervention prescribed (0.85 [0.81–0.89] vs. 0.81 [0.77–0.86];
p=0.0056) and unfit patient according to Droz’s classification
(0.88 [0.86–0.91] vs. 0.83 [0.81–0.86]; p<0.0001). No significant
difference was found for latent class typology and Balducci’s
classification. Sensitivities based on optimal cutoffs were of
similar magnitude for both tools, whereas most specificities
were higher for the modified G8.
Conclusion:
Our findings demonstrate the robustness
of the original and modified G8 to modifications of the
reference gold standard, with evidence of a better diagnostic
performance of the modified G8 for detecting a variety of
health profiles evocative of vulnerability.These results further
support the clinical value of these instruments for detecting