SMJ 2003: 48(1) 17-20
David
Austin
Elizabeth
M Russell, Emeritus Professor of Social Medicine
Address for correspondence and
reprints:
Professor Elizabeth Russell
Department of Public Health
Phone and fax: 01224 861216
Email: e.m.russell@abdn.ac.uk
Abstract
National
statistics indicate that five-year survival from all cancers is worse in older
patients even when reduced life expectancy is accounted for1.
Turner et al.2
summarised disparities between age groups in investigation, treatment and
outcome in cancer. She reviewed the evidence for and against attributing them to
age-based decisions not to treat, but was unable to reach a conclusion on the
underlying reasons.
The
principles of the NHS oppose age-based rationing, and any explicit priorities
have favoured older patients either directly or indirectly because they are the
main sufferers of cancer and heart disease, which are two current priorities3-5.
Traditionally, doctors have been trained that clinical need is the only
stated criterion on which care should be determined.
Age-based decisions to limit
treatment that are not justifiable on grounds of clinical need can therefore be
regarded as ageist. Age Concern feels that ageism exists within the NHS6, yet
there is little empirical data to back-up this claim.
Using
routine data, this study took a different approach to Turner and aimed to
determine whether age related differences in hospital management of cancer
patients were clinically justifiable or not. Colorectal cancer was chosen as the
tracer condition so that the justification of clinical decisions could be
related to available and accepted guidance on good clinical practice7 .
A recent systematic review of surgery for colorectal cancer found
that, although overall survival after colorectal cancer surgery decreased with
age, there was no significant difference in cancer-related mortality with age in
those who underwent curative surgery8. Neither was there a greater
level of co-morbidity (as judged by the presence of conditions such as diabetes
and hypertension) with increasing age in the 84% who had surgery, although
probably there was more co-morbidity in the 16% who did not. However, the
proportion whose tumours had been staged fell significantly with age, which
suggests that in the older old patients something other than stage was used as
the basis for deciding whether or not to undertake curative surgery. Current
evidence is that
·
Although the systematic review
showed that stage of tumour at presentation increases with age, this trend
reflected only a deficit of Stage A and an excess of Stage D. Other studies have
shown no difference in stage at presentation. However, the number of patients
with unknown stage increases with age8-11
·
Age of patient alone is not a predictor
for poor survival or complication rates after surgery11-15
·
Older patients respond as well
to chemotherapy as do younger patients, although they are under represented in
clinical trials and data on toxicity and quality of life is poor 16,17
Thus,
on the basis of this evidence, there is no a
priori reason for managing older patients differently solely on account of
their age. The issue for best practice is the correct selection of patients
according to their clinical condition rather than the simplistic approach of
assuming that older patients will always fare worse and therefore may not merit
aggressive management.
The
studies in the systematic review were, by definition, prospective audit or
research, usually in relatively controlled conditions. In Scotland, because of
the record linkage system for all hospital discharges that has existed for at
least 20 years, there is a rare opportunity to analyse routine hospital patient
care longitudinally regardless of where or by whom it was given. This study
therefore examined the care of colorectal cancer patients in all Scottish
hospitals to see whether any differences by age could be justified on clinical
grounds. It was also recognised that the study would provide a baseline for
future audit of this care in the light of the increasing use of clinical
guidelines for good practice 7.
Methods
The
data set was the Scottish cancer registration records for all patients aged over
17, diagnosed between 1st January 1992 and 31st December 1996 (n=15299).
Computerised probability linking was carried out by ISD Scotland to
connect patients to their individual episodes of care collected through the
Scottish Morbidity Record inpatient and day case form (SMR 01).
SMR 01 details were selected if they contained a diagnosis of colorectal
cancer and were within the time frame of one year pre-registration to one-year
post registration.
Excluded
were patients with missing deprivation category (n=123), missing link numbers
(n=182), anal cancers (n=235) and patients who were not admitted to hospital
(including death certificate only registrations, n=1983).
The revised data set contained 12776 patients.
The
results were adjusted for co-morbid conditions that may affect survival and
therefore treatment choice. The method of adjustment was that used in the
Scottish Clinical Outcome Indicators for all cancers18.
Hospital admissions with a principal diagnosis of any of seven conditions
deemed to be relevant to cancer prognosis were used as the basis for this
variable. The conditions were diabetes, hypertension, ischaemic heart disease,
other heart disease, cerebro-vascular disease, respiratory disease and
arthritis, and the period of two years before to one month after cancer
registration was defined as a clinically relevant time frame.
The co-morbidity variable had three levels: "no admissions",
"one admission" or "two admissions" for any of the relevant
conditions.
Tumour
sub-site was also identified as a clinically relevant factor that is known to
vary with age. . Sub-site was coded
into the categories “right” (caecum to transverse colon), “left” (splenic
flexure to sigmoid colon), “rectal” (rectosigmoid junction, rectal ampulla)
and “unspecified”.
Emergency
admission is known to be associated with a poorer outcome in colorectal cancer9,19,20,
although the reasons are unclear. Route of first inpatient admission for
colorectal cancer was therefore used as one indicator of poor prognosis. Sex and
deprivation (measured by the Carstairs index of deprivation, DEPCAT21)
were included to allow adjustment for socio-demographic factors that might
affect the interpretation. Attempts
were also made to adjust for poor overall fitness. Patients whose registration
was based only on death certification and those with no inpatient record of care
were removed as being many of the "sickest" patients.
This
study used three main indicators of clinical outcome. Histological verification (from the cancer registration data
set) to indicate extent of investigation; the "gold standard" of
investigation for colorectal cancer is a pathological diagnosis. Definitive surgery (coded for on SMR 01) to indicate
potentially curative procedures; all patients with a resectable tumour should be
considered for surgery7. Chemotherapy
(coded for on SMR 01 as procedure X35.2) to indicate adjuvant and palliative
treatment; chemotherapy should be considered for all patients with Dukes stage C
or with advanced disease 7.
SPSS
for Windows was used for all analyses. Age
was initially cross-tabulated against each main outcome to give an indication of
the unadjusted pattern of care. Five
age-bands (17-54,55-64, 65-74, 75-84 and 85+) were used for all the analyses.
All relevant variables (age, comorbidity, sub-site, type of admission,
sex and deprivation category) were then entered into a logistic regression (LR)
analysis for each of the outcome variables.
Only differences significant at the 5% level are reported.
The
mean age of patient was 70.9 (SD ±12.11),
range 17-107. Right-sided and
unspecified tumours were the more common tumours in older groups, and both
co-morbidity and emergency admission increased with increasing age.
Deprivation levels were similarly distributed at each age-band and males
were affected proportionately more than females at younger ages.
The
percentage of patients whose tumours were histologically verified decreased with
age (Mann-Whitney p<0.001,Table
1) The proportion of verified patients (mean 90.6%, n=11581) in the youngest
group was 95.5%, compared with 88.1% in the 75-84 category and 78.1% in the 85+
category. When adjusted for all other relevant factors, increased age remained a
negative predictor for histological verification: 75-84 year olds were 2.7 times
less likely (LR p<0.0001) and 85+ year olds 4.8 times less likely (LR
p<0.0001) to be histologically verified when compared to the youngest group (Figure
1).
The
percentage of patients receiving definitive (ie potentially curative) surgery
(mean 78.1%, n=9977) decreased with age (Mann-Whitney p<0.001, Table
1). After adjusting for the
relative influence of other relevant factors, age continued to be a
statistically significant negative predictor.
Compared to the youngest age band, 75-84 year olds were 20% less likely
to receive definitive surgery (LR p=0.003) and patients of 85 and over were 55%
less likely to receive definitive surgery (LR p<0.0001, Figure
2).
Approximately
one-third (32.8%) of 17-54 year olds and one-fifth (18.5%) of 55-64 year olds
received chemotherapy (Table 1).
Very few patients over 75 received chemotherapy: at 75-84 - only 47
patients (1.4%) - and at 85 and over only 1 patient in Scotland (0.1%) received
the treatment (Mann-Whitney p<0.001). The overall mean was 10.6% (n=1351).
When adjusted for all other relevant factors, increasing age remained a
statistically significant negative predictor for receiving chemotherapy (Figure
3).
Discussion
Our
results show striking age associated differences in the investigation and
treatment of colorectal cancer. Even when adjusted for clinical and demographic
factors, older patients showed decreased rates of histological verification,
potentially curative surgery, and chemotherapy. The rate of potentially curative
surgery was similar to that found in the systematic review8. However,
the percentages of cancers in the Scottish routine data that were not
histologically verified were higher at all ages than the percentages of unstaged
cancers in the review.
What
makes this study valuable is the extent of coverage that the Scottish routine
data allow and the unique insight into the pattern of care within the health
service. While criticism can be
levelled at large scale retrospective studies and a cautious approach to
interpretation should be maintained 22, the results from the study
merit debate. The unique methods of linkage of ISD data were set up in the 1990s
specifically for the purpose of mapping trends and analysing comparative
information. Scotland's cancer
registration system compares favourably with those in other countries. Case ascertainment is believed to be about 95%, data linkage
is highly accurate (error rate <1%), and coding for procedure is also thought
to be correct in over 96% of cases18, 23. Future use of linked SMR 06 data in Scotland has since been
further improved by the addition of variables indicating stage at presentation
and grade of tumour
This
addition permitted a post hoc analysis of cancer registration (SMR 06) data for
1997 to examine the percentage distribution of stage by age at presentation.
When the patients not investigated were excluded, stage did not vary with age
(Figure4). Compared to the recent
systematic review data 8, unknown stage was more likely at all ages
but especially in the 75-84 age group, in whom 15.6% in Scotland in 1992-96 were
not staged (Table 2) compared
with only 9% in the review (covering studies from 1988-2000).
A
further comparison between the 1992-96 data and the systematic review showed
that, while the overall rate of potentially curative surgery was similar (78% in
Scotland compared to 75% in the review), in Scotland this fell from 81% in the
65-74 age group to 61% in those aged 85 plus. In the systematic review, the
comparable fall was from 75% to 67%. If these comparisons are accurate, they may
reflect differences between special studies and routine practice that may be at
the margin of ageism rather than of best clinical practice.
Attempts
to correct for co-morbidity were hampered by the lack of a validated measure in
colorectal cancer. The co-morbidity
indicator used is based on principal diagnosis on hospital admission for
conditions other than colorectal cancer, because principal diagnosis is more
likely to be accurately and completely recorded than are concomitant diagnoses
at the time of the cancer admission, and because the Scottish record linkage
system means that all such admissions in Scotland will be captured. ISD sample
audits have found no systematic bias in completeness of such capture. However, these admissions were recorded only over the two
prior years and so missed co-morbidity
not investigated or treated in hospital during that time. Also, severity of the
co-morbid events is unknown.
Despite
these limitations, the chosen co-morbidity indicator showed variation with age,
histological verification, chemotherapy and definitive surgery and therefore
appeared to have some face validity and be a useful, if not fully comprehensive,
indicator of co-morbidity in colorectal cancer.
Thus,
linkage of cancer registration and hospital episode data provides a very much
larger data set than does either alone, and covers all Scottish hospitals
regardless of participation in special studies or audits. The ability to adjust
the data set also considerably enhances its utility even although no one has yet
devised a reliable measure of co-morbidity. This analysis confirms the existence
of age-related variations in care that may be greater than those found in
special studies. This analysis should act as a baseline for comparison with
subsequent years, during which both more accurate staging data and a growth in
explicit clinical guidelines have become available.
Acknowledgements
We are
extremely grateful to the Cancer Intelligence Unit at ISD for all their help,
and especially Diane Stockton, without whom the several analyses would not have
been possible. Gordon Prescott
provided much needed statistical help. DA is grateful to PPP Healthcare for his
student grant.
References
1.
Scottish Cancer Intelligence Unit. Trends
in cancer survival in Scotland 1971-1995.
Edinburgh: Information and Statistics Division Scotland, 2000.
2.
Turner NJ, Harward RA, Mulley GP, Selby PJ.
Cancer in old age – is it inadequately investigated and treated?
BMJ 1999; 319:309-312.
3.
Department of Health for Scotland. A
national health service. London:
HMSO, 1942:Cmd. 6404.
4.
Scottish Home and Health Department.
The health service – the way ahead (memorandum).
Edinburgh: HMSO 1976.
5.
The Scottish Office, Department of Health.
Towards a Healthier Scotland.
Edinburgh: The Stationary Office, 1999 Cm. 4269.
6.
Age Concern England. Turning your back on us – older people and the NHS.
London: Age Concern, 2000.
7.
Scottish Intercollegiate Guideline Network and Scottish Cancer Therapy
Network.
A national clinical guideline recommended for use in Scotland.
Edinburgh: SIGN 1997.
8.
Colorectal Cancer Collaborative Group. Surgery for Colorectal cancer in
elderly patients: a systematic review. Lancet
2000; 356: 968-74.
9.
Waldron RP, Donovan IA, Drumm J, Mottram SN, Tedman S.
Emergency presentation and mortality from colorectal cancer in the
elderly.
Br J Surg 1986; 73:214-6.
10.
Curless R, French JM, Williams GV, James OFW.
Colorectal carcinoma: do the elderly patients present differently?
Age Ageing 1994; 23:102-7.
11.
Avital S, Kashtan H, Hadad R, Werdin N.
Survival of colorectal carcinoma in the elderly.
A prospective study of colorectal carcinoma and a five-year follow-up.
Dis Colon Rectum 1997; 40: 523-9.
12.
Fabre JM, Rouanet P, Ele N, Fagot H, Gulillon F, Deixonne B, Balmes M,
Domergue J, Baumel H. Colorectal
carcinoma in patients aged 75 years and more: factors influencing short and long
term operative mortality. Int
Surg 1993;78: 200-3.
13.
Payne JE, Meyer H. The
influence of other diseases upon the outcome of colorectal cancer patients.
Aust N Z J Surg 1995; 65:
398-402.
14.
Fitzgerald SD, Longo WE, Gayle L, Daniel RN, Vernava AM.
Advanced colorectal neoplasia in the high-risk surgical elderly patient:
is surgical resection justified? Dis
Colon Rectum 1993;36:161-6.
15.
Violi V, Pietra N, Grattarola M, Sarli L, Choua O, Roncoroni L, Peracchia
A. Curative surgery for colorectal
carcinoma. Dis Colon Rectum 1998; 41: 291-8.
16.
Colorectal Cancer Collaborative Group. Palliative chemotherapy for
advanced colorectal cancer: systematic review and meta-analysis. BMJ 321: 531-5.
17.
Daniele B, Simmonds PD, Best LY, Ross PJ, Cunningham D. Should
chemotherapy be used as a treatment of advanced colorectal cancer (ACC) in
patients over 70 years of age? Eur J
Cancer 35(12): 1640-9.
18.
Clinical Outcomes Working Group. Clinical
outcome indicators. The
Scottish Office 1996.
19.
Irvin GL III, Horsley JS III, Caruana JA Jr.
The morbidity and mortality of emergent operations for colorectal
disease. Ann Surg 1984;199:598-603.
20.
Nilsson E, Gregersen NP, Hartvig B, Sjodahl R.
Carcinoma of the colon and rectum. Results
of treatment in 284 cases. Acta
Chirurgica Scandinavica 1984;150:177-82.
21.
Carstairs V, Morris R. Deprivation and health in Scotland. Aberdeen: Aberdeen University Press 1991.
22.
Davenport RJ, Dennis MS, Warlow C. Effects
of correcting outcome data for case mix: an example from stroke medicine.
BMJ 1996;312:1503-5.
23.
Kendrick S, Clarke J. The
Scottish record linkage system. Health Bull 1993; 51(2):72-79.