Full text of DOI:10.2165/00019053-200119090-00006

Pharmacoeconomics 2001; 19 (9): 947-953
1170-7690/01/0009-0947/$22.00/0
ORIGINAL RESEARCH ARTICLE
© Adis International Limited. All rights reserved.
The Cost of Palliative Care for
Hepatocellular Carcinoma in
Hong Kong
Anthony T.C. Chan,1 Philip Jacobs,2 Winnie Yeo,1 Maria Lai,1 Clarke B. Hazlett,3
Tony S. K. Mok,1 Thomas W. T. Leung,1 Wan Y. Lau4 and Philip J. Johnson1
1
Department of Clinical Oncology, Sir YK Pao Cancer Centre, Chinese University of Hong Kong,
Prince of Wales Hospital, Hong Kong
2
3
Department of Public Health Sciences, University of Alberta, Canada
Office of Educational Services, Chinese University of Hong Kong, Prince of Wales Hospital,
Hong Kong
4 Department of Surgery, Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong
Background: Hepatocellular carcinoma (HCC)is endemic in parts of Asia and
Africa and most patients are not suitable for treatment with a curative approach.
Little is known about the cost of palliative care for HCC.
Abstract
Objective: To determine: (i) patient-specific costs of palliative care of HCC; and
(ii) individual factors that drive patient-specific costs and to develop a model of
cost per case under alternative circumstances.
Methods: 204 patients with inoperable HCC were prospectively tracked from
first hospitalisation until death for health service utilisation. A societal perspec-
tive of cost was taken, including costs of formal and informal services incurred
by payers, caregivers and patients. Observational data from a large Hong Kong
cancer care programme were used. A regression analysis was performed using
formal costs only, with the cost per observed day as the dependent variable.
Results: The median survival was 95 days and the mean observation period was
153 days. The mean value per person for formal healthcare cost was 30 983 Hong
Kong dollars [$HK] ($US3872, 1998 values). The distribution of cost values were
positively skewed. The regression analysis showed that age, days of observation
and survival were negatively related to cost per observed day, and the Child-Pugh
grading of severity of liver cirrhosis was positively related to cost per observed
day. A sensitivity analysis based on the regression equation indicated that non-
survivorship doubles the cost per case, increased severity as measured by the
Child-Pugh Index adds about 50% to the cost, and chemotherapy increases cost
2-fold.
Conclusions: The relatively modest average cost per patient with HCC in Hong
Kong reflects the short median survival and subsequently the limited use of in-
patient care and chemotherapy.

948
Chan et al.
Hepatocellular carcinoma (HCC) is one of the
Methods
world’s most common cancers with a particularly
high incidence in parts of Africa and Asia.[1] In
Hong Kong, it is the second most common malig-
nancy with an age-standardised incidence rate of
24.1 per 100 000 individuals.[2] With a population
of 6 000 000 in Hong Kong, 1707 new case reports
of HCC were registered in 1994.[2]
Surgical resection is the only potentially cura-
tive form of therapy; however, only 10% of patients
have an operable stage of disease at presentation.[3]
Of the remainder, a further 30% (based on experi-
ence at Prince of Wales Hospital) can be offered
active noncurative (i.e. palliative) treatment that
comprises locoregional therapies such as chemo-
embolisation or selective internal irradiation, and
intravenous systemic chemotherapy.[4] Patients
with HCC who have the poorest liver function on
presentation (around 60%) are not candidates for
active treatment and therefore receive best support-
ive care. During the course of best supportive care,
the patients may be transferred to an extended care
facility or a hospice care facility.
Very little is known about the cost of care for
HCC. Four groups of investigators[5-8] conducted
studies that were based on the assignment of costs
to specific care protocols, and thus were not based
on observation. Two observational studies exam-
ined the hospital costs of end-stage liver disease
(but not HCC)[9] and the social costs of colorectal
cancer with liver metastases.[10] Neither addressed
the specific treatment issues of HCC.
Patients and Follow-up
This study included all patients with HCC who
presented to the Department of Clinical Oncology,
Prince of Wales Hospital, a 1337-bed teaching hos-
pital in Hong Kong, between June 1996 and May
1997. The Prince of Wales Hospital is a publicly
funded hospital, and all cancer-related services are
provided free of charge. Excluded were about 10%
of patients who were treated surgically with cura-
tive intent during this period and who did not re-
enter the study. The diagnosis of HCC was made
either by histological examination of tumour tis-
sue, or imaging evidence of a space-occupying le-
sion in the liver together with a serum α-fetoprot-
ein level of above 500 g/L in a known carrier of the
hepatitis B virus surface antigen. Patients were en-
rolled during their first admission to hospital and
were thereafter tracked until death or loss to fol-
low-up.
The severity of cirrhosis was classified at the
time of first hospital admission according to Child-
Pugh grading (table I). The choice of subsequent
palliative strategy was determined at a Joint Hepa-
toma Clinic (including radiologists, oncologists
and surgeons) after discussion with the patients.
The strategies included best supportive care or ac-
tive noncurative strategies for patients with Child-
Pugh grades A and B cirrhosis. For patients with
Child-Pugh grade C cirrhosis, best supportive care
was considered the only appropriate strategy.
All the data for health service utilisation was
The present study had 2 objectives. The first was
to determine the patient-specific costs of palliative
care of HCC from a societal perspective using ob-
servational data from a large Hong Kong cancer
care programme. The second was to determine the
individual factors that drive patient–specific costs,
including patient age, severity of liver cirrhosis and
treatment modality, and to develop a model of cost
per case under alternative circumstances, including
varying lengths of survival, survival or nonsurvi-
val, and mode of care.
Table I. Severity of liver disease (Child-Pugh’s grading)[11]a
Clinical and biochemical Points scored for increasing
measurements
abnormalities
1
2
3
Encephalopathy
Ascites
Bilirubin (mg/100ml)
Albumin (g/L)
Prothrombin (seconds
prolonged)
None
Absent
1–2
>35
1-4
1 and 2
Slight
2-3
28-35
4-6
3 and 4
Moderate
>3
<28
>6
a
Graded according to total points score. A: score 5 to 6; B: 7 to
9; C: 10 to 15.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (9)

Cost of Palliative Care for HCC
949
collected prospectively by the research nurses.
From the date of accrual, all patients were inter-
viewed once per month until death or until the last
follow-up.
the WELCAN[12] workload system. The cost per
workload unit was based on the cost for the Prince
of Wales Hospital laboratory. Radiological ser-
vices were based on the diagnostic imaging com-
mon procedure/examination list of Alberta Health,
Canada.[13] Per minute technical costs were esti-
mated by the Prince of Wales Hospital radiology
services unit.
Informal services and indirect costs were mea-
sured using personal interviews. Respondents
were directly contacted and were requested to es-
timate informal healthcare resources, including the
amount spent for Chinese and herbal medicines,
supplies related to their illness, and the loss of in-
come from work for both patients and for care-
givers. The respondents were interviewed once per
month by one of the study nurses.
For our descriptive analysis, we calculated the
mean value and distribution of costs, by service or
type of resource, for those who used each type of
service or resource. A predictive (regression) anal-
ysis was conducted only for formal healthcare ser-
vices because of the lack of additional variables
which could be used to predict personal out-of-
pocket and indirect costs. As some patients were
lost to follow-up, we used a cost per day of obser-
vation (total formal cost divided by days of obser-
vation) as our dependent variable. Independent
variables included patient age, Child-Pugh grading
at hospitalisation (Child-Pugh grades A and B ver-
sus grade C), period of observation (admission to
hospital until death, loss of follow-up, or conclu-
sion of observation period), type of intervention
(chemotherapy and locoregional therapy versus
best supportive care) and nonsurvival versus sur-
vival at the end of the observation period or loss
to follow-up. As we found that this dependent
variable had a nonlinear relationship with our pre-
dictors, all regression solutions used natural log
transformations of costs. Based on the regression
results, we developed a sensitivity analysis for
measures of cost per case using a number of alter-
native scenarios.
Analytical Approach
A societal perspective was taken that included
all formal and informal services incurred by pay-
ers, caregivers and patients. The cost of any spe-
cific services that were used by a patient had 2
components: the number of units of services that
were used by the patient, and the cost of each of
these units. Measured resources included short-
term inpatient hospital care, continuing inpatient
care and hospice care, outpatient services (clinics,
diagnostic tests, and drugs), Chinese and herbal
medicines, supplies privately obtained, and the in-
come lost by caregivers and patients. All costs are
expressed in terms of Hong Kong dollars ($HK1 =
$US0.125; 1998 values).
All hospital and extended inpatient care was
measured in terms of patient days. Costs for the
number of short term inpatient days were valued in
terms of the per diem total cost for a cancer ward
(including physician services, diagnostic services,
pharmaceuticals and an overhead allocation) ac-
cording to the methodology developed by the Hos-
pital Authority of Hong Kong. Costs for extended
care hospital days (including palliative care) were
valued according to the average inpatient cost per
day (including an overhead allocation) for each
facility.
Outpatient care was divided into three catego-
ries – clinic visits (routine visits, chemotherapy
visits, and chemotherapy follow-up visits), phar-
maceuticals, and diagnostic services. Outpatient
clinic costs included nursing and physician costs
and supplies, and were derived as a cost per visit,
averaged over a sample of clinic days. The cost of
outpatient drugs was based on the formulary cost
at the Prince of Wales Hospital, and included an
additional cost per prescription for professional
pharmacists’ services. Diagnostic services were
valued in terms of workload units. Laboratory
workload units for each type of test were based on
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (9)

950
Chan et al.
land China, and contact was lost with 24. The me-
dian survival was 95 days. The mean observation
period was 153 days [standard deviation (SD) 171
days]. Given that the duration of observation dis-
tribution was skewed to the right, the median was
much less than the mean – 65 days. A total of 136
patients received best supportive care and 68 re-
ceived active noncurative treatment as the first
treatment following enrolment. Of the 68 who re-
ceived active noncurative treatment, 37 received
intravenous systemic chemotherapy and 31 re-
ceived locoregional treatments.
The types of healthcare services provided to the
204 patients were as follows: a total of 114 patients
were treated in both hospital inpatient and out-
patient settings; 45 patients were treated in 3 set-
tings – inpatient care, outpatient clinic, and nursing
homes; 23 in hospital only, without any other ser-
vices; 17 in inpatient hospital and nursing home
only, and 5 in all 4 settings – inpatient and outpa-
tient (clinic) hospitals, nursing homes and home
care.
Patient costs by type of service are summar-
ised in table III. The mean value for formal inpa-
tient short term stay care costs was $HK19 581
($US2481). This value also was markedly skewed
with a median value of $HK11 680 ($US1460).
The mean number of inpatient days per patient was
8.5 (SD 9.8). The average stay for inpatient nursing
care was 16.4 days and for inpatient hospice care
it was 6.5 days.
Results
Descriptive Analysis
204 patients were recruited and their charac-
teristics are listed in table II. The mean age was 57
years (139 were under 65 years of age) and 169
were male; 129 patients were classified as Child-
Pugh grades Aor B, and 75 patients were classified
as grade C. A total of 135 patients died within the
observation period; 14 survived beyond the obser-
vation period, and 55 were tracked until the last
follow-up visit or before being lost to our follow-
up attempts. Of those lost to follow-up, 13 died in
Hong Kong outside of study sites, 18 died in main-
Table II. Patient demographic and disease characteristics of the study
sample
Characteristic
Number of patients
Value
204
% of group
Age
Mean (y)
Standard value (y)
Number <65y
57.15
13.38
139
68
32
Number ≥65y
65
Sex
Male
Female
169
35
83
17
Child-Pugh Grade
Grade A and B
Grade C
129
75
63
37
Outcome
Atotal of 177 persons received some component
of outpatient services. Over all patients, including
those who did not receive any outpatient services,
the mean patient cost was $HK3158 ($US394;
1998 values), excluding chemotherapy. The distri-
bution for outpatient costs was skewed to the right.
For those who received some outpatient services,
the mean values for each type of service are shown
in table III. Patients who made outpatient visits av-
eraged 11.5 visits.
Death
Continuing
Lost to follow-up
135
14
55
66
7
27
Observation period (days)
Mean
Standard deviation
Median
Minimum
Maximum
153
171
65
1
733
Initial treatment at enrolment into study
Best supportive care
136
66.6
With regard to informal costs, 93 persons in-
curred transport costs, at a mean value of $HK1168
($US146). 10 patients incurred costs for Chinese
herbal medicines (mean value of $HK10 795 or
(palliative)
Noncurative treatment
locoregional therapies
systemic therapies
37
31
18.1
15.2
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (9)

Cost of Palliative Care for HCC
951
Table III. Patient costs ($HK; 1998 values)
Cost component
Number of cases
(with costs > 0)
Mean cost (for those Standard deviation Mean cost over all cases
with costs > 0)
(% of total)
Formal costs
Inpatient – acute care
Continuing inpatient care
Outpatient care
204
68
19 581
20 782
23 049
22 489
19 851 (33.6)
6927 (11.7)
accident and emergency
outpatient clinic – cancer clinic
outpatient clinic – diagnostic
outpatient clinic – drugs
chemotherapy
81
1105
1937
1051
1509
8661
1427
2692
1851
1916
4933
439 (0.1)
1101 (1.8)
701 (1.1)
917 (1.5)
1316 (2.2)
31 252 (52.9)
116
136
124
31
Total formal costs
Informal costs
Patient loss of income
Caregiver loss of income
Transportation
Herbal/Chinese medicine
Private costs
62
11
93
10
28
83 734
8500
1168
10 795
6210
96 970
8574
3575
11 689
10 191
25 449 (43)
458 (0.8)
532 (0.9)
529 (0.9)
852 (1.4)
Total informal costs
Total costs (formal + informal)
HK = Hong Kong.
27 821 (47.0)
59 073 (100)
$US1349) and 28 incurred other healthcare costs
(mean value of $HK6210 or $US776). With re-
gard to indirect costs, 62 patients reported a loss
of income, with a mean value of $HK83 734
($US10 466) during the observation period. There
was a wide variation with 12 patients having
costs under $HK30 000 ($US3750) and 6 over
$HK200 000 ($US25 000). 11 caregivers incurred
indirect costs of $HK8500 ($US1062) per care-
giver. The costs were nearly split evenly, with for-
mal costs accounting for 52.9% and informal costs
47.1% of the total.
The mean value per person for healthcare sys-
tem costs was $HK30 983 ($US3872). The stand-
ard deviation was $HK30 768 ($US3846). The
mean value of the cost per observed day was
$HK1049 ($US131) with a standard deviation of
$HK1702 ($US212).
days of observation, age, and survival were nega-
tively related to unit cost, while the Child-Pugh
grading and the two active treatment therapies
(systemic or logoregional) were positively related.
All variables were significant at the 0.05 level or
better.
The relationship between cost per day (trans-
formed from logarithmic into actual values) and
the number of observed days was estimated from
the regression equation for a base case (see table V
for base case assumptions). The estimated cost per
case for the base case was $HK40 052 ($US5006).
A one-way sensitivity analysis, which was con-
ducted by changing individual assumptions, indi-
cated that cost per case was quite sensitive to sur-
vival, severity, chemotherapy and locoregional
therapy. The highest projected cost per case was
$HK85 800 for the base case with chemotherapy,
and the lowest was $HK 19 379 for the base case
(except the patient survived). Thus, nonsurvivor-
ship doubles the cost per case, increased severity
as measured by the Child-Pugh Index (grade C)
adds about 50% to the cost, and chemotherapy in-
creases the costs two-fold (table V).
Regression Analysis
The regression analysis was conducted using
healthcare system costs only. The results of the re-
gression analysis are shown in table IV with pre-
dicted costs specified in log form. The number of
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (9)

952
Chan et al.
Table IV. Regression analysis results. Dependent variable: log of
Most patients with inoperable HCC have a back-
ground history of liver cirrhosis, and advanced
HCC exacerbates the severity of cirrhosis resulting
in a variety of complications in the terminal stages
of the disease. It is therefore not surprising that the
cost of palliative care for patients with HCC pre-
senting with Child-Pugh grade C is higher than the
cost of palliative care for patients with HCC pre-
senting with Child-Pugh grades A and B.
Patients with Child-Pugh grades A and B who
receive active noncurative treatment strategies
incur only moderately greater costs per day than
patients with Child-Pugh grades A and B who are
given best supportive care. None of the active non-
curative strategies have been shown to prolong sur-
vival in a randomised setting;[14-16] even a modest
improvement in disease-free survival for those on
active therapy might warrant the additional costs.
Nonsurvival was shown to be correlated with
costs. This finding is consistent with the presence
of an expenditure ‘bubble’ just prior to death,
which has been noted by some investigators.[17]
The costs for those who are lost to follow-up would
therefore understate the cost of the disease, since
these observations do not include the expenditure
bubble.
cost per observed day ($HK; 1998 values)
Variable
Constant
Coefficient
10.79
Standard errora
0.87
Child-Pugh grade
(A or B = 0, C = 1)
0.43
0.14
Observed days
Age
Systemic chemotherapy
(no = 0, yes = 1)
–0.70
–0.54
0.76
0.04
0.21
0.17
Locoregional therapy
(no = 0, yes = 1)
Survival or lost to follow-up
(yes = 0, no = 1)
R2
0.49
–0.72
0.77
0.17
0.12
a
Indicates coefficient is significant at α level of 0.05.
Discussion
In this study we developed an observational data
base for the palliative care of patients with HCC
who presented to a large public Hong Kong hospi-
tal. We included costs of healthcare services and
other, related costs (including transportation, her-
bal and Chinese medicines and indirect costs of
time loss for patients and caregivers). Our results
showed that healthcare system costs average
$HK30 983 ($US3872). We also developed a pre-
dictive model of formal costs. In this model, age,
the severity of cirrhosis on presentation, the obser-
vational period, and the treatment modality were
all significantly related to the cost per day.
Three previous studies presented estimates for
the cost of HCC, although none were based on
direct observation. Where possible, we have tried
to standardise the results of these studies to those
which we developed, by adjusting for survival or
Table V. Sensitivity analysis
Specifications of scenario
Projected cost per case
$HK (1998 values)
$US (1998 values)
5006
Base case: 150-day survival period, age 59y, Child-Pugh Index A or B, non-survivor 40 052
and not lost to follow-up, no chemotherapy, locoregional therapy
Base case, except the patient survived
Base case except 50-day survival period
Base case except 100-day survival period
Base case except 200-day survival period
Base case except 250-day survival period
Base case except Child-Pugh Index = C
Base case with chemotherapy
19 379
28 869
35 400
43 600
46 500
61 650
85 800
24 300
2422
3609
4425
5450
5813
7706
10 725
3038
Base case, but without locoregional therapy
$HK = Hong Kong dollars.
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (9)

Cost of Palliative Care for HCC
953
agers of Prince of Wales Hospital, Shatin Hospital and Brad-
bury Hospice during the course of this project.
observation time. The 6-month estimate of treat-
ment costs by Sarasin et al.,[8] for Swiss patients,
was $US36 000 (1994 values); this estimate inclu-
ded a hospital stay for palliative care of $US35 000.
Kim et al.[5] derived their estimates from US na-
tional data; they estimated costs for HCC to be
$US10 000 per annum (1994 values). Shiell et al.[7]
estimated the cost of a treatment protocol to be
Australian $A117 000 ($US61 000) [year of cost-
ing not stated]. The time period was not specified
for this protocol. We also note 2 observational
studies on liver-related disease: a US analysis by
Wong et al.[9] on the hospital costs of end-stage
liver disease (primarily oesophageal varices) and a
well-conducted UK study on the societal costs of
liver metastases in colorectal cancer.[10] In the lat-
ter study survival-adjusted costs were similar to
those of our study, though treatment processes do
differ.
The major omissions of our analysis were those
of the diagnostic phase of the treatment, the cost of
investigational agents, and missing follow-up data.
The inclusion of diagnostic costs would not impact
on early interventions, since similar costs would be
incurred, only earlier. These costs would be rele-
vant, in a decision-making phase, to cancer preven-
tion. The costs after loss to follow-up would likely
be higher than prior to follow-up, as the expendi-
ture bubble would occur at that time. We have ad-
justed for this by including survival in the regres-
sion analysis.
References
1. Simonetti RG, Camma C, Fiorella F, et al. Hepatocellular car-
cinoma. Dig Dis Sci 1991; 36: 962-72
2. Hong Kong Cancer Registry. Cancer incidence and mortality in
Hong Kong 1993-1994. Hong Kong: Hospital Authority,
1998
3. Okuda K. Primary liver cancers in Japan. Cancer 1980; 45:
2663-72
4. Shiu W, Dewar G, Leung N, et al. Hepatocellular carcinoma in
Hong Kong: clinical study on 340 cases. Oncology 1990; 47:
241-5
5. Kim WR, Poterucha JJ, Hermans JE, et al. Cost-effectiveness
of 6 and 12 months of interferon-α therapy for chronic hepa-
titis C. Ann Intern Med 1997; 127: 866-74
6. Bennett WG, Inoue Y, Beck R, et al. Estimates of cost-effec-
tiveness of a single course of interferon-α2b in patients with
histologically mild chronic hepatitis C. Ann Intern Med 1997;
127: 855-65
7. Shiell A, Briggs A, Farrell GC. The cost effectiveness of alpha
interpheron in the treatment of chronic active hepatitis C.
Med J Aust 1994; 160: 268-72
8. Sarasin FPGE, Hadengue A. Cost-effectiveness of screening for
detection of small hepatocellular carcinoma in western pa-
tients with Child-Pugh class Acirrhosis. Am J Med 1996; 101:
422-34
9. Wong LL, McFall P, Wong LMF. The cost of dying of end-stage
liver disease. Arch Intern Med 1997; 157: 1429-32
10. Durand-Zaleski I, Earlam S, Fordy C, et al. Cost-effectiveness
of systemic and regional chemotherapy for the treatment of
patients with unresectable colorectal liver cancer. Cancer
1998; 83: 882-8
11. Pugh RN, Murray-Lyon IM, Dawson JL et al. Transection of
the oesophagus for bleeding oesoghageal varices. Br J Surg
1973; 60: 646-9
12. Welsh Office, E.A.S.S.D. Welcan UK. Cardiff. Economics and
Statistical Services Division, Welsh Office, 1998
13. Alberta Health. 1995/6 technical portion of the diagnostic im-
aging common procedure/examination lost (cpel). Edmon-
ton, Alberta, Canada: Alberta Health, 1995
14. Groupe d’Etude et de. Traitement du Carcinome Hepato-
cellulaire. A comparison of lipiodol chemoembolization and
conservative treatment for unresectable hepatocellular carci-
noma. N Engl J Med 1995; 332: 1256-61
Conclusion
15. Lau WY, Ho S, Leung WT, et al. Selective internal radiation
therapy for non-resectable hepatocellular carcinoma with
intraarterial infusion of yttrium-90 microsphere. Int J Radiat
Oncol 1998; 40: 583-92
16. Lai CL, Wu PC, Chan GC, et al. Doxorubicin versus no anti-
tumor therapy in inoperable hepatocellular carcinoma: a pro-
spective randomized trial. Cancer 1988; 62: 479-83
17. Felder S, Meier M, Schmitt H. Health care expenditures in the
last month of life. J Health Econ 2000; 19: 679-95
The relatively modest average cost per patient
with HCC in Hong Kong reflects the very short
median survival and, subsequently, the limited use
of inpatient care and chemotherapy in these pa-
tients relative to those with other diseases treated
for extended periods.
Acknowledgements
This project is supported by a Full Project Grant of the
Health Services Research Committee, Hospital Authority,
Hong Kong. The authors would like to acknowledge the
support of the Hospital Chief Executives and Finance Man-
Correspondence and offprints: Dr Anthony T.C. Chan, De-
partment of Clinical Oncology, Chinese University of Hong
Kong, Prince of Wales Hospital, Shatin, Hong Kong, China.
E-mail: anthonytcchan@cuhk.edu.hk
© Adis International Limited. All rights reserved.
Pharmacoeconomics 2001; 19 (9)