Journal of Gastroenterology and Hepatology (1992) I, 463-468

LIVER A N D BILIARY

Analysis of cost-effectiveness of different strategies for hepatocellular carcinoma screening in hepatitis B virus carriers J. Y. K A N G , T. P. LEE, I. Y A P A N D K . C. L U N * Division of Gastroenterology, Department of Medicine and *Department of Community, Occupational and Family Medicine, National University Hospital, Singapore

Abstract A mathematical model was used to calculate the efficacy of screening to detect hepatocellular carcinoma at a resectable stage in hepatitis B virus carriers. Data relating to tumour incidence, efficacy of screening tests and tumour growth times were obtained from a literature review. Various tests were costed according to charges currently prevailing at the authors’ institution. T h e cost per early tumour detected is inversely proportional to tumour incidence. It is relatively low for populations with high incidences of hepatocellular carcinoma for example, male carriers over the age of 30. Both the costs and the proportions of early tumour detected increase with increasing frequency of screening. However, the use of ultrasonography at 10 monthly intervals or both ultrasonography and a-fetoprotein estimation at yearly intervals will detect 90% of tumours early at a cost of S$20 000 (US$11 800) per early tumour detected. T h e results would be significantly altered if tumour growth times were markedly different from those reported in the literature.

Key words: alpha-fetoprotein, cirrhosis, incidence, ultrasonography. INTRODUCTION Hepatocellular carcinoma (HCC) is probably the most common malignant tumour in males worldwide. It affects about 800 000 men every year.’ The median survival in patients with clinically apparent HCC is 2-4 months, and 2 year survival is virtually nil.293However, if HCC is detected at a pre-symptomatic stage it can be surgically resected with at least a potential for cure. There are marked geographical variations in the incidence of HCC, its frequency being higher in parts of Asia and Africa than in Western countries. In the USA, for example, the annual incidence of HCC is less than four per 100 000 compared with 150 per 100 000 in Taiwan.’ Hepatocellular carcinoma has been aetiologically linked to chronic hepatitis B infection and carriers of the virus are 10-100 times more likely to develop HCC compared with non-carriers.’ The annual incidence of HCC among hepatitis B virus (HBV) carriers in Asia is in the order of 350 per 100 000 in asymptomatic male subjects: 826 per 100 000 in patients with chronic hepatitis5 and as high as 8000 per 100 000 in cirrhotics6 It has been suggested that screening for HCC may be beneficial in high-risk populations.’ The number of HBV carriers worldwide is estimated to be 200 M.8 These subjects are at increased risk of development of HCC and

can potentially benefit from a screening programme. The incidence of hepatoma varies geographically and in different sex-age groups. Although several screening programmes have been set up in different parts of the ~ o r l d $ ~ -it’ is~ unclear which screening test or combination of tests is the most cost-effective. T h e optimal time interval between screening tests is also unknown. Decision analysis techniques will help to answer these questions.’

METHODS The variables The main variables determining the cost-effectiveness of screening for HCC in HBV carriers are as follows: incidence of hepatoma; sensitivity and specificity of screening tests; interval between successive screening tests; growth rates of untreated HCC; result of surgery; and cost of various diagnostic and therapeutic procedures. T h e two tests that have been extensively evaluated for HCC screening are serum a-fetoprotein estimation (AFPE) and abdominal ultrasound examination (US). It has been assumed that false-positives and false-negatives from one test are independent of those of the other.

Correspondence: Associate Professor J. Y. Kang, Division of Gastroenterology, Department of Medicine, National University Hospital, Lower Kent Ridge Road, Singapore 051 1. Accepted for publication 8 May 1992.

J. Y.Kang et al.

464 outcome

-i:e False Porltlve

SC,se"l"g

False Negative

'healthy HBV Carller'

Late

m

Since the sensitivity of AFPE is only 82% for symptomatic HCC,I4 we have arbitrarily set a maximum sensitivity for early HCC of 80% for the present calculations. A number of subjects without HCC have false-positive screening test results and undergo unnecessary further investigations. Their numbers are denoted by E/C where E is the false-positive rate of the screening test. The number of HCC detected by a screening programme is:

AB - AB (C - D ) / C

Neqalive

Nspative

'healthy HBV Carrler'

Figure 1 Possible outcomes of screening for HCC.Screening

test results.

Outcome of screening A diagram showing the possible outcomes of screening is shown in Fig. 1. The aim of screening is to detect HCC at an 'early' and therefore resectable stage rather than a 'late' stage which would be uniformly fatal. Although small size does not equate necessarily with an early lesion, it is generally thought that results of surgical resection are better for small HCC compared with large ones.' In this study, 'early' HCC have been defined as lesions under 3 cm in diameter. Some positive screening test results are false-positives. These patients would undergo further investigations (assumed for this discussion to be computerized tomography, hepatic angiography and liver biopsy). Individuals with early HCC would then undergo resection. Some negative screening results are false-negative results. These patients eventually present with late HCC. The majority of patients with negative screening tests remain 'healthy' hepatitis B virus carriers.

Number of early hepatomas detected by screening The number of early HCC detected depends on the incidence of HCC(A) and the sensitivity of the screening test(s) (B). If the interval between screening tests (C) is longer than the time taken for a HCC to grow from a detectable stage to a late stage (D),a number of HCC will be diagnosed only at an incurable stage. The number of such failures is proportional to (C - D ) and is denoted by (C - D)/C. If D > C, on the other hand, some early HCC missed by one screening test may be picked up by repeat screening. This number is proportional to ( D - C ) and is denoted by (1 - B) ( D - C)/C provided D 5 2C. If D > 2C, some early HCC missed by a second screening test may be picked up by a third or subsequent test. However, this would not be taken into account because very few HCC are picked up with US (sensitivity in the order of 88%) at third and subsequent tests. When D > 2C, the number of HCC picked up after having been missed at initial screening is still (1 - B) (D- C)/C where D = 2C. With AFPE (sensitivity in the order of 68%; Table 1) one repeat test increases its sensitivity to 90%.

=

ABD/C when C > D

or

AB +AB (1 - B) ( D - C ) / C= A B / C [ C + (1 - B) ( D - C)] when D > C provided D S 2 C . If D > 2C, the second formula still applies but the value of D would be fixed at 2C. When AFPE is used its maximum sensitivity is set at 80% regardless of the number of times the screening test is used. The cost of the whole screening programme is F/C + G(AB + E / C ) where F and G are the costs of the screening test@)and investigations, respectively. The cost incurred for every HCC detected would be:

F/C

+ G ( A B + E / C ) - F + G(ABC + E ) ABD/C

or

ABD

F/C + G ( A B + E / C ) A B / C [ C + (1 - B ) ( D - C)] F + G(ABC + E ) AB[C + (1 - B ) ( D - C)] depending on whether C > D or D > C and provided D 5 2C. If D > 2C, the second formula still applies but the value of D would be fixed at 2C. The proportion of HCC detected by a screening programme would be:

AB - AB' ( A when C > D or AB

')Ic

shortening to BD/C

+ AB(1 - B ) ( D - C)/C

shortening to A B / C [ C + (1 - B ) ( D - C)]

when D > C and provided D 5 2C. If D > 2C, the second formula is used but D is fixed at 2C.

Literature review A MEDLINE literature search was conducted on HCC for the years 1975-March 1989. The search was under the following headings: epidemiology, mortality, diagnosis, screening and surgery. Citations were generated by the Australian MEDLARS Service using programs supplied by the US National Library of Medicine and Compact Cambridge MEDLINE CD-ROM. Studies on the use of AFPE and US in early HCC were selected and data were averaged to obtain an assessment of the sensitivity of these two tests. Results of studies in

Hepatocellular carcinoma screening

46 5

Table 1 Estimates of various variables pertaining to early

Incidence of heDatoma in heoatitis B virus carriers’(Ma1e > 30 y’ears ~ l d ) 927/105/year ~ * ~ 68.0% Sensitivity of AFPE;-~;;;;-~~ Specificity of AFPE 19.9% Sensitivitvof uS5,’6,20-22,25-27,35 87.8% 98.1% Specificity of US^*^^ Growth time of HCC from 1-3 cm17-18,28,36 1.63 (0.38-8.96) years [median (range)] Cost of AFPE $30.00* cost of us Cost of liver biopsy, CAT scan and

angiography *S$l .OO

=

$100.00* $2 075.00*

90 80 70 60 50 -

100

hepatomas

2 b 0

m

s vI

40 30

-

20 10

0

I



I



I

.

I

.

I

.

I

.

US $0.59 (approx.).

which HBV carriers or cirrhotics were screened for early HCC were used to obtain an average specificity for positive test results. One centimetre is taken as the minimum detectable size of a HCC. The time taken for a HCC to grow from a diameter of 1-3 cm is known for 65 patients in the l i t e r a t ~ r e . ’ ” ’ The ~ ~ ~ distribution ~~~~ of these 65 tumours is skewed (mean = 2.25 years, median = 1.63 years, s.d. = 1.94 years, range = 0.388.96 years). In our analyses we have used each of these values in 65 separate calculations and taken the mean of the results. T h e calculations therefore assume that the growth times of these 65 HCC are representative of HCC in general. The costs (S$l.OO = US$0.59, approximately) of various diagnostic and therapeutic procedures used for the analysis are those pertaining to the author’s institution at the time of writing (Table 1).

RESULTS Factors affecting the proportion of early HCC detected The proportion of early HCC detected is directly related to the sensitivity of the screening test(s). It is inversely related to the interval between screening tests and the time interval taken for a HCC to grow 1-3 cm.

Choice and frequency of screening tests Ultrasound is more sensitive but also more costly compared with AFPE. Concurrent use of both tests increases both the sensitivity and the cost incurred. Irrespective of the choice of screening test($, increasing the frequency of screening will increase both the cost of the programme as well as the proportion of HCC detected at an early stage (Fig. 2).

Cost per early HCC detected The cost per early HCC detected is dependent on the incidence of HCC in the population screened (Fig. 3). It

60

55 50

-

I

0

4



1

8

.

I

12

.

I

16

20

24

Screening interval (months)

Figure 2 Influence of screening intervals on (a) cost per early HCC detected and (b) proportion of HCC detected at an early stage. For this analysis, the HCC incidence was taken to be 927/105/year (a) AFPE; (4) US; (+) BOTH.

remains relatively low as long as the HCC incidence is above 500/100 000/year, for example, in male HBV carriers more than 30 years old.439Frequencies of around 4800/100 000/year, such as may be encountered in cirrhotic^,^' would result in a cost per HCC diagnosed of under S$10 000, using average values for the different variables listed in Table 1 and applying both screening tests once a year. Disregarding the proportion detected, the lowest cost per early HCC detected occurs with a screening interval of 14 months if AFPE were used or 18-20 months if US alone or both US and AFPE were used. Increasing the interval between screening tests does not reduce the cost per early HCC detected.

Cost versus proportion of early HCC detected The lowest cost per early H C C detected can be achieved using AFPE alone at a screening interval of 14 months. However, with this screening strategy only around 65% of HCC will be detected at an early stage. Figure 4 compares the cost per early HCC detected with the propor-

3. Y. Kung et al.

466 50

Sensitivity testing

40

Sensitivity of AFPE and US The effect of varying sensitivities of the screening tests on

35 30 25

20 15

-

10 -

5-

0

1000

2000

3000

4000

5000

Incidence of HCC (per 100 0001yOar)

the screening intervals required to detect particular proportions of HCC at an early stage using different strategies was examined. For strategies involving AFPE the screening intervals remain relatively constant when the sensitivity varies 60-83% (Fig. 5a; as in 15 of 16 studies). The sensitivity of US has little influence on screening intervals (Fig. 5b) provided it does not fall below 78%. Only one of nine studies reported a sensitivity below this level. This remains so if sensitivities of either screening test are varied when both tests are used concurrently (Fig. 5c).

Figure 3 Influence of the incidence of HCC on cost per early HCC detected Screening was carried out annually. (U) AFPE; (+) US; (+) BOTH.

+ 60%

*

68%

75%

I 83.3%

164

100

-

.0 0

P2

80

60 70

Y)

50

-

30 20 40

10

6 ln

5

-

50

80

70

60

90

100

4

24

-

+ 59%

(b)

-E+

60

70

80

90

1 00

HCC detected (%)

Figure 4 Cost effectiveness plot of three different screening strategies using different screening intervals. (-El-) AFPE; (4) US; (+) BOTH. Numbers refer to screening intervals in months.

or (iii) concurrent estimation use atof yearly both screening intervals tests. would be alone; a-Fetoprotein the most cost-effective strategy to diagnose 70% of HCC at an early stage, costing only S$7500 per early HCC detected. To diagnose 80% of HCC at an early stage, the use of AFPE at 4 monthly intervals, US at 16 monthly intervals or both AFPE and US at 18 monthly intervals are approximately equal in terms of cost-effectiveness. A cost of approximately S$15 000 per early HCC diagnosed is incurred for each of these three strategies. This means

(d

>

L

87.8%

* 93%

E

'6

a, E

'2

E

8

C

a,

*

0

.

.

\

.

,

-r, --g ln

24

-

16

L

m

.-c

0,

.5 C al

?!

.

.

,

.

.

.

. . .

I

90

80

I

100

- I

+ 94% 961%]

-

(c)

20-

2

.

ro

60

-

-87%

99.5%

8-

'2

4 -

v)

O

T

.

*

.

, .. :

,

. . .

, .

.

(i)

98% 77 9%

,

,

(11)

467

Hepatocellular carcinoma screening

Effect of variation in HCC growth times The time taken for HCC to grow from 1-3 cm has a major influence on the optimal frequency of screening (Fig. 6) whether one or both tests were used.

Cost of screening tests The cost of AFPE at the National University Hospital is 0.3 times that of US. If AFPE cost less than 0.3 relative to US, the strategies using AFPE would be more costeffective. Conversely if serology was more than 0.3 times as costly as US, strategies using US, would become more attractive. However, the cost ratio of AFPE : US (within the range 0.1-0.5) did not affect the frequency of screening required to diagnose different proportions of HCC at an early stage when both modalities were used together.

DISCUSSION The question of whether detection of early HCC prolongs patient survival has not been addressed because the prognosis of symptomatic HCC with a median survival rate of approximately 4 months is so dismal, it seems logical to detect and treat HCC at a pre-symptomatic stage. Survival rates of 90 and 59% at 1 and 4 years have been rep~rted.~’.However lead-time bias has not been excluded and the better survival rates of early HCC patients after hepatic resection may be more apparent than real. Hepatic function appears to be a major determinant of survival.38Survival rates of 96% at 1 year have, for example, been reported for untreated early HCC in Child’s A patients.28 In another recent study in which patients with cirrhosis from different causes were screened regularly only 17 of 59 (29%) HCC detected were potentially operable.39 Sixty per cent of those undergoing surgery had tumour recurrence within 1 year. Despite these uncertainties screening programmes are being carried out in various T h e present study compares the cost-effectiveness of various screening programmes in terms of the number of early HCC

1.63 years

i

v)

fc 20 -

:

.

1.00 year

o ! . 60

.

1

1

70

.

.

y

. , .

.. , . . . ,

80 90 HCC detected (X)

100

Figure6 Effect of varying HCC growth times on optimal screening intervals to detect particular proportions of early HCC using both tests. For this analysis, the HCC incidence was taken to be 927/105/year.

detected. The results suggest that the cost per early HCC diagnosed remains relatively low as long as the frequency of HCC in the population is at least 500/100000/year. T h e most cost-effective frequency for HCC screening is using both AFPE and US annually. This strategy will detect 90% of HCC at an early stage. The calculations are, however, dependent on the time taken for early HCC to grow 1-3 cm. Should many HCC grow faster than what is suggested in the current literature the proportion diagnosed at an early stage will be reduced. More data on the growth rates of early HCC are required to confirm the present observations. It was assumed that hepatic resection is the only curative treatment for HCC. Hepatic transplantation may in time become a recognized treatment, in which case curative surgery may be possible for HCC considered ‘late’ by present standard^.^' Screening tests may then be performed less frequently. These findings are relevant in planning health care for the 200 M HBV carriers worldwide. Knowing the costs involved at particular incidences of HCC, the feasibility of screening programmes and the interval between screening tests can be better assessed.

REFERENCES 1. RUSTGIV. K. Epidemiology of hepatocellular carcinoma. Gastroenterol. Clin. North Am. 1987; 16: 545-51. K., OHTSUKI T., OBATAH. er al. Natural history of 2. OKLJDA hepatocellular carcinoma and prognosis in relation to treatment. Study of 850 patients. Cancer 1985; 56: 918-28. 3. KEWM. C. & GEDDESE. W. Hepatocellular carcinoma in rural southern African blacks. Medicine 1982; 61: 98-108. R. P., HWANGL. Y., LIN C. C. & CHIEN.C. S . 4. BEASLEY Hepatocellular carcinoma and hepatitis B virus. A prospective study of 22 707 men in Taiwan. Lancet 1981; ii: 112932. 5. LIAWY. F., TAID. I., CHUC. M. et al. Early detection of hepatocellular carcinoma in patients with chronic type B hepatitis. Gastroenterology 1986; 90: 263-7. N., MoTolKE Y.era/. Prospective study 6. OBATAH., HAYASHI on the development of HCC from liver cirrhosis with persistent hepatitis B virus infection. Cancer 1980; 25: 741-7. D. J. Screening studies and markers. 7. JONESD. B. & KOOREY Gastroenterol. Clin. North Am. 1987; 16: 563-73. F. & GUSTI. D. Viral hepatitis. Bull. WHO 8. DEINHARDT 1982; 60: 661-91. W. L., LANIERA. P., MCMAHONB. J., FITZGER9. HEYWARD ~1.11M. A., KILKENNY S. & PAPROCKI T. R. Early detection of primary hepatocellular carcinoma. Screening for primary hepatocellular carcinoma among persons infected with hepatitis B virus. 3AMA 1985; 254: 3052-4. 10. LOHIYA G., LOHIYA S., FISCHER T., NGUYEN H., Vuu T. & HUMEREZ R. Serum a-fetoprotein in 714 mentally retarded carriers of hepatitis B surface antigen. Eur. J. Epiderniol. 1988; 4: 181-4. 11. SHANGHAI COORDINATING GROITFOR RESEARCH ON LIVER SHANGHAI. Diagnosis and treatment of primary CANCER,

468

hepatocellular carcinoma in early stage : Report of 134 cases. Chinese Med. J. 1979; 92: 801-6. 12. TANG Z. Y., YANGB. H., TANG C. L., Yu Y. Q., LINZ. Y. & WENGH. Z. Evaluation of population screening for hepatocellular carcinoma. Chinese Med. J. 1980; 93: 795-9. M. C., FINEBERG H. V., ELSTEINA. S. et al. 13. WEINSTEIN Clinical Decision Analysis. W. B. Saunders, Philadelphia, 1980; 228-65. 14. WEPSICH. T. 81 KIRKPATRICK A. Alpha-fetoprotein and its relevance to human disease. Gastroenterology 1979; 77: 787-96. 15. SHEUJ. C., SUNGJ. L., CHEND. S. et al. Early detection of hepatocellular carcinoma by real-time ultrasonography. A prospective study. Cancer 1985; 56: 660-6. 16. EBARAM., OHTO M. & KONDOF. Strategy for early diagnosis of hepatocellular carcinoma (HCC). Ann. Acad. Med. Sing. 1989; 18: 83-9. 17. SHEUJ. C., SUNGJ. L., CHEND. S. et al.Growth rate of atymptomatic hepatocelluar carcinoma and its clinical implications. Gastroenterology 1985; 89: 259-66. M., OHTOM., SHINAGAWA T. et al. Natural history of 18. EBARA minute hepatocellular carcinoma smaller than three centimeters complicating cirrhosis. A study in 22 patients. Gastroenterology 1986; 90: 289-98. 19. CHEND. S., SUNGJ. L., SHEUJ. C. et al. Serum alpha fetoprotein in the early stage of human hepatocellular carcinoma. Gastroenterology 1984; 86: 1404-9. T., OHTOM., KIMURA K. et al. Diagnosis and 20. SHINACAWA clinical features of small hepatocellular carcinoma with emphasis on the utility of real-time ultrasonography. A study in 51 patients. Gastroenterology 1984; 86: 495-502. 21,, SHEUJ. C., SUNGJ. L., CHEND. S. et al. Ultrasonography of small hepatic turnours using high-resolution linear-array real-time instruments. Radiology 1984; 150: 797-802. 22. CHEND. S., SHEUJ. C., SUNGJ. L. et al. Small hepatocellular carcinoma-A clinicopathological study in thirteen patients. Gastroenterology 1982; 83: 1109-19. Y., OHNISHIK., NOMURA F. & IIDAS. Plasma 23. TANABE abnormal prothrombin levels in patients with small hepatocellular carcinoma. Am. J. Gastroenterol. 1988; 83: 1386-9. 24. OKUDAK., NAKASHIMA T., OBATAH. & KUBOY. Clinicopathological studies of minute hepatocellular carcinoma. Analysis of 20 cases, including 4 with hepatic resection. Gastroenterology 1977; 73: 109-1 5. 25. HIGASHI T., TOBEK., ASANOK. et al. Ultrasonographic characteristic of small hepatocellular carcinoma. Acta Med. Okayama 1988; 42: 151-7. 26. OHTOM., TSUCHIYA Y. & SAISHO H. Carcinoma of the liver and the biliary system. Asian Med. J. 1983; 27: 307-16.

3. Y.Kung et al. 27. KOBAYASHI K., SUGIMOTO T., MAKINOH. et al. Screening methods for early detection of hepatocellular carcinoma. Hepatology 1985; 5: 1100-5. R., Fusco G. et al. Asymptomatic 28. COTTONEM., VIRDONE hepatocellular carcinoma in Child’s A cirrhosis. Gastroenterology 1989; 96: 1566-71 T., KUBOTA Y., MIZUNO T. & SAMESHIMA 29. SEKIT., NONAKA Y. Ultrasonically guided percutaneous ethanol injection therapy for hepatocellular carcinoma. Am. 3. Gasrroenterol. 1989; 84: 1400-7. 30. LOK A. & LAI C. L. Alpha-fetoprotein monitoring in Chinese patients with chronic hepatitis B virus infection: Role in the early detection of hepatocellular carcinoma. Hepatology 1989; 9: 110-15. 31. COTTONEM., TURRI M., CALTAGIRONE M. et al. Early detection of hepatocellular carcinoma associated with cirrhosis by ultrasound and alpha-fetoprotein : A prospective study. Hepatogastroenterology 1988; 35: 101-3. S. J., WOODJ. R., KLEEG. G. & Go 32. CZAJAA. J., BEAVER V. L. W. Frequency and significance of serum alphafetoprotein elevation in severe hepatitis B surface antigennegative chronic active hepatitis. Gastroenterology 1987, 93: 687-92. D. G., ENDOY. & CLARKE C. Alpha fetoprotein 33. WOODFIELD in liver disease in Papua New Guinea. Aust. N.Z. 3. Med. 1976; 6: 117-20. 34. WEBER P., WEBEROVA D. & MARTINEKK. Alphafetoprotein, carcinoembryonic antigen and various biochemical tests in patients with tumorous and inflammatory liver diseases. Neoplasma 1988; 35: 605-13. T., MATSUIO., SUZUKI M. & IDA M. Diagnosis 35. TAKASHIMA and screening of small hepatocellular carcinomas. Radiology 1982; 145: 635-8. 36. YOSHINO M. Growth kinetics of hepatocellular carcinoma. Jpn. 3. Clin. Oncol. 1983; 13: 45-52. N., YUKAYA H., CHANGY. C. et al. Appraisal of 37. NAGASUE hepatic resection in the treatment of minute hepatocellular carcinoma associated with liver cirrhosis. Br. J. Surg. 1987; 74: 836-8. 38. OHNISHIK., TANABE Y., RYU M. et al. Prognosis of hepatocellular carcinoma smaller than 5 cm in relation to treatment: study of 100 patients. Hepatology 1987; 7: 1285-90. M., FRANCHIS R. D., NINNO I. D. et al. Hepato39. COLOMBO cellular carcinoma in Italian patients with cirrhosis. N . Engl. J. Med. 1991; 325: 675-80. F., BELLIL. S. et al. Reappraisal of 40. BELLIL., ROMANI surgical treatment of small hepatocellular carcinomas in cirrhosis. Dig. Dis. Sci. 1989; 34: 1571-5.

Analysis of cost-effectiveness of different strategies for hepatocellular carcinoma screening in hepatitis B virus carriers.

A mathematical model was used to calculate the efficacy of screening to detect hepatocellular carcinoma at a resectable stage in hepatitis B virus car...
492KB Sizes 0 Downloads 0 Views