|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2003 American Society for Clinical Oncology Simple Clinical Prognostic Model for Hepatocellular Carcinoma in Developing Countries and Its ValidationFrom the Departments of Gastroenterology and Internal Medicine, Singapore General Hospital; and Division of Clinical Trials and Epidemiological Sciences, National Cancer Center, Singapore. Address reprint requests to Chee Kiat Tan, FRCP, Department of Gastroenterology, Singapore General Hospital, Outram Rd, Singapore 169608, Republic of Singapore; email: gm2tck{at}sgh.com.sg.
Purpose: More than 80% of hepatocellular carcinomas (HCCs) worldwide occur in developing countries, especially in Asia. It often presents at an advanced stage beyond treatment. In this circumstance, a simple prognostic model is useful. Previous prognostic models require radiologic and laboratory investigations that are not readily available in developing countries. Our aim is to formulate and then validate a simple clinical prognostic model for HCC in an Asian population using only clinical parameters and with serum alpha-fetoprotein (AFP) as the sole laboratory test. Patients and Methods: Cox regression modeling was performed on several clinical parameters and serum AFP level in 397 patients with HCC who received only supportive care in Singapore. A later group of 324 HCC patients from an Asia-Pacific–wide randomized trial was then used to validate the model. Results: Ascites, physical performance status, and serum AFP were independently predictive of survival. Cox analysis yielded a simple score based on these three variables that categorizes patients into low-, medium-, and high-risk groups with 6-month survivals of 43%, 21%, and 5%, respectively. The prospective validation data provided corresponding estimates of 33%, 15%, and 3% and give confirmation of the utility of the simple model. Conclusion: We have formulated and prospectively validated a simple prognostic score for untreated HCC that only requires a clinical evaluation for ascites and physical performance status and measurement of serum AFP. This simple model is particularly apt for developing country circumstances and can also be used to select patients for treatment trials.
HEPATOCELLULAR CARCINOMA (HCC) is one of the most common cancers in the world. However, more than 80% occur in developing countries in Asia.1 Most cases are related to chronic hepatitis B and C infection and, as hepatitis B is endemic in Asia, it is not surprising that the highest incidence of HCC is also in Asia.1 Unfortunately, most cases of HCC in developing countries present late, and curative treatment via resection or liver transplantation is not possible.2,3 Although Singapore is no longer classified as a developing country, this late presentation situation is also paralleled in this country in South-East Asia, where the incidence is high at 19 per 100,000 population per year4; only approximately 5% are amenable for curative surgery.3 As most patients present with untreatable HCC, it is useful to be able to predict their probability of surviving the next few months. However, although there are several published prognostic models for HCC, they were all based on developed Western populations, included treated patients, and often involve complicated calculations and measurement of laboratory parameters that are not readily available in much of Asia.5–7 Because HCC is very common in most countries in the region, the use of alpha-fetoprotein (AFP) measurements (a straightforward laboratory test) has become common practice despite other health service provision difficulties. MEDLINE search from 1966 until 1997 revealed only a single report based in a developing nation that specifically examined prognostic factors in patients with untreated HCC, but the study cohort consisted of only 70 patients.8 This situation remained unchanged until some 5 years later, when a study of Hong Kong Chinese (also including treated patients) describing a prognostic index was published; this study included the laboratory measures of bilirubin and alkaline phosphatase.9 The aim of our study was to first develop, and then validate on new data, a simple prognostic model for patients with untreated HCC for use in developing countries.
Study Populations Index group. A total of 587 patients with HCC were seen in the Department of Gastroenterology, Singapore General Hospital over the period of July 1989 through October 1997. HCC was diagnosed based on histology, diagnostic AFP level with a mass lesion on imaging, or a positive angiography with lipiodol uptake. Of these, 49 (8.3%) were operable and received surgery, 45 (7.7%) had chemoembolization, and 94 (16%) received systemic chemotherapy. The remaining 399, except two who died on the day of diagnosis at admission, received only supportive care and were used to develop the prognostic index. Supportive care included counseling, hospice care, and medications such as diuretics and analgesics when indicated. Several clinical parameters were recorded at the time of diagnosis. These included age, sex, ethnicity, history of alcohol intake, physical performance as assessed by Zubrod performance score (ZPS), and presence of ascites on physical examination. An alcohol intake of more than 60 g/d for more than 5 years was considered significant. Serum AFP was measured using a commercial assay system (Axsym system; Abbott Laboratories, Abbott Park, IL). Patients were categorized into five groups based on AFP value as shown in Table 1  . All patients were reviewed in the clinic at least every 1 to 2 months.
Survival census was performed on March 31, 1999, with the Singapore National Registry of Births and Deaths. The law requires that the registry be notified of all deaths in Singapore, with details of date and cause of death. Validation group. Patients from the Asia-Pacific–wide multicenter trial for treatment of inoperable HCC and recruited over the period of April 4, 1997, to June 8, 2000, were used to validate the prognostic model.10 There were 329 patients from 10 centers in nine countries in the Asia-Pacific region (Myanmar, Hong Kong, Singapore, Indonesia, Thailand, Malaysia, South Korea, Australia, and New Zealand) who were randomized in a double-blind fashion to either tamoxifen treatment or placebo. Survival information was available on all but five of these patients. All consenting patients had their baseline characteristics recorded at the time of diagnosis of HCC, after which they were randomly assigned to the trial and routine follow-up until death. The trial was initiated through the National Medical Research Council, Singapore.
Statistical Analysis
Validation group. Survival was calculated from the date of randomization. The SS derived from the index group was used to calculate for each patient in this validation group their corresponding SS. On the basis of this score, the patient was then assigned to one of the low-, medium-, and high-risk groups as previously defined, and the survival experience summarized using the Kaplan-Meier technique. From these curves the proportion alive at 6 months was estimated and the corresponding observed separation (OSEP = pHigh - pLow) calculated.12 All calculations were performed using standard statistics software (SPSS version 10.0.5; SPSS Inc, Chicago, IL).
Index Group Patient characteristics of the 397 patients of the index group are shown in Table 1  (for convenience, those characteristics ultimately selected for the SS of Table 3 are presented first). Most of the patients were ethnic Chinese. As expected, male patients far outnumbered female patients. Median age at the time of diagnosis of HCC was 63.5 years. Chronic hepatitis B was the most commonly associated etiology. Most of the patients were not treatable because of advanced malignant disease (majority were International Union Against Cancer tumor-node-metastasis system stage III or worse). At the census only seven patients (1.8%) remained alive. Of those who had died, all died of a liver-related cause. None died of unnatural causes. Overall survival at 6 months was 23%.
Individual (univariate) Cox regression analysis of all the clinical parameters listed in Table 1 and serum AFP level showed that the major variables influencing survival are ZPS, presence of ascites, and AFP (Table 2). Once these variables were assessed, little extra information was provided by age, sex, ethnicity, or significant alcohol history.
On the basis of the odds ratios in Table 2
Examination of Table 3
The Cox score was then divided by the smallest regression coefficient, namely 0.290, and rounded off to the nearest integer to yield a simplified survival score SS = 2A + P + F. On this basis, the minimum SS = 0 and the maximum SS = 7. Plots of the corresponding eight survival curves suggested that three groups that have significantly different prognosis are constituted (Table 5 ; Fig 1). From these, we arrive at PSEP = 0.947-0.571 = 0.376.
Validation Group Calculating the SS for the validation group gave a higher proportion in the low-risk category than for the index group and correspondingly fewer in the high-risk group (Table 5 ). The survival curves, corresponding to those of Figure 1, are shown in Figure 2. From these, we arrive at OSEP = 0.972-0.666 = 0.306.
HCC is a common malignancy in Asia and other developing countries. Unfortunately, many patients present late, at a stage when the malignancy is too advanced or the underlying liver function is too poor to allow curative resection.2,3 Therefore, in addition to offering supportive care, it is often important to provide some guidance as to the likely survival time. None of the prognostic models published so far is suitable for use in many Asian or developing countries. This is because they were based on developed (mainly Western) populations, have not been validated in other patient groups, and require several laboratory and radiologic tests for input into the prognostic model.5–7,9 It is impossible to apply such models in most developing countries where many laboratory and radiologic investigations are not readily available because of poor socioeconomic conditions and underdeveloped transport infrastructure. A search of MEDLINE from 1966 to 1997 yielded only one study that is directly comparable with ours; that is, on prognostic factors in Asian patients with untreated HCC.8 However, this study was based on a population of only 70 patients. Furthermore, several laboratory tests as well as liver scintigraphy were required, and the authors also admitted that they were unable to devise a useful clinical staging model. Some 5 years later, a prognostic index based on the Hong Kong Chinese patients with HCC has now been reported.9 Although the study group included both treated and untreated patients and has not been validated on external data, their prognostic index included the same three parameters that we propose and so confirmed their prognostic role. However, their full index required advanced radiologic staging of the HCC as well as several other blood tests. The simplest prognostic model would be one that requires only variables that can be documented simply from the clinical history and physical examination without the need for sophisticated tests. Hence, in deriving our prognostic model by Cox regression analysis, we included only variables that can be determined from the clinical history and physical examination. Although encephalopathy is also a clinical sign that can be easily determined, we did not include it in our analysis, as only 13 patients had encephalopathy. As for jaundice, we did not include it as a clinical variable because of the recognized individual variability of clinicians in the detection of mild jaundice. Although the measurement of serum AFP level is a laboratory test, it is usually available even in developing countries, because HCC is very common and the test is recognized as basic for the diagnosis of HCC. It may be argued that in deriving a prognostic model that is intended for application in regions where facilities are not readily available, we used sophisticated technology to determine that our patients had untreatable disease. In Singapore, we are fortunate that such facilities are available, and we are able to use them to determine a patient’s suitability for treatment. It is only through these modalities of investigation are we were able to conclude that the HCC is not treatable and select these patients for entry into the prognostic modeling.
There was a relatively high proportion (20%) of index patients in which the ZPS had not been reported by the attending physician. The prognosis of these patients, however, when treated as a single group, was intermediate, between ZPS = 1 and ≥ 2 (Table 2 Our prognostic model requires only three variables. Two of the three variables are readily determined from just the clinical history (physical performance status based on Zubrod performance score) and physical examination (presence of ascites). Physical performance status has been shown by other studies to be associated with prognosis in HCC5,6,9 and indeed with other malignancies.13 The presence of ascites has also been shown by others to be significantly related to survival, either by itself or as part of the Child-Pugh score.5,7,9,14 Although it can be argued that these parameters may simply reflect severe end-stage liver disease rather than tumor morbidity per se, this issue is not pertinent in clinical practice, as most patients will eventually die from progressive liver failure or complications of end-stage liver cirrhosis. None of our patients died from a non–liver-related or unnatural causes. Therefore, regardless of the exact liver-related causation of death, the proposed model can still be used to predict the patient’s survival. As for the other clinical parameters analyzed, ethnicity was not a significant determinant of survival. This is expected, as the Asian ethnicities of Chinese, Malays, and Indians are closely related, and differences have not been found even between races as disparate as whites and blacks.15 Sex has also been shown by several studies not to be a significant determinant of survival.2,5,6,9,14,15 Like our experience, age has been shown by most studies to be unimportant in prognosis,2,6,8,9,14 although one small study of 56 patients found poorer prognosis with increasing age at diagnosis.15 Alcohol consumption has not been in analyzed in most studies. Two studies that did include alcohol consumption in the analysis did not find it a significant determinant of survival.5,14 Thus our findings are in agreement with most other studies albeit both treated and untreated patients are included in these studies. Serum AFP level has to be assayed in making a diagnosis of HCC and is available in most developing countries. Hence, we included it in our prognostic model, and it was found to be a significant independent variable as well. Similarly, serum AFP level has also been shown by others to be significantly related to survival in HCC.6,9,16,17
The process of developing a simple model balances practical with statistical considerations. Practical issues include using categories for serum AFP, classifying any unknown characteristics into the worst group, and rounding the regression coefficients to obtain the relative weights attached to each variable. Statistical considerations suggest that enough data are available to ensure that the number of events, here patient deaths, is sufficient for the modeling process. This is expressed as the number of events per candidate variable (EPV). For the index cases with n = 397 cases and the v = 11 candidate variables of Table 1 The PSEP is a suggested measure of the estimated ability of a rule to separate individuals into different prognostic groups.12 Using an independent cohort of patients to validate our prognostic model, the OSEP of 0.31 is close to the PSEP of 0.38, thus demonstrating validity of our prognostic model. Although the validation group was obtained from a treatment (tamoxifen) trial, the treatment was a palliative one, and the survival differences were minimal.10 By assigning the categories for the variables into corresponding numerical scales for the Cox regression analysis and simplifying the final Cox score, we have derived a simple prognostic model for untreated HCC in an Asian population. The survival score formula of 2A + P + F can be calculated by the bedside once the serum AFP level is known. This prognostic model has also been well validated in a separate population of patients in the Asia-Pacific region.
We thank Drs W.C. Chow, C.K. Yap, and K.Y. Ng from the Department of Gastroenterology of the Singapore General Hospital for allowing their patients to be enrolled onto the study. We also thank Drs M.W. Khin (Yangoon General Hospital, Myanmar), P. Johnson (Chinese University of Hong Kong, Hong Kong SAR), K.C. Soo (National Cancer Centre, Singapore), P. Chow (Singapore General Hospital, Singapore), K. Thiravud (National Cancer Institute, Thailand), W.M. Tjakra (University of Udayana, Bali, Indonesia), A. Haron (Universiti Kebangsaan Malaysia, Malaysia), S.H. Bae (Catholic University of Korea, South Korea), M. Findlay (Wellington Regional Oncology Unit, New Zealand), and J. Cebon (Ludwig Institute for Cancer Research, Sydney, Australia) for allowing their patients to be used to validate our prognostic model. We also thank B.C. Tai of Clinical Trials and Epidemiology Research Unit, Singapore, for her help with data from the AHCC01 trial.
1. International Agency for Research in Cancer: Globocan 2000: Cancer incidence, mortality and prevalence worldwide. http://www-dep.iarc.fr/globocan/globocan.html 2. Lai CL, Lam KC, Wong KP, et al: Clinical features of hepatocellular carcinoma: Review of 211 patients in Hong Kong. Cancer 47:2746–2755, 1981[CrossRef][Medline] 3. Lim TC, Teh LB, Kwok KC, et al: Hepatocellular carcinoma: A clinical study. Ann Acad Med Singapore 15:158–161, 1986[Medline] 4. Chia KS, Lee HP, Seow A, et al: Trends in cancer incidence in Singapore 1968–92. Singapore, Singapore Cancer Registry, Singapore Cancer Registry Report No. 4, 1996 5. Calvet X, Bruix J, Gines P, et al: Prognostic factors of hepatocellular carcinoma in the West: A multivariate analysis in 206 patients. Hepatology 12:753–760, 1990[Medline] 6. Chevret S, Trinchet JC, Mathieu D, et al: A new prognostic classification for predicting survival in patients with hepatocellular carcinoma. J Hepatol 31:133–141, 1999[CrossRef][Medline] 7. The Cancer of the Liver Italian Program (CLIP) Investigators: A new prognostic system for hepatocellular carcinoma: A retrospective study of 435 patients. Hepatology 28:751–755, 1998[CrossRef][Medline] 8. Kim UB, Doo CJ, Baek SH, et al: Natural history and prognostic factors of primary hepatocellular carcinoma: Study of 70 untreated cases. Korean J Intern Med 4:136–141, 1989[Medline] 9. Leung TWT, Tang AMY, Zee B, et al: Construction of the Chinese University Prognostic Index for hepatocellular carcinoma and comparison with the TNM staging system, the Okuda staging system, and the Cancer of the Liver Italian Program staging system. Cancer 94:1760–1769, 2002[CrossRef][Medline] 10. Chow PKH, Tai BC, Tan CK, et al: High-dose tamoxifen in the treatment of inoperable hepatocellular carcinoma: A multicentre randomised controlled trial. Hepatology 36:1221–1226, 2002[CrossRef][Medline] 11. Parmar MKB, Machin D: Survival Analysis: A Practical Approach. Chichester, NY, Wiley, 1995 12. Altman DG, Royston P: What do we mean by validating a prognostic model? Stat Med 19:453–473, 2000[CrossRef][Medline] 13. Stanley KE: Prognostic factors for survival in patients with inoperable lung cancer. J Natl Cancer Inst 65:25–32, 1980[Medline] 14. Akashi Y, Koreeda C, Enomoto S, et al: Prognosis of unresectable hepatocellular carcinoma: An evaluation based on multivariate analysis of 90 cases. Hepatology 14:262–268, 1991[CrossRef][Medline] 15. Zhao G, Su S, Borek D, et al: Long survival and prognostic factors in hepatocellular carcinoma. J Surg Oncol 45:257–260, 1990[Medline] 16. Izumi R, Shimizu K, Kiriyama M, et al: Alpha-fetoprotein production by hepatocellular carcinoma is prognostic of poor patient survival. J Surg Oncol 49:151–155, 1992[Medline] 17. Williams R, Melia WM, Johnson PJ: Serum alpha-fetoprotein in hepatocellular carcinoma: Value in diagnosis and prognosis. Ann Acad Med Singapore 9:245–250, 1980[Medline] Submitted March 24, 2001; accepted March 29, 2003.
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2003 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
36, which is many more than the minimum of 10 (or the safer 20) which has been suggested.
