Statistics from Altmetric.com
- hepatocellular carcinoma
- prognostic score
- CLIP score
- BCLC staging system
- Okuda staging
- staging systems
- liver cirrhosis
Are prognostic models for hepatocellular carcinoma useful in predicting survival?
Over the past decade, non-surgical therapies for hepatocellular carcinoma (HCC) occurring on a background of cirrhosis have been established as effective. Hepatic arterial chemoembolisation has a proven survival advantage in selected patient groups,1,2 and there is now strong evidence that ablative therapies, both percutaneous alcohol injection3 and radiofrequency ablation,4 have survival benefit. In addition, HCC incidence is increasing in the Western world, mainly due to the hepatitis C epidemic and this cancer now has an increasing impact on health services.
A further change in the clinical setting in which we encounter HCC has occurred. The advent of ablative therapy for HCC, which can destroy tumour nodules up to approximately 5 cm in diameter, has focused the need for detection of tumours at an early stage in cirrhotic patients and has led to the widespread use of screening in most of the world. There is no doubt that screening using ultrasound and α fetoprotein will detect HCC at a smaller size.5 This has produced the need for new prognostic models given that the initial Okuda system6 was developed some 30 years ago when the size of HCCs at presentation was substantially larger than today.
Thus prognostic models are important from both a clinical perspective, of being able to give patients and their relatives accurate information on survival, and from a research perspective, to assess outcome of new therapies to help define prognosis as accurately as possible. A number of prognostic models for HCC have been developed. All models have been established in patient cohorts and validated in other cohorts. All available models have established two key factors influencing outcome: how good is the underlying liver function and how extensive is the tumour.
In this issue of Gut, two articles address this issue: one comparing current prognostic models—Cancer of the Liver in Italy (CLIP),7 Barcelona Clinic Liver Cancer (BCLC),8 and the Okuda system6—in a single Italian centre,9 and the other by Tateishi and colleagues10 assessing a new model developed in a large cohort of Japanese patients (see page 411 and page 419). It should be emphasised that these are not natural history based models; almost all patients had therapy and some patients had more than one treatment modality. This form of modelling of survival therefore reflects the outcome likely to be seen in a real clinic setting but is subject to the impact of many treatment related effects on survival which may not be reproducible in standard clinical practice. It is also important to recognise that these studies were aimed at prognosis in patients with “early” HCC, essentially presymptomatic patients detected primarily by screening, and therefore represent a subgroup of the patient population.
The comparative study, in which 87% of patients had screening detected HCC,9 showed that the long established Okuda system is no longer useful in patients with relatively early HCC; not surprising given that Okuda I (early) is defined as less than 50% of liver volume replaced by tumour, which can still include massive tumours which would be far too large to treat with ablative therapy. The other systems performed well, with the best predictor being the BCLC system. This has the disadvantage of complexity; if strictly followed it requires measurement of portal pressure whereas the CLIP system is simple, uses clinical parameters, and can be used at the bedside. The accuracy of the CLIP model is interesting as many of the same parameters as Okuda (Child-Pugh score and tumour extent) are used, yet simple changes such as the classification of tumours as uni or multifocal dramatically improves its predictive value over Okuda. The Barcelona Clinic Model (BCLC) presents features in more detail with more categories, and uniquely includes performance status, a factor which has been key in predicting response to treatment in most other oncology settings. In large clinical trials of therapy for HCC, the BCLC system would seem to have significant advantages in the level of detail of stratification which is possible. From a clinical perspective, the CLIP score will give patients and clinicians an almost equally reliable guide as to outlook.
The Japanese model was developed entirely in patients treated by ablative therapy, which meant that 93% of patients had tumours smaller than 5 cm in diameter.10 Again, the analysis produced a simple clinical algorithm based on parameters known to reflect liver function (bilirubin and albumin) and tumour extent (size and number). In this system, the ability of these simple parameters to predict survival was high. The overall five year survival rates varied from almost 80% in patients with good liver function and small tumours to 15% in those with poor liver and large or multiple tumours. This model will require further validation; perhaps its major drawback was its validation in a cohort undergoing another form of therapy (surgical resection). However, it was interesting that it still predicted survival in a group being treated with an alternative strategy, suggesting that the parameters used will hold true for most situations where early HCC is being treated. The simplicity of this score is attractive; it only takes into account tumour size and number, albumin, and bilirubin, parameters which will be available for all patients early on in the clinical assessment. It differs from the CLIP score in not using α fetoprotein and portal vein invasion and excludes the other components of the Child Pugh score. Further validation of this system is required to see if it performs as well as CLIP and BCLC in other parts of the world, where the aetiology of liver disease may be different, and with treatment modalities such as arterial chemoembolisation.
The overriding importance of tumour size and liver function comes across in all of these scoring systems. The real difference between them is in which components of liver function to include and if tumour size and number is enough or if other factors of tumour extent such as portal invasion are better markers. One area which has not been examined specifically in the models is the cause of death. In patients with advanced liver disease the tumour may be incidental and one would expect that most would die of liver failure. In patients with good liver function the outlook may be much more determined by tumour extension. In this latter group, second tumour development and the potential to treat the underlying liver disease are also likely to affect outcome; there is compelling evidence that therapy for hepatitis C will reduce cancer incidence.11
Perhaps the most important point from both the comparative studies and this new algorithm is that in using a combination of ablative therapy and chemoembolisation survival rates are very high. The best comparator studies were two natural history series in untreated patients with small HCC and good liver function at presentation which showed an approximate three year survival rate of 25%.12,13 The three year survival in the Grieco study was 40% with a comparable aetiology of liver disease. This is in contrast with overall survival values which remain poor worldwide, with only 5–8% five year survival rates in the USA and Europe. These values emphasise that good results with a change in outcome can be achieved but only if tumours are diagnosed at an early stage on a background of good liver function. The treatment option in patients with decompensated liver disease is limited to liver transplantation and there is no possible therapy for those with both a poorly functioning liver and a large HCC because of the high tumour recurrence rates post transplantation. In most parts of the world, the typical patient with HCC still presents with advanced disease which explains the poor overall survival figures. Can we improve the situation and detect more HCC at a treatable stage? Screening of high risk cirrhotic patients has gained wide acceptance and there is clear evidence that such strategies can detect cancers of a smaller size5,14; this can translate into more treatable HCC with survival rates as high as seen in these two cohorts.15 What has not been established is if this strategy can alter overall survival, and if the costs, both financially and psychologically, are worthwhile. The main problems with screening are that a significant minority of HCC patients present with their tumour at the same time as the present with their underlying liver disease, and that the screening tests (α fetoprotein and ultrasound) are either relatively insensitive for small HCC or very operator dependant. The risk of HCC varies according to sex and the cause of cirrhosis, adding further complexity to the debate. There is a need for studies which better identify patients at risk and examine new modalities of screening, such as magnetic resonance imaging. However, there is no doubt that if small HCC are detected they can be treated. We also have prognostic models which allow both clinical prediction of outcome and clinical trials of therapy to be undertaken.
Are prognostic models for hepatocellular carcinoma useful in predicting survival?
Conflict of interest: None declared.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.