 |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Current Issue | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
LECTURE James
Arthur DICKINSON1 and Yuk Tsan WUN2 Abstract
Introduction
How should doctors respond? Many
hepatologists and oncologists recommend screening with ultrasound,
alpha-fetoprotein (AFP) or both, at varying intervals based on their
experience of seeing many cases of cirrhosis and liver cancer. Most
hospital-based specialty clinic studies include a population with
more severe diseases, often with cirrhosis or even symptomatic liver
disease; thus, it is not surprising that such patients have high rates
of complications or death. We cannot extrapolate their results to
the type of patients we see in primary care clinics. Therefore, we
must ask what should primary care doctors advise their patients? We
searched the literature to find the answers to these important clinical
questions. We
conclude there are insufficient quality studies on this topic to provide
adequate information on age and sex groups and for the possibility that
risks are different according to racial groups or social development. As
a consequence of the limited data, we developed a mathematical model to
predict risk using Hong Kong data. In this model we presume that nearly
all of the liver cancer and cirrhosis occurs among the proportion of the
population who are hepatitis B carriers. Full details of this work are
described in our paper12
but in brief, we found that among male carriers, the risk of developing
liver cancer (hepatocellular carcinoma [HCC]) is so high that between
the ages of 35 and 55 mortality from HCC among male carriers is higher
than for all other causes of death combined. For women, mortality rates
of carriers are raised but the additional risk never exceeds the normal
mortality rate. When we add the mortality rates for cirrhosis, liver cancer
and deaths from other causes to estimate total mortality among male carriers,
it starts to rise noticeably from about the age of 35 to be twice the
total mortality for non-carriers, while for women the difference is always
quite small (Fig.
2). We then used a life-table approach that takes a hypothetical
newborn population of 100 000 people and assumes they would have the current
death rates from all causes over their whole lifetime. This produces curves
that show that survival for male carriers begins to drop noticeably from
the age of 35-40, and leads to about 10% mortality by age 50 and 20% by
age 60. Their average life expectancy is about 72 years, whereas non-carriers
in Hong Kong would live 7 years longer, approximately 79 years. For female
carriers the survival rate is always better than for male non-carriers,
beginning to drop from about the age of 60 so that mortality reaches 10%
by age 70. Half the female carriers live to around 80, whereas female
non-carriers on average live to approximately 82 (Fig.
2). Thus
we have shown that the risk is high, especially for men, probably at about
the same risk as cigarette smokers, half of whom will die in middle age
as a result of smoking.13 What
can we do for these patients? Treatment or screening? Two
treatments are now available for hepatitis B: interferon and lamivudine.
Interferon is effective in reducing liver inflammation but has substantial
side-effects for a considerable period. A much better tolerated drug with
very few side-effects is lamivudine, which reduces hepatic inflammation
and the changes of cirrhosis. If continued for more than a year, a substantial
rate of sero-conversion to e-antigen-negative does occur but over time
there is an increasing number of drug-resistant mutations which escape
suppression, so that liver damage continues.15
Both of these drugs are expensive, especially for people in less developed
countries and since these approaches are still new, the long-term outcome
for reduction of liver cancer is still unclear. Many hepatologists recommend
chronic long-term lamivudine, but at the moment this is hopeful rather
than based on data, since this drug has not been available long enough
to produce the necessary supportive evidence. Clearly
it would be desirable to prevent liver cancer from occurring, but if we
cannot, is screening effective? Potential screening tools are AFP and/or
ultrasound. A variety of recommendations are made on who to screen, how
often to screen and effectiveness of these programs. Therefore, we undertook
a Cochrane systematic review of the value of screening people with chronic
hepatitis B infection for liver cancer.16
One trial in China directly addressed the question.17
These researchers detected many more tumors in the screened group and
these cancer patients had a longer survival rate post-diagnosis than the
patients who presented clinically in the unscreened group.17
This trial is sometimes quoted as proving the value of screening, but
when we reanalyzed to compare the overall liver cancer death rates, they
were not significantly different.16
Thus they produced one of the classic biases of screening: the lead-time
bias.18
One other randomised controlled trial fulfilled the quality criteria,
but compared AFP against AFP and ultrasound19
without comparison to controls. One trial in Alaska compared AFP against
an historical control group without screening.20
The data are thus very limited and not encouraging. While
these studies detected liver cancer at the preclinical stage, and also
detected a number of small liver cancers at a stage when surgical resection
could potentially make a difference, the effort and cost of screening
is high. The Chinese trial required 1006 AFP, 329 ultrasounds or 403 AFP
plus ultrasounds to detect one liver cancer smaller than 5 cm diameter.
The researchers' optimistic estimate of delaying 10 deaths would require
screening 937 patients annually for 5 years. Sherman's group performed
nearly 800 AFP tests to find one small liver cancer19
and in Alaska McMahon's group undertook 800 AFP to find one liver cancer
or 2000 to find one small liver cancer.20
Across 4159 patients in this study, sensitivity of AFP appears to be about
72.5%, and specificity 93.7%. Given the model described above12 and assuming
the prevalence of liver cancer is twice the calculated incidence, the
positive predictive value would be approximately 6% in men aged 40-49,
12% in those aged 50-59, and 24% in those aged 60-69. Further investigation
is then needed to demonstrate that most patients with positive AFP tests
do not have a liver cancer. Thus
we conclude that while AFP conducted annually or twice yearly can detect
more small and even operable liver cancers with a longer survival rate
than those unscreened, this is likely due to lead-time bias. There is
no trial evidence that screening postpones or decreases the death rates
of carriers or improves their quality of life. If
empiric evidence for a screening test is not available, we use a set of
standard criteria for screening tests. When we compare liver cancer against
these (Table 2) we conclude
that screening fails. While this is an important disease, this cancer
grows relatively rapidly and screening tests only work effectively for
chronic, slowly developing disease. At the moment we are unclear on the
value of the two potential screening tests, and even worse, it is not
yet clear that early treatment for this cancer is truly effective, especially
as these cancers tend to be multifocal, rising at many points in a damaged
liver, so that many patients are inoperable because of severe underlying
liver disease, while there is a high operative mortality and postoperative
recurrence rate, even for those previously operated upon. Liver transplant
might be a potential solution, but in many countries where this disease
is a major problem, donation of organs is not socially acceptable, while
the operation is complex and extremely costly. Even where it is performed,
there will never be sufficient donor livers available. Performing
a screening program would cause substantial harm because of the high rate
of false positives, with the resultant anxiety, worry and high cost for
these people. We must recognize the aphorism of Muir Gray: 'All screening
programs do harm; some can do good as well. The harm from a screening
program starts immediately; the good takes longer to appear. Therefore
the first effect of any program, even an effective one, is to impair the
health of the population.'20 Conclusion
The
Malaysian Consensus Committee on screening for hepatocellular cancer21
agrees with our assessment that there is no evidence of value and there
are poor data for improving survival, probably due to length bias, but
they still recommended screening. It is common for authoritative committees
to make recommendations that do not follow from the evidence, since all
of us at the emotional level want to do something for patients facing
this unpleasant disease. However, our assessment of the evidence is that
screening chronic hepatitis B patients for liver cancer would provide
minimal benefit, while being burdensome and costly, so it should not be
offered or recommended References
Table 1 Incidence rate or mortality rates of hepatocellular cancer among hepatitis B carriers by age and sex, obtained from population cohort studies
--------------------------------------------------------------------------------------------------------------------------------------------------------------- Table 2 Matching screening chronic hepatitis B paitients for liver cancer against criteria for screening
--------------------------------------------------------------------------------------------------------------------------------------------------------------- |
One trial, reported
sketchily; outcome: no difference16