Treatment
of hepatitis B virus infection with interferon Factors predicting response to interferon
Several ranoomised controlled trials have been undertaka chronic hepatitis B. In patients with HBe antigen-positive 25-50%.
In those infected at birth, response
to evaluate the efficacy of a-intexferoq in the therapy of disease acquired in adult life the wsponse rates vary from
rates are lower. Twenty-one
pretreatment
variables
were assessed for their
significance in response prediction using data from 114 patients given o-interferon for chronic hepatitis B virus infection. In those patients who had received a minimum of 90 million units per m2 total dose over 12 weeks, a negative anti-human immunodeficiency virus antibody stattts fp < O.OOl), chronic active hepatitis on liver biopsy @ < 0.005). high AST level (p < O.oOl), low hepatitis B virus DNA level @ < O.OiJl) and a history of acute hepatitis (p 4 0.005) were all associated with an increased likelihood of response on uniwiate analysis. On stepwise logistic regression analysis, hepatitis B virus DNA, AST and a history of acute hepatitis predicted response independently @ < 0.05). The most reliable combination of predictive factors was a negative anti-human immunodeficiency virus antibody status, with either a positive history of acote ictefic hepatitis and AST > 45 IIJ per liter or no history of acute icteric hepatitis and AST > 85 HJ per liter, which predicted response in 77% with a specificity of 79% @ < 0.001). The loss of HBsAg in addition to HBeAg and hepatitis B virus DNA was more likely to occur in patients with chronic infection of %I million units per m’ a-inter-
and south of France as Medi-
feron; 43 of these patients had responded to therapy with
symptoms were considered
chronic malaise, anorexia and right hypochondrial
to he pain.
a sustained loss of HBeAg these had additionally not signtficantly unwariate
and HBV
lost HBsAg.
DNA,
associated with treatment
mfection.
response on
analysis are listed in Table 2. However,
tive history of acote icteric hepatitis (AH) hepetitisw.
or minimat disease).
and negative an&HIV
AST
a posi-
at the onset of
presence of more severe inflammation
biopsy (chromcanive &is
and 13 of
Factors which were
on liver
chronic perswen! level, HBV
DNA
hep level
antibody ~tatos were signihcanrly
associated with response (Table 3).
S6 Of patienls with a past history of AH. those who lost HBsAg m addition to HBeAg and HBV DNA were sigmficantly more likely to give a history of AH within the 24 months poor to therapy wheo compared with responders who did not loose HBsAg (10113 “s. 019. p < 0.001).
As AH, 4ST. HBV DNA, were all significantly sssociated
anti-HIV and histology with response on univa-
DiSCiWiOll This analysis shows that it is possible to predict which patients with adult acquired HBV infection will respond to monorherapy with o-interferon. The only requirements are a knowledge of the patient’s anti-HIV status, clinical history of acute icteric hepatitis and P single pretreatment
riate anatysia, they were combmed in stepwise logistic regression analysis. The combination of anti-HIV, AST and AH proved to be the most useful variable for predicting response. Histology and HBV DNA added no further improvement in discrimination. Response could be predicted with a sensitivity (correctly predicted responders/all responden) of 77% (33143) and specificity (correctly predicted nonresponderslall nonresmnders) of 79% I56i71) (I < 0.0011 from a nee?:we anti-HIVgntihody &tos an‘d eitheranAST > 85 Iv per liter with a negative history of AH, or an AS7 > 45 IU per liter with a positive history of AH. One hundred per-
measurement of AST. However. this simple model could well he idiosyncratic to the patients investigated, and it is important to test it in other study populations. The predictive factors indentified in this study can be divided into two separate groups. Hwological activity, aminotransfcrase level and a history of AH are each indirect indicators of the level of immune activity directed against the virus (13-16). Similarly, anti-HIV positive patients frequently show immune defects (9.10.17). HBV DNA, however, is a measure of virus levels. It is therefore surprising !hat the use of HEiV DNA levels adds no further power to AST, AH and anti-HIV states in cur predictive model. Utilizing such easily acquired data, the physician may optimize the patient’s chance of response
cent response could be expected in anti-HIV-negative patients with a AST > 330 IU per liter, AH = 0 or AST > I65 IU per liter. AH = I, hot with a sensitivity of only 15%. HBV DNA, AST and a history of acute hepatitis were
by commencing therapy when response is likely, or alternatively consider other therapeutic approaches (11,18) if nonresponse is predicted. The variables known to he esso&ted with early spontaneous serownversion are similar to those that are predictive of treatment response (l-6,9,
independent predictive variables f_n 4 0.05). Although anti-HIV states (p = O.OlS) also correlated with response, this was not significant at the 5% level. When this model was app!ied separately to the groups of patients treated with different a-interferons, there was no statislical difference in success of prediction betwwn treatment types.
10,18-23).
Despite
this, treatment
is still necessary
in
such patients for several reasons. In all trials of a-interferon conducted at this centre, the annual rate of serocc-+ version in randomized untreated control patients has been less than one-fifth of the rate reported in treated patients. This early treatment induced seroconversion may be expected to he beneficial by preventing the significant hepatic fibrosis or cirrhwir iint is found on liver biopsy of some patients before they undergo natural seroconversion (24-26). Treatment can also he justified by the observation, supported by this study, that response occurring within 2 years of vinls acquisition leads sdditionally to the loss of HBsAg (3,6) which may reduce the patient’s risk of developing hepatocellular carcinoma (27.28). Although this model increases the sensitivity of TLsponse prediclion to 77% in the patients studied, there is still room for improvement. This may emanate from other more accurate measurements of the pretreatment immune xyr,-me to HBV. Alternatively, them may be influences separate from those studied here. such as development of anti-interferon antibodies during therapy (29). Until such improvements occur, the predictive model described may be useful to physicians using o-interferon when it is given a product license for the treatment of chronic HBV infection.