|
- 现金
- 62111 元
- 精华
- 26
- 帖子
- 30437
- 注册时间
- 2009-10-5
- 最后登录
- 2022-12-28
|
1832
Non-invasive prediction of liver disease in a multi-genotype
cohort of hepatitis B e antigen negative patients
Gayatri Chakrabarty1, Ivana Carey2, Kosh Agarwal2, Daniel M.
Forton1; 1Gastroenterology and Hepatology, St Georges University
of London, London, United Kingdom; 2Liver Unit, Kings College
Hospital, Denmark Hill, London, United Kingdom
Introduction: HBeAg-ve patients with liver disease require
early intervention, yet can be difficult to identify. Liver stiffness
measurement (LSM) can stage liver fibrosis but is not always
available. Fibrosis prediction models using baseline laboratory
parameters exist in HCV infection. Models in HBV have
generally dealt with eAg+ve patients. The majority of patients
starting antivirals in the UK are eAg-ve. Aim To identify factors
that predict liver disease in eAg-ve HBV infection; to establish a
model to predict liver disease at baseline assessment. Methods
270 HBeAg-ve patients (genotypes A to E) from 2 urban teaching
hospitals underwent fibroscan or liver biopsy. No liver
disease (NLD) was defined as LSM< 6kPa or liver biopsy Ishak
F<2 and/ or HAI4 < 4; liver disease (LD) was LSM ≥ 6kPa or
F≥2 and/or HAI ≥ 4 or radiological evidence of cirrhosis. Basic
demographic and lab parameters were recorded and qHBsAg
and IP 10 levels were measured. The patients were randomly
allocated to training (80%) and validation (20%) cohorts, using
SPSS. Univariate analyses were performed between LD and
NLD in the training set. Significant variables (p<0.02) were
then subjected to multivariate stepwise binary logistic regression.
ROC curves were generated for models, which were
tested in the validation cohort. Results Gender, platelet count,
baseline ALT, HBV DNA log10, qHBsAg log10, HBV genotype
and IP10 level were significantly different between LD and NLD
on univariate analyses. All the variables identified in the univariate
analysis were used in the regression analysis. The model
incorporating ALT, HBV DNA, IP 10 and platelet count gave
the best AUROC at 0.79. (PP = Exp (-1.54010 + 0.00340 x IP
10 + 0.01976 x ALT + 0.44801 x HBV DNAlog10 - 0.00909
x Platelet count) / 1+ Exp (-1.54010 + 0.00340 x IP 10 +
0.01976 x ALT + 0.44801 x HBV DNAlog10 - 0.00909 x
Platelet count). The AUC in the small validation cohort was
0.67. A PP cut-off value of >0.20 had a sensitivity of 92% for
LD. A higher cut-off value of <0.60 identified 92% (120/130)
of NLD. Using these 2 cut-offs, 53% of patients could be categorised
with a NPV of 89% and PPV of 81% for LD. Discussion
In a heterogeneous cohort of HBeAg-ve, multi-ethnic, mixed
genotype, treatment naive patients from 2 hospitals, a model
incorporating baseline ALT, HBV DNA, IP 10 and platelets
can usefully identify liver disease (AUROC 0.79). It compares
well with a published model for significant fibrosis in mainly
eAg+ve patients (AUROC 0.80) (Hui et al 2005). Notably
qHBsAg levels were not independently associated with liver
disease. Further work is required to validate in larger cohorts
and to test as a prognostic model over time.
Disclosures:
Ivana Carey - Grant/Research Support: Gilead, Roche; Speaking and Teaching:
BMS
Kosh Agarwal - Advisory Committees or Review Panels: Gilead, BMS, Novartis,
Janssen, AbbVie; Consulting: MSD, Janssen, Achillion, Intercept; Grant/Research
Support: Roche, Gilead, BMS, Arbutus; Speaking and Teaching: Astellas, Gilead,
BMS, GSK
Daniel M. Forton - Advisory Committees or Review Panels: Gilead, Merck, Abbvie;
Speaking and Teaching: Janssen, BMS
The following people have nothing to disclose: Gayatri Chakrabarty
|
|
发表于 2016-10-25 16:52