Clinical characteristics of patients with hepatitis and cirrhosis and the construction of a prediction model

Abstract

BACKGROUND

Hepatitis B-associated cirrhosis is an important disease burden in China. However, there is a lack of effective predictors in clinical practice to drive delivery and enable early treatment to delay disease progression.

AIM

To analyzing the clinical characteristics of patients with hepatitis and cirrhosis, the nomogram model was established and validated.

METHODS

The clinical data of 1070 patients with hepatitis B who were treated in our hospital from October 2015 to July 2022 were collected. In a 7:3 ratio, 749 cases were divided into training cohorts and 321 cases were divided into validation cohorts. In addition, the training cohort and validation cohort were further divided into hepatitis group and hepatitis B-related cirrhosis group based on whether the patient progressed to cirrhosis. Binary logistic regression was used to analyze the influencing factors of hepatitis progression to cirrhosis. A roadmap prediction model was established, and the predictive effect of the model was evaluated by patient-subject receiver operating characteristic curve (ROC), and the effectiveness of the model was evaluated by decision curve analysis.

RESULTS

Binary logistic regression analysis was performed using hepatitis B-related cirrhosis = 1 and hepatitis = 0 as dependent variables, and univariate analysis of serological indicators was used as covariates. The results showed that glutamic oxaloacetate aminotransferase/glutamate acetone aminotransferase levels, prothrombin time activity, and hepatitis B e antigen levels were all contributing factors to the progression of hepatitis to cirrhosis. The area under the ROC curve was 0.693 [95% confidence interval (CI): 0.631 to 0.756] for the training cohort and 0.675 (95%CI: 0.561 to 0.790) for the validation cohort. In addition, the decision analysis curves of the prediction models of both the training cohort and the validation cohort confirmed the effectiveness of the nomogram prediction model.

CONCLUSION

Three independent factors influencing the progression to cirrhosis in patients with hepatitis B were identified. The construction of a nomogram prediction model from hepatitis to cirrhosis has high application value as a tool for predicting the occurrence of liver cirrhosis in hepatitis B patients.

Key Words: Hepatitis; Hepatitis B-related cirrhosis; Clinical features; Influencing factors; Nomogram

Core Tip: In order to analyze the clinical characteristics of patients with hepatitis and cirrhosis, a nomogram model was established and verified. The clinical records of 1070 patients with hepatitis B who were treated in our hospital were selected for study, and they were divided into 749 cases in the modeling cohort and 321 cases in the model validation cohort in a 7:3 ratio. At the same time, the model validation cohort was divided into hepatitis group (n = 688) and hepatitis cirrhosis group (n = 61), and the model validation cohort was divided into hepatitis group (n = 295) and hepatitis cirrhosis group (n = 26) according to whether the patients had liver cirrhosis. The levels of albumin hepatitis B E antigen, white blood cell, red blood cell, hemoglobin, absolute neutrophil count and absolute lymphocyte count in patients with hepatitis cirrhosis were significantly lower in the modeling cohort than those in patients with hepatitis, while the levels of the ratio of aspartate aminotransferase, plasma prothrombin time, hyaluronic acid, laminin and platelet count were significantly higher than those in patients with hepatitis (all P < 0.05). The area under the receiver operating characteristic curve (AUC) of the nomogram prediction model established by cohort modeling was 0.693 [95% confidence interval (CI): 0.631 to 0.756], and AUC of the nomogram prediction model was 0.675 (95%CI: 0.561 to 0.790). The actual result curves of the nomograms generated by the modeling and validation cohorts deviate from the calibration curves with little tolerance.

INTRODUCTION

The incidence of hepatitis B virus (HBV) infection is relatively high in China, and the mortality rate has obviously increased in recent years[1]. At the same time, as a virus which is difficult to be completely cured, HBV can invade human liver tissue and cause hepatitis. In the early stage of hepatitis B, patients only have mild symptoms, even no discomfort, so the onset is hidden and the symptoms are atypical, which leads to the continuous deterioration of liver function, the patient’s condition is gradually aggravated, and even progress to cirrhosis. Patients with hepatitis B cirrhosis are one of the important causes of clinical primary liver cancer and hepatic encephalopathy[2]. Therefore, timely identification and early intervention in liver cirrhosis can not only promote the prognosis of patients with liver cirrhosis due to hepatitis B to a certain extent, but also improve the efficiency of clinical diagnosis.

Liver biopsy, as the most objective, true and reliable diagnostic criterion for detecting liver function, is not only traumatic, but also technically difficult. In addition, patients’ acceptance of it is low, so its application is limited. With the progress of science and technology, the application of technology in the medical field can deal with complex and huge data and produce more accurate prediction models. However, because of the complexity of the models, it is difficult for clinicians to explain it to patients, which increases the difficulty of practice. In order to be concise and practical, the model is modified to a simpler scoring system in clinic, but there are still some deviations.

It is of great clinical significance to find new markers and predictive models that can safely and effectively detect the liver function of patients with hepatitis B cirrhosis. As a graphical tool, the nomogram can not only accelerate more complicated calculations, but also present the influence of various prediction factors on the results through graphical tools, so that we can better understand and use[3]. In addition, this modeling method is not only easy to operate, but also safer. Besides, it can be observed repeatedly and accepted by patients. Based on this, 1070 cases of hepatitis patients were included in this study. By collecting their general clinical data and routine laboratory indicators, a corresponding nomogram prediction model of hepatitis complicated with liver cirrhosis was established, aiming at providing reference for early clinical detection and treatment of liver cirrhosis.

MATERIALS AND METHODS

Subjects

A retrospective study was conducted on the clinical medical records of 1070 hepatitis B patients who were treated in our hospital from October 2015 to July 2022. They were divided into a model cohort of 749 cases and a model validation cohort of 321 cases according to the ratio of 7:3. The model validation cohort was divided into hepatitis group (n = 21) and hepatitis cirrhosis group (n = 45). The study was approved by the Hospital Ethics Committee. Inclusion criteria: (1) All patients met the diagnostic criteria of hepatitis B in the Guidelines for the prevention and treatment of chronic hepatitis B[4], and the patients with cirrhosis of liver met the criteria of hepatitis and cirrhosis in the guidelines for the prevention and treatment of chronic hepatitis B[4]; (2) The patient has reached the age of 18; and (3) The clinical data of patients are complete. Exclusion criteria: (1) Patients with combined liver cancer; (2) Patients with severe cardiac system diseases; (3) Liver cirrhosis caused by other reasons; and (4) Pregnant or lactating patients.

Observation indicators

Clinical data of patients in the hepatitis group and the liver cirrhosis group were collected through electronic medical records in our hospital. General data of the two groups were compared, including gender, age, hepatitis classification, allergy history, disease history, operation history, smoking history, alcohol consumption history, family disease history, and treatment plan. Meanwhile, laboratory indexes of two groups of patients are counted, albumin (Alb), lactate dehydrogenase (LDH), total bilirubin, total bile acid, prealbumin, cholinesterase (CHE), urea nitrogen, total cholesterol, high density lipoprotein cholesterol, low density lipoprotein cholesterol, triglyceride, apolipoprotein A, apolipoprotein B, creatine kinase, leucine aminopeptidase, direct bilirubin, glutamyl transpeptidase, the ratio of aspartate aminotransferase and alanine aminotransferase (GOT/GPT), international normalized ratio (INR), plasma prothrombin time (PT), prothrombin activity (PTA), hepatitis B surface antigen, hepatitis B surface antibody, hepatitis B E antigen (HbeAg), hepatitis B E antibody, hepatitis B core antibody IgG, hyaluronic acid (HA), laminin (LN), platelet count (PLT), white blood cell (WBC), red blood cell (RBC), hemoglobin (HGB), absolute neutrophil count (ANC), absolute lymphocyte count (ALC) etc.

To analyze the influencing factors of patients with hepatitis cirrhosis, and establish and verify the road map prediction model.

Statistical analysis

Statistical product and service solutions 25. 0 software and R software were used to analyze the collected data. The collected measurement data were subjected to Shapiro-Wilk normal distribution test, where P > 0.05 meant the normal distribution data were expressed as mean ± SD, and t test, where P < 0.05 meant the non-normal distribution data were described as the median (quartile), followed by Mann-Whitney U test. Enumeration data collected were expressed as % and disordered data were analyzed using χ² or Fisher exact test, and ordered data were analyzed using Mann-Whitney U test. Univariate and multivariate logistic regression analysis was used to analyze the influencing factors of hepatitis cirrhosis in patients, and the nomogram prediction model was established. The discriminant ability of the verification set and the calibration map were used to evaluate the accuracy of the nomogram. The area under the receiver operating characteristic (ROC) curve was used to evaluate the resolution of nomograms. The correction curve of the model was calculated, and the consistency of the model was verified by Hosmer-Lemelshaw test. Decision curve analysis was also performed to evaluate the model’s discriminating ability. P < 0.05 indicated that the difference was statistically significant.

RESULTS

Baseline clinical features

A total of 1070 patients with hepatitis B were included, including 749 cases in the modeling cohort and 321 cases in the verification cohort. The average age of the patients was (46.84, 11.73) years old. There were 679 males (63.46%) and 391 females (36.54%). In the modeling cohort, 487 cases (65.02%) were male and 262 cases (34.98%) were female. In the validation cohort, there were 192 males (59.81%) and 129 females (40.19%). There was no significant difference between the two groups in baseline data (P > 0.05) as shown in Table 1 and Supplementary Table 1.

Table 1 Comparison of baseline data between modeling and verification cohorts, mean ± SD/n (%).

Index		Total cases (n = 1070)	Modeling queue (n = 749)	Verification queue (n = 321)	t/χ2/Z	P value
Gender	Man	679 (63.46)	487 (65.02)	192 (59.81)	2.627	0.105
	Woman	391 (36.54)	262 (34.98)	129 (40.19)
Age (years)		46.84 ± 11.73	46.58 ± 11.77	47.45 ± 11.63	1.112	0.266
Hepatitis classification	Mild	242 (22.62)	161 (21.50)	81 (25.23)	0.849	0.396
	Moderate	753 (70.37)	538 (71.83)	215 (66.98)
	Serious	75 (7.01)	50 (6.68)	25 (7.79)
Alb (× 109/L)		43.11 ± 4.70	43.02 ± 4.65	43.32 ± 4.79	0.958	0.338
LDH (U/L)		210.73 ± 93.70	210.07 ± 73.36	212.28 ± 129.41	0.353	0.724
CHE (U/L)		7376.59 ± 2247.89	7298.93 ± 2190.67	7557.79 ± 2369.69	1.728	0.084
WBC (× 109/L)		5.19 ± 1.53	5.19 ± 1.51	5.20 ± 1.59	0.098	0.922
RBC (× 1012/L)		4.74 ± 0.50	4.73 ± 0.51	4.74 ± 0.49	0.297	0.766
HGB (g/L)		145.45 ± 16.13	145.56 ± 15.96	145.22 ± 16.54	0.316	0.752
ANC (× 109/L)		2.84 ± 16.13	2.81 ± 1.07	2.89 ± 1.25	1.064	0.288
ALC (× 109/L)		1.86 ± 0.65	1.88 ± 0.66	1.84 ± 0.64	0.917	0.360
GOT/GPT		0.76 ± 0.43	0.75 ± 0.42	0.78 ± 0.46	1.040	0.299
PTA (%)		109.50 ± 18.79	108.16 ± 40.22	112.61 ± 40.40	1.545	0.123
HBeAg (S/CO)		1745.74 ± 32177.8	2355.23 ± 38449.88	323.60 ± 526.52	0.946	0.344
INR		1.02 ± 0.12	1.03 ± 0.31	1.00 ± 0.31	1.451	0.147
PT (second)		11.76 ± 1.39	11.84 ± 4.38	11.57 ± 3.38	0.986	0.325
HA (ng/mL)		141.29 ± 142.55	147.90 ± 173.41	125.87 ± 160.61	1.946	0.052
LN (μg/L)		54.47 ± 42.36	56.44 ± 55.54	49.87 ± 43.43	1.886	0.060
PLT (× 109/L)		362.35 ± 4394.84	439.84 ± 5251.83	181.54 ± 63.11	0.881	0.379

Alb: Albumin; LDH: Lactate dehydrogenase; CHE: Cholinesterase; WBC: White blood cell; RBC: Red blood cell; HGB: Hemoglobin; ANC: Absolute neutrophil count; ALC: Absolute lymphocyte count; PTA: Prothrombin activity; GOT: Aspartate aminotransferase; GPT: Alanine aminotransferase; HBeAg: Hepatitis B E antigen; INR: International normalized ratio; PT: Plasma prothrombin time; HA: Hyaluronic acid; LN: Laminin; PLT: Platelet count.

Comparison of clinical data between the hepatitis group and the hepatitis cirrhosis group in the modeling cohort

In the modeling cohort, the levels of Alb, LDH, CHE, PTA, HBeAg, WBC, RBC, HGB, ANC and ALC of patients with hepatitis cirrhosis in the hepatitis cirrhosis group were significantly lower than those in the hepatitis group. The levels of GOT/GPT, INR, PT, HA, LN and PLT were significantly higher than those in the hepatitis group (P < 0.05). See Table 2 and Supplementary Table 2.

Table 2 Comparison of clinical data between the hepatitis group and the hepatitis cirrhosis group in the modeling cohort, mean ± SD/n (%).

Index		Total cases (n = 749)	Hepatitis group (n = 688)	Hepatitis cirrhosis group (n = 61)	t/χ2/Z	P value
Gender	Man	487 (65.02)	447 (64.97)	40 (65.57)	0.009	0.924
	Woman	262 (34.98)	241 (35.03)	21 (34.43)
Age (years)			46.12 ± 22.80	51.79 ± 10.06	1.925	0.051
Hepatitis classification	Mild	161 (21.50)	152 (22.09)	9 (14.75)	1.342	0.180
	Moderate	538 (71.83)	491 (71.37)	47 (77.05)
	Serious	50 (6.68)	45 (6.54)	5 (8.20)
Allergy history		44 (5.87)	40 (5.81)	4 (6.56)	0.002	0.963
Disease history	Hypertension	41 (5.47)	38 (5.52)	3 (4.92)	0.605	0.545
	Diabetes	38 (5.07)	35 (5.09)	3 (4.92)
	Heart disease	30 (4.01)	29 (4.22)	1 (1.64)
Operation history		99 (13.22)	90 (13.08)	9 (14.75)	0.137	0.712
Smoking history		95 (12.68)	86 (12.50)	9 (14.75)	0.257	0.612
Drinking history		90 (12.02)	82 (11.92)	8 (13.11)	0.769	0.380
Family history of disease		572 (76.37)	520 (75.58)	52 (85.25)	2.900	0.089
Treatment regimen	Monotherapy	623 (83.18)	568 (82.56)	55 (90.16)	2.317	0.128
	Combination therapy	126 (16.82)	120 (17.44)	6 (9.84)
Alb (× 109/L)		43.02 ± 4.65	43.10 ± 4.60	42.13 ± 1.06	2.891	0.004
LDH (U/L)		210.07 ± 73.36	210.33 ± 11.13	207.17 ± 10.49	2.133	0.033
CHE (U/L)		7298.93 ± 2190.67	7336.36 ± 1184.27	6876.75 ± 1236.59	2.895	0.004
PTA (%)		108.16 ± 19.22	108.92 ± 19.13	99.62 ± 18.22	3.653	0.000
HBeAg (S/CO)		2355.23 ± 38449.88	2543.73 ± 314.87	229.18 ± 303.99	55.175	0.000
WBC (× 109/L)		5.19 ± 1.51	5.25 ± 1.51	4.55 ± 1.22	3.520	0.001
RBC (× 1012/L)		4.73 ± 0.51	4.74 ± 0.31	4.64 ± 0.23	2.460	0.014
HGB (g/L)		145.56 ± 15.96	145.70 ± 6.12	143.97 ± 3.97	2.167	0.031
ANC (× 109/L)		2.81 ± 1.07	2.85 ± 1.09	2.45 ± 0.81	2.798	0.005
ALC (× 109/L)		1.88 ± 0.66	1.90 ± 0.66	1.69 ± 0.60	2.399	0.017
GOT/GPT		0.75 ± 0.42	0.74 ± 0.42	0.86 ± 0.36	2.162	0.031
INR		1.03 ± 0.11	1.02 ± 0.12	1.07 ± 0.13	3.097	0.002
PT (second)		11.84 ± 1.38	11.79 ± 1.37	12.35 ± 1.43	3.049	0.002
HA (ng/mL)		147.90 ± 145.41	144.64 ± 142.14	184.75 ± 175.34	2.069	0.039
LN (μg/L)		56.44 ± 45.54	55.72 ± 24.77	64.63 ± 23.16	2.706	0.007
PLT (× 109/L)		439.84 ± 5251.83	179.83 ± 54.36	3372.49 ± 18283.84	4.612	0.000

Alb: Albumin; LDH: Lactate dehydrogenase; CHE: Cholinesterase; WBC: White blood cell; RBC: Red blood cell; HGB: Hemoglobin; ANC: Absolute neutrophil count; ALC: Absolute lymphocyte count; PTA: Prothrombin activity; GOT: Aspartate aminotransferase; GPT: Alanine aminotransferase; HBeAg: Hepatitis B E antigen; INR: International normalized ratio; PT: Plasma prothrombin time; HA: Hyaluronic acid; LN: Laminin; PLT: Platelet count.

Logistic regression analysis of risk factors for hepatitis cirrhosis

Hepatitis cirrhosis = 1, and hepatitis = 0 were taken as the dependent variables, and the factors significantly different in the above single factor analysis were taken as the covariates for binary logic regression analysis. The results showed that the levels of GOT/GPT, PTA and HBeAg were all the influencing factors of hepatitis cirrhosis.

Log (P) = 6.572 + 0.640 × GOT/GPT - 0.049 × PTA - 0.001 × HBeAg. See Table 3.

Table 3 Logistic regression analysis of risk factors for hepatitis cirrhosis.

	B	SE	Wals	P value	OR	95%CI
Alb	-0.033	0.043	0.608	0.736	0.967	0.890 to 1.052
LDH	-0.002	0.002	1.211	0.271	0.998	0.993 to 1.002
CHE	0.000	0.000	1.565	0.211	1.000	1.000 to 1.000
GOT/GPT	0.640	0.305	4.401	0.036	1.896	1.043 to 3.448
INR	14.952	9.843	2.307	0.129	2.536	0.013 to 7.450
PT	-1.566	0.877	3.184	0.074	0.209	0.037 to 1.167
PTA	-0.049	0.022	5.252	0.022	0.952	0.913 to 0.993
HBeAg	-0.001	0.000	5.310	0.021	0.999	0.999 to 1.000
HA	0.000	0.001	0.088	0.767	1.000	0.998 to 1.003
LN	0.000	0.004	0.015	0.901	1.000	0.992 to 1.007
PLT	0.000	0.000	0.574	0.449	1.000	1.000 to 1.000
WBC	-1.527	0.794	3.701	0.054	0.217	0.046 to 1.029
RBC	-0.311	0.586	0.281	0.596	0.733	0.233 to 2.310
HGB	0.027	0.018	2.129	0.145	1.027	0.991 to 1.065
ANC	1.302	0.864	2.268	0.132	3.676	0.676 to 20.002
ALC	0.963	0.831	1.341	0.247	2.619	0.513 to 13.364
Constant	6.572	5.590	1.382	0.240	715.135

OR: Odds ratio; CI: Confidence interval; Alb: Albumin; LDH: Lactate dehydrogenase; CHE: Cholinesterase; WBC: White blood cell; RBC: Red blood cell; HGB: Hemoglobin; ANC: Absolute neutrophil count; ALC: Absolute lymphocyte count; PTA: Prothrombin activity; GOT: Aspartate aminotransferase; GPT: Alanine aminotransferase; HBeAg: Hepatitis B E antigen; INR: International normalized ratio; PT: Plasma prothrombin time; HA: Hyaluronic acid; LN: Laminin; PLT: Platelet count.

The establishment of nomogram model for predicting the occurrence of hepatitis cirrhosis

The obtained three independent risk factors (GOT/GPT, PTA and HBeAg) were used to construct a prediction model using R software, and the nomogram model was established (Figure 1A). After the generated nomogram was calibrated (Figure 1B), the predicted event was in high consistency with the actual event. The area under the ROC curve of the nomogram prediction model was 0.693 [95% confidence interval (CI): 0.631 to 0.756] (Figure 1C). The decision analysis curve is shown in Figure 1D, where the X axis represents the threshold probability, the Y axis represents the net gain, and the solid black line represents the net gain using the nomogram prediction model. The curve shows a high yield, further confirming the effectiveness of the nomogram prediction model.

Figure 1 Establishment of a nomogram model for predicting the occurrence of hepatitis cirrhosis. A: Nomogram; B: Calibration curve of modeling queue; C: Receiver operating characteristic curve of modeling queue; D: Modeling queue decision analysis curve. PTA: Prothrombin activity; GOT: Aspartate aminotransferase; GPT: Alanine aminotransferase; HBeAg: Hepatitis B E antigen; AUC: Aera under curve.

Validation of nomogram model

Based on clinical data from a validation cohort (n = 321) of patients (Table 4), the ROC curve was used for external validation of the hepatitis cirrhosis risk nomogram, and the area under the ROC curve was 0.675 (95%CI: 0.561 to 0.790) (Figure 2A). The slope of the generated nomogram calibration curve was close to 1 (Figure 2B), and the results of Hosmer-Lemeshow test: χ² = 7.884, P = 0.445 > 0.05, The decision curve shows a high net benefit for the model (Figure 2C), suggesting that the nomogram model performed well in the validation group.

Figure 2 Validation of the nomogram model. A: Validation queue receiver operating characteristic curve; B: Validation queue calibration curve; C: Verify the queue decision analysis curve. AUC: Aera under curve.

Table 4 Comparison of clinical data between the validation cohort hepatitis group and the hepatitis cirrhosis group, mean ± SD/n (%).

Index		Total cases (n = 321)	Hepatitis group (n = 295)	Hepatitis cirrhosis group (n = 26)	t/χ2/Z	P value
Gender	Man	192 (59.81)	180 (61.02)	12 (46.15)	2.196	0.138
	Woman	129 (40.19)	115 (38.98)	14 (53.85)
Age (years)			46.89 ± 21.68	53.85 ± 9.03	1.623	0.106
Hepatitis classification	Mild	81 (25.23)	78 (26.44)	3 (11.54)	1.718	0.086
	Moderate	215 (66.98)	195 (66.10)	20 (76.92)
	Serious	25 (7.79)	22 (7.46)	3 (11.54)
Alb (× 109/L)		43.32 ± 4.79	43.44 ± 3.70	41.97 ± 2.64	1.981	0.049
LDH (U/L)		212.28 ± 129.41	203.94 ± 66.37	306.92 ± 390.42	3.979	0.000
CHE (U/L)		7557.79 ± 2369.69	7617.54 ± 1352.37	6879.85 ± 1506.37	2.642	0.009
PTA (%)		112.61 ± 17.40	114.05 ± 16.15	96.36 ± 22.54	5.166	0.000
HBeAg (S/CO)		323.60 ± 526.52	340.11 ± 333.91	136.59 ± 395.57	2.933	0.004
WBC (× 109/L)		5.20 ± 1.59	5.28 ± 1.57	4.30 ± 1.48	3.065	0.002
RBC (× 1012/L)		4.74 ± 0.49	4.76 ± 0.48	4.46 ± 0.47	3.060	0.002
HGB (g/L)		145.22 ± 16.54	145.96 ± 15.98	136.77 ± 20.42	2.744	0.006
ANC (× 109/L)		2.89 ± 1.25	2.94 ± 1.26	2.24 ± 0.96	2.761	0.006
ALC (× 109/L)		1.84 ± 0.64	1.85 ± 0.45	1.63 ± 0.41	2.406	0.017
GOT/GPT		0.78 ± 0.46	0.77 ± 0.46	1.00 ± 0.44	2.452	0.015
INR		1.00 ± 0.11	0.99 ± 0.22	1.11 ± 0.12	2.743	0.006
PT (second)		11.57 ± 1.38	11.45 ± 0.96	12.95 ± 3.35	5.576	0.000
HA (ng/mL)		125.87 ± 134.61	124.44 ± 34.51	142.08 ± 37.40	2.482	0.014
LN (μg/L)		49.87 ± 33.43	48.29 ± 31.21	67.75 ± 49.89	2.878	0.004
PLT (× 109/L)		181.54 ± 63.11	185.44 ± 61.16	137.27 ± 69.18	3.809	0.000

Alb: Albumin; LDH: Lactate dehydrogenase; CHE: Cholinesterase; WBC: White blood cell; RBC: Red blood cell; HGB: Hemoglobin; ANC: Absolute neutrophil count; ALC: Absolute lymphocyte count; PTA: Prothrombin activity; GOT: Aspartate aminotransferase; GPT: Alanine aminotransferase; HBeAg: Hepatitis B E antigen; INR: International normalized ratio; PT: Plasma prothrombin time; HA: Hyaluronic acid; LN: Laminin; PLT: Platelet count.

DISCUSSION

Hepatitis cirrhosis is a severe, chronic liver disease that seriously affects the quality of life and health of patients and places a heavy burden on the health-care system. With the social and economic development and the improvement of living standards, the prevalence of hepatitis and cirrhosis has gradually increased and become an important issue in the field of global public health[5]. Therefore, understanding the clinical characteristics of patients with hepatitis and cirrhosis and building prediction models are of great significance for the early diagnosis, treatment and management of the disease.

In this study, the patients were divided into a modeling cohort and a verification cohort, and the clinical data of the patients were compared. The results showed that in the modeling cohort, there were 688 patients with hepatitis (91.86%) and 61 patients with hepatitis cirrhosis (8.14%), indicating that the probability of liver cirrhosis for patients with hepatitis was high. This result was also close to that in the previous study by Liao et al[6]. At the same time, through the comparison of clinical data between the two groups in the modeling cohort, it was found that the levels of Alb, LDH, CHE, PTA, HBeAg, WBC, RBC, HGB, ANC and ALC of patients with hepatitis cirrhosis in the modeling cohort were significantly lower than those of the hepatitis group. The levels of GOT/GPT, INR, PT, HA, LN and PLT were significantly higher than those of the hepatitis group. Hepatitis cirrhosis = 1, and hepatitis = 0 were taken as the dependent variables, and the factors significantly different in the above single factor analysis were taken as the covariates for binary logic regression analysis. The results showed that the levels of GOT/GPT, PTA and HBeAg were all the influencing factors of hepatitis cirrhosis. The analysis was performed for the following reasons: GOT and GPT are sensitive indicators of hepatocyte injury. In patients with hepatitis, hepatitis virus infection further aggravates the inflammatory reaction in the liver, and in this process, the release of immune cells in the body will be further increased. When the liver tissue is infiltrated and activated by the immune cells, the inflammatory reaction in the tissue is further triggered, resulting in the massive release of inflammatory mediators and cytokines that directly damage the hepatocytes to a certain extent, leading to the increased release of GOT and GPT to a certain extent[7,8]. With the continuous progress of inflammation, liver tissue undergoes further fibrosis reaction, which causes the normal structure of the liver to be replaced by fibrous tissue, thus affecting the normal regeneration ability and metabolic function of the liver, aggravating the necrosis of hepatocytes and the increase in the release of GOT and GPT[9,10]. Therefore, the ratio of GOT to GPT can also be used as a predictive index for the occurrence of hepatitis cirrhosis.

Related studies have found that patients' liver functions are damaged, which further leads to the reduction in the synthesis of coagulation factors in the body, and thus to a certain extent causes patients to have a certain degree of coagulation dysfunction[11]. This also suggests that coagulation factors can be used as indicators in patients with cirrhosis. As an important indicator to assess the coagulation function in patients with cirrhosis of liver, PTA shows a significant decline due to the reduction in the synthesis of coagulation factors, which in turn leads to the prolongation of coagulation time to varying degrees[12]. And You et al[13] also found that the PTA level of patients with hepatitis and cirrhosis was significantly reduced. This is also consistent with the results of this study. The analysis was mainly performed because prothrombin was activated into thrombin through a series of enzymatic reactions in the body, which involved a variety of coagulation factors and cofactors such as factor VII and factor X[14]. When an inflammatory reaction occurs in the liver of a patient, it will lead to liver dysfunction of the patient, resulting in impaired metabolism and clearance of coagulation factors, and further blocked activation of prothrombin, which increases the deactivation rate of prothrombin in the body, and thus reduces the activity of prothrombin to a certain extent[15,16]. Relevant studies have shown that in patients with HBeAg-positive cirrhosis, HBV replication is active in the body and it is highly infectious. When the virus replicates in a large amount, it will aggravate the damage to liver function to a certain extent and accelerate the progression of liver cirrhosis[17]. Meanwhile, Wang et al[18] found a significant correlation between the pathological progress of hepatitis patients and their HBeAg dynamic level in the propensity score matching study. This further indicates that HBeAg level can be used as an important indicator for predicting the progression of hepatitis patients to cirrhosis. The main analysis reasons are as follows: HBeAg-positive state is usually accompanied by the activation of cluster of differentiation (CD) 4 + T cells, CD8 + T cells and natural killer cells, which are mainly responsible for clearing away the infected hepatocytes by releasing cytotoxic substances and producing cytokines. However, in this process, healthy hepatocytes are also damaged to different degrees, thus aggravating the process of liver inflammation and fibrosis[19,20]. Low HBeAg level, on the other hand, reflects the mutation or deletion of HBeAg epitope, which to some extent leads to the mutation of virus strains, and further weakens the virus replication capacity, thus slowing the progression of liver inflammation and fibrosis to a certain extent.

In order to determine the predictive value of GOT/GPT, PTA and HBeAg levels in patients with hepatitis cirrhosis, in this study, the nomogram model was established using the modeling cohort. The area under the ROC curve of the nomogram prediction model was larger, and the prediction performance was good. The probability of liver failure in the verification group was predicted by the subjects in the verification cohort using the nomogram, which indicated that it had certain predictive value. In addition, the factors of this model are all medical records of patients, so it is easy to obtain and has high clinical adaptability. In addition, verification of the cohort calibration curve revealed a small deviation of the actual result curve from the calibration curve, indicating a high agreement between the predicted and actual events. It could be seen from the decision analysis curve of the verification queue that the decision analysis curve at the top right corner generally indicated that the model had high true positive rate and low false positive rate, which meant that the model had certain accuracy and reliability.

The prediction model in this study is based on retrospective data. The data collection is from a single medical center, and there may be selection bias. Moreover, the sample size is relatively limited, and only 1070 patients were included. This may influence the stability and generalization ability of the model. The accuracy of the predicted results of this model is also influenced by the interpretation and application of the results by clinicians. In actual clinical decision-making, it may be necessary to combine other clinical indicators and doctors’ experience to make comprehensive judgments.

CONCLUSION

GOT/GPT, PTA and HBeAg are all independent factors influencing the progression of hepatitis to cirrhosis in patients with hepatitis. The nomogram prediction model for the occurrence of hepatitis cirrhosis constructed in this study showed good prediction ability, and the predicted events were highly consistent with the actual events. As a tool for predicting the occurrence of hepatitis cirrhosis in hepatitis patients, the model has high application value. However, due to the limited clinical data of our patients in this retrospective study, which may also affect the results of this study to some extent, further prospective studies are required to establish more comprehensive prediction models.

ACKNOWLEDGEMENTS

Special thanks to Dr. Xu C of Dalian Institute of Public Health and Dalian Public Health Clinical Center for providing scientific guidance and manuscript revision for this manuscript.