Abstract
Background and Goals:
The Fibrosis-4 Index (FIB-4) has demonstrated a strong association with severe liver disease (SLD) outcomes in primary care, but previous studies have only evaluated this relationship using 1 or 2 FIB-4 scores. In this study, we determined the association of FIB-4 as a time-varying covariate with SLD risk using time-dependent Cox regression models.
Study:
This retrospective cohort study included primary care patients with at least 2 FIB-4 scores between 2012 and 2021. The outcome was the occurrence of an SLD event, a composite of cirrhosis, complications of cirrhosis, hepatocellular carcinoma, and liver transplantation. The primary predictor was FIB-4 advanced fibrosis risk, categorized as low-(<1.3), indeterminate-(1.3≤FIB-4<2.67), and high-risk (≥2.67). FIB-4 scores were calculated and the index, last, and maximum FIB-4s were identified. Time-dependent Cox regression models were used to estimate hazard ratios (HR) and their corresponding 95% confidence intervals (CI) with adjustment for potentially confounding covariates.
Results:
In the cohort, 20,828 patients had a median of 5 (IQR: 3–11) FIB-4 scores each and 3% (n=667) suffered an SLD outcome during follow-up. Maximum FIB-4 scores were indeterminate-risk for 34% (7,149) and high-risk for 24% (4,971) of the sample, and 32% (6,692) of patients had an increase in fibrosis risk category compared to their index value. The adjusted Cox regression model demonstrated an association between indeterminate- (HR 3.21; 95%CI 2.33 – 4.42)) and high-risk (HR 20.36; 95%CI 15.03 – 27.57) FIB-4 scores with SLD outcomes.
Conclusions:
Multiple FIB-4 values per patient are accessible in primary care, FIB-4 fibrosis risk assessments change over time, and high-risk FIB-4 scores (≥ 2.67) are strongly associated with severe liver disease outcomes when accounting for FIB-4 as a time-varying variable.
Keywords: Non-invasive tests, fibrosis, cirrhosis, chronic liver disease
Introduction
The Fibrosis-4 Index (FIB-4) is a non-invasive, serologic composite score for assessing advanced fibrosis risk in patients with chronic liver disease (CLD) [1–3]. Advanced fibrosis is the single best predictor of CLD that progresses to severe liver disease (SLD) outcomes including cirrhosis, complications of cirrhosis, hepatocellular carcinoma (HCC), and need for liver transplantation[4–9]. Detection of advanced fibrosis historically relied on liver biopsy, but non-invasive tools such as FIB-4 have emerged as viable alternatives to gauge advanced fibrosis risk [2, 3, 10]. FIB-4 is calculated from commonly ordered blood tests, including aminotransferases (aspartate [AST] and alanine [ALT]) and platelets (FIB-4=[Age x AST]/[Platelets x √ALT]), and is accessible in primary care [1, 11]. Population-level and primary care-based studies have demonstrated that FIB-4 scores at high-risk for advanced fibrosis (FIB-4 ≥ 2.67) are associated with SLD outcomes in patients with and without known CLD [12–14].
FIB-4’s relationship with future SLD events argues for more regular and routine FIB-4 calculation, facilitated by the electronic health record (EHR), to assess advanced fibrosis risk in patients with known CLD and to signal the need to pursue diagnosis in those patients without known CLD [13, 14]. However, work to date demonstrating the strong relationship between high-risk FIB-4 scores and SLD only evaluated one or two FIB-4 values for each patient, despite the availability of numerous unique FIB-4 scores longitudinally in primary care data [12–14]. FIB-4 can change over time and is likely better treated as a time-varying covariate than a covariate fixed at the initial value during a study’s inception or follow-up period. The relationship between FIB-4 risk and SLD may not be constant over time [15].
In this study, we determined the association of FIB-4 advanced fibrosis risk as a time-varying covariate with the outcome of an SLD event. This analysis provides estimates of the association of FIB-4 history with the occurrence of an SLD outcome. We also evaluated the relationship of a patient’s maximum FIB-4 score during the follow-up period with the occurrence of SLD to understand whether there is a threshold effect on the relationship between FIB-4 fibrosis risk and SLD outcomes.
Patients and Methods
Study Design, Setting, and Data Source
This retrospective study of EHR data from patients in a primary care clinic evaluates the association between FIB-4 as a time-varying covariate with the risk of an SLD event using a time-dependent Cox regression analysis. EHR data came from patients receiving care at the internal medicine patient-centered medical home at the Medical University of South Carolina (MUSC) between 2012 and 2021. This study was approved by the Institutional Review Board at MUSC.
Patients
Patients with at least one visit to the primary care clinic and one occurrence of laboratory testing for aminotransferases and a platelet count were considered for inclusion. After identification of these patients, FIB-4 scores were calculated for each set of aminotransferases where both AST and ALT were ≤ 350 IU/L, and there was a platelet count at the time of, or within 6 months of, the AST and ALT values. Limits on the AST and ALT values were chosen to avoid calculating FIB-4 scores from inputs that more likely reflect acute liver inflammation than advanced fibrosis risk. Beginning with the chronologically first acceptable inputs for FIB-4 calculation, FIB-4 scores were serially calculated for each patient each time there was an AST/ALT result. We required at least 6 months between each FIB-4 value to ensure unique AST, ALT, and platelet count inputs. Study cohort inclusion required at least 2 FIB-4 scores. Patients with an SLD outcome by ICD-9/10 code (Supplementary Table 1) prior to their index FIB-4 score were excluded [13, 14, 16–18].
Outcomes
The occurrence of a severe liver disease (SLD) event was the primary outcome of interest. SLD events were composites of cirrhosis, complications of cirrhosis, hepatocellular carcinoma, and liver transplantation identified by ICD-9/10 codes. Complications of cirrhosis included portal hypertension, esophageal varices, ascites, hepatorenal syndrome, and hepatic encephalopathy. Previously studied algorithms of ICD-9/10 codes for detecting cirrhosis from administrative EHR data were used for SLD ascertainment, and all outcomes were confirmed on chart review through identification of cirrhosis or its complications in a note, evidence of nodularity or cirrhosis on imaging, evidence of a prolonged prothrombin time or thrombocytopenia, and the absence of an alternative cause for the SLD diagnostic code (Supplementary Table 1) [16–18].
Primary Predictor Variable
FIB-4 advanced fibrosis risk, as a repeated, time-varying predictor, was the primary predictor variable of interest. Each calculated FIB-4 score for every patient included in the cohort was categorized as low- (FIB-4 < 1.3), indeterminate- (1.3 ≤ FIB-4 < 2.67), or high-risk (FIB-4 ≥ 2.67) for advanced fibrosis [2, 19]. The first FIB-4 (index), the most recent FIB-4 (last), and the maximum FIB-4 (max) was identified for each cohort patient.
Other Covariates
Potentially confounding variables included sex (female / male), race (Black, non-Black), smoking history (ever/never), marital status (married/unmarried), and body mass index (continuous). Possible confounders also included comorbidities which were identified using Elixhauser ICD-9/10 algorithms comprised cardiovascular disease, chronic kidney disease, diabetes mellitus, hyperlipidemia, hypertension, and hypothyroidism [20, 21]. These covariates were chosen a priori using risk factors associated with chronic liver disease and health outcomes [22–25]. Age was not included as a covariate out of concern for endogeneity with FIB-4, as age is a component of the FIB-4 equation [1]. We included diagnosis codes for CLD etiologies, including alcohol-related liver disease, nonalcoholic fatty liver disease (NAFLD), viral hepatitis, a combination of etiologies, and other (e.g. hemochromatosis, autoimmune hepatitis, etc.), using composites of ICD-9/10 codes validated in the literature (Supplementary Table 1) [18]. CLDs were included if the ICD-9/10 codes were detected before or during the follow-up period.
Statistical Analysis
Descriptive analyses were performed for the cohort and included the median and interquartile range (IQR) of FIB-4 scores analyzed per included patient. The proportion of the cohort experiencing an SLD outcome during the follow-up period was calculated, and proportions of the sample having an SLD event were calculated by index, last, and maximum FIB-4 advanced fibrosis risk category.
A Kaplan-Meier analysis of time to SLD was performed for the overall sample by time-dependent FIB-4 advanced fibrosis risk category. Time-dependent Cox regression models were developed for the outcome of the occurrence of an SLD event with a primary predictor variable of FIB-4 advanced fibrosis risk strata (low, indeterminate, and high) as a time-varying covariate [26]. We used the counting process set up and created multiple rows of data for each subject to account the time varying values of FIB4. Each additional row represents a time interval and the corresponding FIB4 values for that interval. The event variable for that row is 1 if the time interval ends with SLD and it is zero otherwise. It is important to note that the multiple rows of data per subject doesn’t require adjustment for correlation due to clustering (frailty) since the Cox likelihood at any time point only uses one copy of the subject. An unadjusted model was developed followed by a model adjusting for the potentially confounding demographic, clinical, and comorbidity covariates. Estimated hazard ratios and 95% confidence intervals were calculated for each model. Models were assessed for goodness of fit. Interactions were explored and multicollinearity was assessed using variance inflation factor. A sensitivity analysis was performed in the subset of cohort patients with at least 3 qualifying FIB-4 scores.
Using the advanced fibrosis risk category of the maximum FIB-4 score calculated for each patient, we calculated the sensitivity, specificity, positive (PPV), and negative predictive value (NPV) of the max FIB-4 (≥ 2.67 and ≥ 1.3) for detecting a future SLD outcome.
Previous studies have demonstrated advancing age as a potential confounder in FIB-4’s ability to predict advanced fibrosis [27]. To address this concern, we performed a sensitivity analysis by creating subgroups based on patient age at the time of index FIB-4 (age<65 and age≥65 years). Unadjusted and adjusted Cox regression models were developed within each age subgrouping. Alcohol use, in alcohol-related liver disease and other CLD etiologies, is associated with an increased risk of cirrhosis [28, 29]. Though alcohol use documentation in the EHR is suboptimal, we performed a sensitivity analysis including alcohol use disorder and alcohol-related liver disease as a covariate (by ICD-9/10 codes in Supplementary Table 1) in an adjusted Cox regression model. Additionally, FIB-4’s performance has been most extensively validated in NAFLD; thus we developed Cox regression models including NAFLD as a covariate. SAS 9.4 (SAS, Cary, NC) and higher was used for all analyses.
Results
The cohort included 20,828 unique patients during the study period (Supplementary Figure 1). Patients had a mean age of 52.1 years (±17.4), were 65% female, 43% Black, and had a mean BMI of 29.7 kg/m2 (Table 1). By the end of follow-up, 13.9% of the cohort had a diagnosis of CLD, with 4.7% diagnosed with alcohol-related liver disease (or alcohol use disorder), 3.4% diagnosed with NAFLD, and 2.4% diagnosed with viral hepatitis. Of the sample, 67% had hypertension, 30% had diabetes, 19% had chronic kidney disease, and 31% had cardiovascular disease. Patients were followed for a median of 7.3 (IQR: 4.6–8.8) years. Included patients had a median of 5 (IQR: 3–11) FIB-4 scores per patient and 3% (n=667) suffered an SLD outcome during the follow-up period (Figure 1).
Table 1.
Characteristics of the study cohort.
| Overall | |
|---|---|
| Characteristics | n=20,828 |
| FIB-4 scores / patient, median (IQR) | 5 (3 – 11) |
| Index FIB-4 Fibrosis Category | |
| Low | 13,146 (63.1%) |
| Indeterminate | 6,091 (29.2%) |
| High | 1,490 (7.2%) |
| Maximum FIB-4 Fibrosis Risk Category | |
| Low | 8,708 (41.8%) |
| Indeterminate | 7,149 (34.3%) |
| High | 4,971 (23.9%) |
| CLD Diagnoses (%) | |
| CLD Overall | 2,900 (13.9%) |
| Alcohol | 984 (4.7%) |
| NAFLD | 700 (3.4%) |
| Viral | 495 (2.4%) |
| Other | 294 (1.4%) |
| Mixed | 427 (2.1%) |
| Demographics (%) | |
| Age, mean (SD) | 52.1 (17.4) |
| Sex | |
| Female | 13,470 (64.7%) |
| Male | 7,358 (35.3%) |
| Race | |
| Black | 9,022 (43.3%) |
| Other | 642 (3.1%) |
| White | 11,164 (53.6%) |
| Married | 9,315 (44.7%) |
| Smoking history | 2,429 (11.7%) |
| Comorbidities (%) | |
| BMI, mean (SD) | 29.7 (7.9) |
| Hypertension | 13,977 (67.1%) |
| Diabetes | 6,141 (29.5%) |
| Hyperlipidemia | 11,774 (56.5%) |
| CVD | 6,456 (31.0%) |
| Hypothyroidism | 3,683 (17.7%) |
| CKD | 3,879 (18.6%) |
| SCD | 569 (2.7%) |
BMI=body mass index (kg/m2). CKD=chronic kidney disease. CLD=chronic liver disease. CVD=cardiovascular disease. FIB-4=Fibrosis-4 Index. IQR=inter-quartile range.
NAFLD=nonalcoholic fatty liver disease. SCD=sickle cell disease.
Figure 1.
Kaplan-Meier curve for severe liver disease risk free survival by time-dependent FIB-4 risk category for the entire sample.
Of the index FIB-4 scores, 63%, 29%, and 7% were low-, indeterminate-, and high-risk for advanced fibrosis, respectively. Patients’ last FIB-4 scores were 60% (12,560) low-risk, 29% (6,035) indeterminate-risk, and 11% (2,233) high-risk for advanced fibrosis (Table 2). Maximum FIB-4 scores were indeterminate-risk for 34% (7,149) and high-risk for 24% (4,971) of the sample. Of the cohort, 27% (5,524) had a change in advanced fibrosis risk from their index to last FIB-4 score, with 16% (3,296) having an increase in fibrosis risk category and 11% (2,228) transitioning to a lower-risk advanced fibrosis category (Table 2). Using the maximum FIB-4 scores for each patient, 32% (6,692) of patients had an increase in advanced fibrosis risk category compared to their index value. Of patients with low-risk FIB-4 scores throughout the follow-up period, 0.5% (43/8,708) experienced an SLD event.
Table 2.
Occurrence of SLD by combinations of (a) index and last FIB-4 scores, and (b) index and maximum FIB-4 scores in patients prior to their terminal event (SLD or end of study).
| SLD | No SLD | Total | ||
|---|---|---|---|---|
| Index FIB-4 | Last FIB-4 | |||
| Low | 135 (1.0%) | 13,011 (99.0%) | 13,146 | |
| Low | 77 (0.7%) | 10,763 (99.3%) | 10,840 | |
| Indet. | 29 (1.5%) | 1,938 (98.5%) | 1,967 | |
| High | 29 (8.6%) | 310 (91.4%) | 339 | |
| Indet. | 216 (3.5%) | 5,875 (96.5%) | 6,091 | |
| Low | 23 (1.5%) | 1,518 (98.5%) | 1,541 | |
| Indet. | 109 (3.1%) | 3,451 (96.9%) | 3,560 | |
| High | 84 (8.5%) | 906 (91.5%) | 990 | |
| High | 316 (19.9%) | 1,275 (80.1%) | 1,591 | |
| Low | 7 (3.9%) | 172 (96.1%) | 179 | |
| Indet. | 31 (6.1%) | 477 (93.9%) | 508 | |
| High | 278 (30.8%) | 626 (69.2%) | 904 | |
| Total | 667 (3.2%) | 20,161 (96.8%) | 20,828 | |
| SLD | No SLD | Total | ||
| Index FIB-4 | Max FIB-4 | |||
| Low | 135 (1.0%) | 13,011 (99.0%) | 13,146 | |
| Low | 43 (0.5%) | 8,665 (99.5%) | 8,708 | |
| Indet. | 37 (1.1%) | 3,275 (98.9%) | 3,312 | |
| High | 55 (4.9%) | 1,071 (95.1%) | 1,126 | |
| Indet. | 216 (3.5%) | 5,875 (96.5%) | 6,091 | |
| Low | 0 | 0 | 0 | |
| Indet. | 102 (2.7%) | 3,735 (97.3%) | 3,837 | |
| High | 114 (5.1%) | 2,140 (94.9%) | 2,254 | |
| High | 316 (19.9%) | 1,275 (80.1%) | 1,591 | |
| Low | 0 | 0 | 0 | |
| Indet. | 0 | 0 | 0 | |
| High | 316 (19.9%) | 1,275 (80.1%) | 1,591 | |
| Total | 667 (3.2%) | 20,161 (96.8%) | 20,828 |
FIB-4=Fibrosis-4 Index. Indet=Indeterminate-risk for advanced fibrosis. SLD=severe liver disease.
In the unadjusted Cox regression model with FIB-4 advanced fibrosis risk category as a time-varying covariate, indeterminate-risk FIB-4 scores were associated with 3-fold risk of SLD (HR: 3.25; 95%CI 2.39–4.42) and high-risk FIB-4 scores were associated with a 24-fold hazard of an SLD event (HR: 24.59; 95%CI 18.56 – 32.57) [Table 3]. The adjusted Cox regression model demonstrated an association between a history of indeterminate- (HR 3.21; 95%CI 2.33 – 4.42)) and high-risk (HR 20.36; 95%CI 15.03 – 27.57) FIB-4 scores with SLD events.
Table 3.
Estimated Hazard Ratios and 95% Confidence Intervals using time-dependent Cox regression models with repeated FIB-4 measurements as the primary predictor variable for the outcome of severe liver disease.
| Regression Models | ||||
|---|---|---|---|---|
| Unadjusted Model | Adjusted Model | |||
| Predictors | HR | 95% CI | HR | 95% CI |
| FIB-4 Risk | ||||
| Low (ref.) | ||||
| Indeterminate | 3.25 | 2.39 – 4.42 | 3.21 | 2.33 – 4.42 |
| High | 24.59 | 18.56 – 32.57 | 20.36 | 15.03 – 27.57 |
| Black | 0.64 | 0.51 – 0.81 | ||
| Male | 1.19 | 0.95 – 1.50 | ||
| Unmarried | 1.71 | 1.34 – 2.19 | ||
| Smoking history | 2.44 | 1.89 – 3.15 | ||
| BMI | 1.00 | 0.98 – 1.01 | ||
| Hypertension | 1.18 | 0.85 – 1.63 | ||
| Diabetes | 2.17 | 1.70 – 2.76 | ||
| Hyperlipidemia | 0.50 | 0.39 – 0.63 | ||
| CVD | 1.10 | 0.87 – 1.37 | ||
| Hypothyroidism | 0.95 | 0.71 – 1.27 | ||
| CKD | 1.66 | 1.30 – 2.10 | ||
FIB-4=Fibrosis-4 index. HR=hazard ratio. CI=confidence interval. Ref=reference.
BMI=body mass index. CVD=cardiovascular disease. CKD=chronic kidney disease.
Maximum FIB-4 scores at high-risk for advanced fibrosis had a sensitivity of 72.7%, a specificity of 77.7%, a PPV of 9.8%, and an NPV of 98.9% for detecting SLD (Table 4). A maximum FIB-4 value at indeterminate risk or greater for advanced fibrosis had a sensitivity of 93.6%, a specificity of 34.0%, a PPV of 5.1%, and an NPV of 54.6% for an SLD occurrence. Sensitivity analyses of cohort subgroups based on age (Supplementary Tables 2a and 2b) and models including NAFLD (Supplementary Table 3) and alcohol use diagnosis codes (Supplementary Table 4) yielded similar findings.
Table 4.
Performance characteristics of (a) FIB-4 ≥ 2.67 and (b) FIB-4 ≥ 1.30 for detecting SLD.
| Outcomes | |||
|---|---|---|---|
| Max FIB-4 | SLD | No SLD | Total |
| ≥ 2.67 | 485 | 4,486 | 4,971 |
| < 2.67 | 182 | 15,675 | 15,857 |
| Total | 667 | 20,161 | 20,828 |
| Sensitivity: 485 / 667 = 72.7% | |||
| Specificity: 15,675 / 20,161 = 77.7% | |||
| PPV: 485 / 4,971 = 9.8% | |||
| NPV: 15,675 / 15,857 = 98.9% | |||
| Max FIB-4 | SLD | No SLD | Total |
| ≥ 1.30 | 624 | 11,496 | 12,120 |
| < 1.30 | 43 | 8,665 | 8,708 |
| Total | 667 | 20,161 | 20,828 |
| Sensitivity: 624 / 667 = 93.6% | |||
| Specificity: 8,665 / 20,161 = 43.0% | |||
| PPV: 624 / 12,120 = 5.1% | |||
| NPV: 8,665 / 15,857 = 54.6% | |||
FIB-4=Fibrosis-4 Index. NPV=negative predictive value. PPV=positive predictive value.
SLD=severe liver disease (composite of cirrhosis, complications of cirrhosis, hepatocellular carcinoma, and need for liver transplantation).
Discussion
This study demonstrates the frequent availability of FIB-4 scores in primary care practice, as the 20,868 patient cohort had a median of 5 unique FIB-4 scores per person and more than 75% of the sample had at least 3 FIB-4 values (IQR: 3–11) during the 2012–2021 study period. Of the patients studied, 32% had a FIB-4 score at increased risk for advanced fibrosis following their index value, and 16% of the sample had a lower-risk future FIB-4 value. Taking account of these changes in FIB-4 advanced fibrosis risk during the follow-up period, the Cox regression models featuring FIB-4 risk as a time-varying predictor variable revealed a strong association between indeterminate (HR 3.21; 95%CI 2.33 – 4.42) and high-risk (HR 20.36; 95%CI 15.03 – 27.57) FIB-4 scores with SLD events, when adjusting for potentially confounding variables. Lastly, having a high-risk FIB-4 score at any point in time may be an important signal for detecting future SLD outcomes, as 1 in 10 (9.8%) of patients with a maximum FIB-4 ≥ 2.67 experienced an SLD event and this maximum FIB-4 threshold had a negative predictive value of 98.9% for SLD. In patients with exclusively low-risk FIB-4 scores throughout the study period, only 0.5% (43/8,708) experienced an SLD event.
The ubiquity of FIB-4 inputs in primary care and the accessibility of these inputs within the EHR provide clinicians with the potential to longitudinally follow FIB-4 advanced fibrosis risk over time [30]. Current guidelines recommending FIB-4 for fibrosis risk stratification focus on the calculation at a single point in care, with ambiguity surrounding when FIB-4 should be reassessed (e.g. every 2–3 years) and little guidance as to how (or if) more historically distant FIB-4 scores should be incorporated into future SLD risk assessments [2, 31, 32]. As our data show, FIB-4 risk can change over time and 32% of the patients in our cohort had a subsequently higher-risk FIB-4 score compared to their index value. The EHR can facilitate FIB-4 score calculation with every liver test panel ordered and provide context with regard to changes in FIB-4 risk over time. Leveraging the EHR to alert primary care clinicians to the increased risk of future SLD may help to improve CLD diagnosis and optimize linkage to specialty hepatology care for patients at greatest risk of future severe liver outcomes [33].
Studies evaluating the association between FIB-4 risk and SLD outcomes have thus far focused on only 1 or 2 FIB-4 scores per patient [12–14]. As our results demonstrate, patients in usual primary care often have the inputs for many more FIB-4 risk assessments and this study is novel in its treatment of FIB-4 fibrosis risk as a time-varying covariate. Here, FIB-4 risk is measured at baseline, but its relationship with the SLD outcome is likely to change during follow-up. This change in FIB-4’s impact on SLD risk over time may not fulfill the assumptions of proportional hazards and may benefit from a methodology that handles FIB-4 as a time-varying covariate in a time-dependent Cox regression model [15, 34, 35]. This time-dependent Cox model allows for analysis of covariates that change with time by comparing the current covariate value (i.e. FIB-4 fibrosis risk) of the patient with the outcome (i.e. SLD) to the current covariate values of the other patients at risk at each event time [15, 34, 36]. We apply this method in this study and again demonstrate a strong association between high-risk FIB-4 scores and SLD outcomes.
These data provide evidence that multiple FIB-4 scores per patient are available in the course of usual primary care and that a single occurrence of a high-risk FIB-4 value (≥ 2.67) may portend future severe liver outcomes. To explore this idea further, we calculated the performance characteristics of a FIB-4 score ≥ 2.67 for detecting SLD using each patient’s maximum FIB-4 value during the study period. Of the patients that had at least one high-risk score at any point during the study period, nearly 10% (9.8%) had an SLD event, whereas never possessing a high-risk FIB-4 had a negative predictive value for SLD of 98.9%. Occurrences of high-risk FIB-4’s in the primary care setting, whether present or past, should prompt heightened vigilance for CLD and CLD progression to severe liver disease outcomes. When the etiology of CLD is known, indeterminate- and high-risk FIB-4 scores should prompt confirmatory advanced fibrosis risk assessment with liver stiffness measurements (e.g. vibration-controlled elastography) or serologic testing (e.g. Enhanced Liver Fibrosis [ELF] test), in accordance with specialty society guidelines [31, 37]. Known CLD and advanced fibrosis should also prompt referral to a hepatologist. Future research will need to consider the cost-benefit ratio of routine liver chemistry and platelet testing in primary care patients for the purposes of CLD screening. A multi-center, pragmatic trial of current primary care patients, with variability in the frequency of and intervals between these lab results, would help to answer questions on FIB-4 testing in primary care.
We recognize limitations in this work. First, our outcomes rely upon ICD-9/10 codes in the electronic health record which have historically lacked sensitivity for detecting clinical conditions [20]. However, we attempted to mitigate this concern by using composites of ICD-9/10 codes for cirrhosis, its complications, hepatocellular carcinoma, and need for liver transplantation previously studied in the literature and found to improve case finding sensitivity while preserving relatively high specificity [16–18]. Also, as a retrospective study, we do not have perfect follow-up data for our entire cohort and it is likely that some patients left the health system prior to the end of the study period. To mitigate this potential bias, we assumed all such patients did not experience an SLD event in an attempt to provide conservative estimates of hazard ratios in our Cox regression models. Next, our primary predictor variable was a FIB-4 score that was calculated from labs not necessarily drawn for monitoring or assessing liver disease. We also did not include known CLD as a predictor variable in our models because CLD (especially non-alcoholic fatty liver disease) is underdiagnosed in primary care and we desired to study the relationship of FIB-4 with SLD even when knowledge of CLD is not yet known. Alcohol use as a risk was not included because the data are unreliably collected and are often not available within the EHR. We included diagnoses of NAFLD, alcohol use disorder, and alcohol-related liver disease in sensitivity analysis. Table 1 included the CLD diagnoses to better contextualize the patient cohort. Also, this cohort only included patients with at least 2 sets of lab inputs to calculate at least 2 FIB-4 scores. These inclusion criteria pose threats from selection bias, as the cohort possibly includes patients more closely monitored with lab tests, more likely to have CLD, and more likely to have suffered an acute liver injury. A prospective study with uniform FIB-4 testing would address this limitation. Lastly, this data come from a single institution and has a high burden of obesity, diabetes mellitus, and other risk factors for NAFLD which could threaten the generalizability of the findings.
FIB-4 values can be frequently calculated within primary care, FIB-4 advanced fibrosis risk assessments change over time, and high-risk FIB-4 scores (≥ 2.67) are strongly associated with severe liver disease outcomes. FIB-4 calculation and incorporation of historical context within the EHR could provide useful clinical signals to improve liver disease diagnosis and outcomes in primary care.
Supplementary Material
Grant Support:
National Institute of Diabetes and Digestive and Kidney Diseases (NIH/NIDDK K23DK118200 PI: Schreiner; R03DK129558 PI: Schreiner; P30DK123704 PI: Rockey). This project was also supported by the South Carolina Clinical & Translational Research Institute with an academic home at the Medical University of South Carolina CTSA National Center for Advancing Translational Sciences (NCATS) of the National Institutes of Health (NIH) under UL1 TR001450.
Abbreviations
- ALT
alanine aminotransferase
- AST
aspartate aminotransferase
- CLD
chronic liver disease
- EHR
electronic health record
- FIB-4
Fibrosis-4 Index
- HCC
hepatocellular carcinoma
- HR
hazard ratio
- IQR
interquartile range
- NPV
negative predictive value
- PPV
positive predictive value
- SLD
severe liver disease
Footnotes
Authorship: All contributors to this work appear in the author byline. All named authors contributed to the study design, data interpretation, manuscript editing, and submission decision. Dr. Schreiner and Ms. Zhang led the writing of the original draft. Dr. Gebregziabher and Ms. Zhang developed the analysis plan, and Dr. Gebregziabher and Ms. Zhang performed the analyses. Ms. Zhang, Mr. Marsden, and Dr. Schreiner performed the data collection.
Disclosures: All authors report no conflicts of interest with this work.
Data Availability
The study protocol and individual participant data contributing to these reported results reported in this article will be made available to investigators whose proposed use of the data has been approved by an independent review committee. Data will only be made available after de-identification (text, tables, figures, and appendices) and compliance with the Health Insurance Portability and Accountability Act of 1996 and the Institutional Review Board at MUSC is assured.
References
- 1.Vallet-Pichard A, Mallet V, Nalpas B, et al. FIB-4: an inexpensive and accurate marker of fibrosis in HCV infection. comparison with liver biopsy and fibrotest. Hepatology. 2007;46:32–36. [DOI] [PubMed] [Google Scholar]
- 2.European Association for the Study of the L, List of panel m, Berzigotti A, et al. Easl Clinical Practice Guidelines (Cpgs) On Non-Invasive Tests For Evaluation Of Liver Disease Severity And Prognosis- 2020 Update. J Hepatol. 2021. [DOI] [PubMed] [Google Scholar]
- 3.Sterling RKD-R A; Patel K; Asrani SK; Alsawas M; Dranoff J; Fiel MI; Murad MH; Leung DH; Levine D; Taddei TH; Taoli B; Rockey DC; AASLD Guidelines Committee. Executive Summary: Non-invasive Liver Disease Assessments (NILDA) of hepatic fibrosis, steatosis, and portal hypertension: 2021 Practice Guidelines and Guidance from teh American Association for the Study of Liver Diseases. In: Committee AG, ed. The Liver Meeting. Anaheim, CA: AASLD; 2021. [Google Scholar]
- 4.Hagstrom H, Nasr P, Ekstedt M, et al. Fibrosis stage but not NASH predicts mortality and time to development of severe liver disease in biopsy-proven NAFLD. J Hepatol. 2017;67:1265–1273. [DOI] [PubMed] [Google Scholar]
- 5.Sanyal AJ, Van Natta ML, Clark J, et al. Prospective Study of Outcomes in Adults with Nonalcoholic Fatty Liver Disease. N Engl J Med. 2021;385:1559–1569. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Lackner C and Tiniakos D. Fibrosis and alcohol-related liver disease. J Hepatol. 2019;70:294–304. [DOI] [PubMed] [Google Scholar]
- 7.Naveau S, Gaude G, Asnacios A, et al. Diagnostic and prognostic values of noninvasive biomarkers of fibrosis in patients with alcoholic liver disease. Hepatology. 2009;49:97–105. [DOI] [PubMed] [Google Scholar]
- 8.Rockey DC, Bell PD and Hill JA. Fibrosis--a common pathway to organ injury and failure. N Engl J Med. 2015;372:1138–1149. [DOI] [PubMed] [Google Scholar]
- 9.Younossi ZM and Henry L. Are Noninvasive Scoring Systems for Persons With Chronic Liver Disease Ready for Prime Time? Gastroenterology. 2020;158:40–42. [DOI] [PubMed] [Google Scholar]
- 10.Rockey DC. Noninvasive assessment of liver fibrosis and portal hypertension with transient elastography. Gastroenterology. 2008;134:8–14. [DOI] [PubMed] [Google Scholar]
- 11.Schreiner AD, Zhang J, Durkalski-Mauldin V, et al. Advanced liver fibrosis and the metabolic syndrome in a primary care setting. Diabetes Metab Res Rev. 2021;37:e3452. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Hagstrom H, Talback M, Andreasson A, et al. Repeated FIB-4 measurements can help identify individuals at risk of severe liver disease. J Hepatol. 2020;73:1023–1029. [DOI] [PubMed] [Google Scholar]
- 13.Schreiner AD, Moran WP, Zhang J, et al. The Association of Fibrosis-4 Index Scores with Severe Liver Outcomes in Primary Care. J Gen Intern Med. 2022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Schreiner AD, Zhang J, Moran WP, et al. FIB-4 and incident severe liver outcomes in patients with undiagnosed chronic liver disease: A Fine-Gray competing risk analysis. Liver Int. 2022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.K JD and P RL. The Statistical Analysis of Failure Time Data. Hoboken, NJ, USA: John Wiley & Sons, Inc.; 2002. [Google Scholar]
- 16.Mapakshi S, Kramer JR, Richardson P, et al. Positive Predictive Value of International Classification of Diseases, 10th Revision, Codes for Cirrhosis and Its Related Complications. Clin Gastroenterol Hepatol. 2018;16:1677–1678. [DOI] [PubMed] [Google Scholar]
- 17.Nehra MS, Ma Y, Clark C, et al. Use of administrative claims data for identifying patients with cirrhosis. J Clin Gastroenterol. 2013;47:e50–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Hagström H, Adams LA, Allen AM, et al. Administrative Coding in Electronic Health Care Record-Based Research of NAFLD: An Expert Panel Consensus Statement. Hepatology. 2021;74:474–482. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Shah AG, Lydecker A, Murray K, et al. Comparison of noninvasive markers of fibrosis in patients with nonalcoholic fatty liver disease. Clin Gastroenterol Hepatol. 2009;7:1104–1112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Quan H, Li B, Saunders LD, et al. Assessing validity of ICD-9-CM and ICD-10 administrative data in recording clinical conditions in a unique dually coded database. Health Serv Res. 2008;43:1424–1441. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43:1130–1139. [DOI] [PubMed] [Google Scholar]
- 22.Chalasani N, Younossi Z, Lavine JE, et al. The diagnosis and management of nonalcoholic fatty liver disease: Practice guidance from the American Association for the Study of Liver Diseases. Hepatology. 2018;67:328–357. [DOI] [PubMed] [Google Scholar]
- 23.Chu CM and Liaw YF. Incidence and risk factors of progression to cirrhosis in inactive carriers of hepatitis B virus. Am J Gastroenterol. 2009;104:1693–1699. [DOI] [PubMed] [Google Scholar]
- 24.de Torres M and Poynard T. Risk factors for liver fibrosis progression in patients with chronic hepatitis C. Ann Hepatol. 2003;2:5–11. [PubMed] [Google Scholar]
- 25.Hedenstierna M, Nangarhari A, Weiland O, et al. Diabetes and Cirrhosis Are Risk Factors for Hepatocellular Carcinoma After Successful Treatment of Chronic Hepatitis C. Clin Infect Dis. 2016;63:723–729. [DOI] [PubMed] [Google Scholar]
- 26.Kalbfleisch JDPR. The Statistical Analysis of Failure Time Data. New York, NY: John Wiley & Sons; 2002. [Google Scholar]
- 27.McPherson S, Hardy T, Dufour JF, et al. Age as a Confounding Factor for the Accurate Non-Invasive Diagnosis of Advanced NAFLD Fibrosis. Am J Gastroenterol. 2017;112:740–751. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Huang DQ, Mathurin P, Cortez-Pinto H, et al. Global epidemiology of alcohol-associated cirrhosis and HCC: trends, projections and risk factors. Nat Rev Gastroenterol Hepatol. 2023;20:37–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Roerecke M, Vafaei A, Hasan OSM, et al. Alcohol Consumption and Risk of Liver Cirrhosis: A Systematic Review and Meta-Analysis. Am J Gastroenterol. 2019;114:1574–1586. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Schreiner AD, Zhang J, Durkalski-Mauldin V, et al. Advanced Liver Fibrosis and the Metabolic Syndrome in a Primary Care Setting. Diabetes Metab Res Rev. 2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Rinella ME, Neuschwander-Tetri BA, Siddiqui MS, et al. AASLD practice guidance on the clinical assessment and management of nonalcoholic fatty liver disease. Hepatology. 2023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Kanwal F, Shubrook JH, Adams LA, et al. Clinical Care Pathway for the Risk Stratification and Management of Patients With Nonalcoholic Fatty Liver Disease. Gastroenterology. 2021;161:1657–1669. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Srivastava A, Gailer R, Tanwar S, et al. Prospective evaluation of a primary care referral pathway for patients with non-alcoholic fatty liver disease. J Hepatol. 2019;71:371–378. [DOI] [PubMed] [Google Scholar]
- 34.Zhang Z, Reinikainen J, Adeleke KA, et al. Time-varying covariates and coefficients in Cox regression models. Ann Transl Med. 2018;6:121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Bellera CA, MacGrogan G, Debled M, et al. Variables with time-varying effects and the Cox model: some statistical concepts illustrated with a prognostic factor study in breast cancer. BMC Med Res Methodol. 2010;10:20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Therneau T and Grambsch P. Modelling Survival Data: Extending the Cox Model. NY, USA: Springer; 2000. [Google Scholar]
- 37. .Hepatitis C Guidance 2018 Update: AASLD-IDSA Recommendations for Testing, Managing, and Treating Hepatitis C Virus Infection. Clin Infect Dis. 2018;67:1477–1492. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The study protocol and individual participant data contributing to these reported results reported in this article will be made available to investigators whose proposed use of the data has been approved by an independent review committee. Data will only be made available after de-identification (text, tables, figures, and appendices) and compliance with the Health Insurance Portability and Accountability Act of 1996 and the Institutional Review Board at MUSC is assured.