Indirect serum biomarkers perform sub optimally in screening for significant liver fibrosis among HIV-infected and uninfected adults in Uganda

Introduction Indirect serum bio-markers present an acceptable noninvasive and cheap alternative for screening of significant liver fibrosis (SLF). Evaluation of their use in resource limited settings is important to determine their utility. Methods We conducted a cross sectional study among 520 HIV infected and HIV uninfected adults attending care clinics in Kampala Uganda. Presence of SLF was determined using Fibroscan® liver stiffness measurement of ≥7.2KPa. The diagostic value of indirect serum bio-markers for diagnosis of SLF was evaluated using the area under the receiver operating characteristics curve (AUROC) using Fibroscan® as gold standard. Results Overall AUROC values for Age Platelet Index (API), Aspartate to Alanine Ratio (AAR), AST-to-Platelet Ratio Index (APRI), Fibrosis Index based on 4 Factors (FIB-4) and Gamma glutamyl transferase to Platelet Ratio Index (GPR) were 0.52, 0.49, 0.55, 0.55 and 0.54 respectively. Among HIV-infected participants AUROC values were slightly improved at predicting presence of SLF but still under 70%. Conclusion Despite APRI and FIB-4 being more likely to identify participants with SLF, the overall diagnostic value of all serum bio-markers was poor with and without stratification by HIV status. We recommend the use of Fibroscan® technology as more accurate non-invasive diagnostic method for screening of SLF.


Introduction
Mortality and morbidity resulting from significant liver fibrosis (SLF) is becoming more prevalent 1, 2 . Regardless of the cause, chronic liver injury leads to SLF which is the primary risk factor for complications, such as cirrhosis and liver cancer 1, 3 . Significant liver fibrosis occurs when there is excess deposition of extracellular matrix by the stellate cells of the liver as a response to chronic injury 2,4 . As changes alter the structure and or function of the liver, assessment of the liver structure and function is used to clinically detect and gauge the extent of liver fibrosis. Evaluation of SLF is critical in clinical practice to prognosticate patients, make treatment choices, as well as monitor response to therapy and disease progression [3][4][5] . The gold standard evaluation for SLF traditionally is liver biopsy, but it is limited by sampling and observational error 6,7 . In addition, it requires special expertise to be performed and for pathological specimen to be read; all of which limit its use within resource limited settings. By virtual of its invasive nature, it is not readily acceptable by patients 7 .
Recent advancements in diagnostic methods for SLF advocate for the use of non-invasive techniques. Non-invasive techniques for diagnosis of SLF are broadly categorized as imaging or blood based techniques, some of which usually have standardized interpretation requiring less technical expertise. Due to the ease of their administration and reduced patient risk they can be performed repetitively, an advantage in monitoring disease progression and response to therapy 8 . Imaging techniques include transient elastograpy (Fibros-can®) which has been shown to be comparable in performance to liver biopsy for most liver disease conditions 5,8,9 . The volume of the liver explored by Fibroscan® is at least 100 times more than that explored by liver biopsy and results are immediate 10 . However Fibroscan® machines remain costly for most low and middle income countries which limits their use to research settings rather than widespread clinical use 11 . Indirect serum bio-markers present a more feasible option given that they are derived from cheap and routinely performed tests and standardized indices are easily interpretable by non-specialized clinical personnel aiding decentralized use. 7,8,12 . Given the limitations in use of liver biopsy and or Fibro-scan® in low resource settings, indirect serum bio-marker indices may suffice as the next available option to screen for SLF to facilitate early referral and management of liver disease.
Early screening purposes to prevent and halt progression of SLF to complicated forms of liver disease for which tertiary management is very costly and most times out of reach. In addition, within the sub Saharan African region, the limited studies on the validation of these indirect serum biomarkers in an African population, have evaluated their use in single disease conditions, mainly viral hepatitis, and their remains disparity in their performance. There is strong recommendation for more similar studies in the region to inform on threshold values for these indirect serum bio-markers when applied to various liver disease entities 11,13 . Furthermore, available recommendations for the use of indirect serum bio-markers by World Health Organization in locations with limited access to the Fibroscan® mainly focuses on use of APRI among persons with chronic hepatitis B yet there is an increase in risk factor profile for chronic liver disease in the region presently for which the use of APRI may not necessary be suited.
We therefore conducted a cross sectional study to determine the diagnostic value and level of agreement for the diagnosis of SLF for five specified indirect serum bio-markers (API, APRI, AAR GPR and FIB-4) using Fibroscan® as gold standard among adults attending care in urban Uganda.

Study design and population
We conducted a cross sectional study between January 2015 and March 2020 involving 260 HIV-infected and 260 HIV-uninfected adults attending out-patient clinics within the National referral hospital in Kampala. The HIV-infected and HIV-uninfected patients were identified from the Adults Infectious Diseases Clinic (AIDC) and Ear Nose and Throat (ENT) out-patient clinics respectively. The AIDC is a model center of excellence for the care and management of HIV-infected adults. The prevalence of chronic hepatitis B among patients attending the AIDC is estimated at 11% and sero-prevalence of hepatitis C at 2%. The description of the site and prevalence of other risk factors for liver disease such as tobacco and alcohol use is well described elsewhere [14][15][16] . The ENT clinic is government supported sub specialized surgical out-patient clinic that receives stable patients seeking related services mainly as self-referrals as well as consultative referrals within Mulago national referral hospital, Kampala, Uganda.

Participant enrolment HIV-infected participant enrolment
We sampled the HIV-infected participants from already collected data from a prospective study, the HIV and Hepatocellular carcinoma in Uganda (H2U) at the AIDC, a study that aims to determine the prevalence and correlates of pre-malignant liver cirrhosis. Using a randomly generated computer list, we selected 260 HIV-infected study participants from a pool of 2000 participants in the H2U database. We included participants with records of liver stiffness measurement (LSM) done by Fibroscan®. Pregnant women and patients with medical implants were excluded from having Fibroscan done.

HIV-uninfected participants
Using compiled information on age and gender of the HIV-infected participants from the AIDC, we used these demographic characteristics to enroll the HIV-uninfected participants at the ENT clinic, matched by gender and age group. We accessed a daily clinic register on attendance, identified and approached potential participants requesting initial consent to participate in the study and screen for HIV serology in accordance with the national guidelines 17 . Patients that tested HIV sero-negative were then taken through detailed informed consent. Adults who had no known history of chronic liver disease were eligible to participate in the study. Pregnant women and persons with implanted medical devices were excluded from participation.

Study Procedures Questionnaire
The information collected at both clinics included socio-demographic information (age, gender); lifestyle habits (on self-reported use of alcohol, tobacco and herbal medicines); anthropometric data (weight and height) and clinical information (viral hepatitis serology, biochemical data, ART related information and drug history).

Physical measurements
All anthropometric measurements were done using Seca 761 mechanical scale to the nearest 1kg and a Seca Leicester stadiometer to the nearest 0.1 cm for weight and height respectively. Laboratory testing: All blood testing was done at the Central laboratory at the Infectious Disease Institute in the Makerere University John Hopkins University (MU-JHU) laboratory, a CAP certified laboratory. Complete blood count was done using a Beckman Coulter AcT 5diff analyzer and liver biochemistry was analyzed using HitachiCobas C311. Hepatitis B serology was performed using an enzyme immunoassay (Monolisa HBsAg Ultra 3; Bio-Rad). Hepatitis C antibody testing was done using 3 rd generation enzyme immunoassay (Bio-rad Monolisa Anti-HCV PLUS). Transient elastography (Fibroscan®): Liver stiffness measurements (LSM) for participants from both clinics were done using Fibroscan® (Echosens, Paris France) and the median result of ten Fibroscan® readings with the IQR <30% and accuracy of 60% and above was considered as the final result11. At the time of enrolment for the HIV-infected participants under the H2U study, the only available Fibroscan® probe was the M-probe. In contrast during the enrolment of the HIV-uninfected participants from the ENT clinic the Fibroscan® was upgraded to have both an M and XL probe. The XL probe being suitable for LSM in persons with body mass index (BMI) ≥30 and the M-probe being suitable for persons with BMI less than 30. Participants were not asked to fast prior to the Fibroscan® procedure.

Computing the indirect serum biomarker indices
Using parameters from the liver biochemistry and platelet count, indices were computed for the selected serum bio-markers (table 1). The indirect serum bio-markers included the API, AAR, APRI, FIB-4 and GPR. The outcome SLF was assessed as a binary outcome for both Fibroscan® and the selected indirect serum bio-markers. On Fibrosca® SLF was defined as having a LSM of >7.1KPa18. The upper limit of normal for aspartate transaminase (AST) and gamma glutamyl transferase (GGT) used were 32 U/L and 55 U/L respectively.

Data Analyses
For the descriptive statistics, continuous variables were presented as means and medians. Categorical data were described as proportions and comparisons between two proportions done using chi square test. The diagnostic value of API, AAR, APRI, FIB-4, GPR and Fibros-can® was evaluated using receiver operating characteristic (ROC) analysis. The sensitivity (Se), specificity (Sp), positive predictive value (PPV), negative predictive value (NPV), diagnostic accuracy (DA), positive likelihood ratio (LR+) and negative likelihood ratio (LR-) were calculated based on the cut-off points as mentioned above.
Significance was set at p-value <0.05. Definition of the diagnostic value of the indirect serum bio-markers was as follows; excellent for AUROC value ≥0.9, good for AUROC value 0.80-0.89, moderate for 0.70-0.79 and poor for AUROC value <0.70. Agreement testing for categorical data comparing two raters was done for Fibroscan® and each indirect serum bio-marker. Cohen's Kappa statistic for agreement testing was used where the prevalence of SLF was ≥10% for each rater. In instances where the prevalence of SLF was lower than 10% for anyone rater, a prevalence adjusted bias adjusted kappa was applied. Interpretation of the Kappa statistic was done as shown below (table 2) 21 . Four of the HIV negative participants were excluded from the analyses because of an invalid LSM.

Results
There were more female participants in the study (56% vs 44%). The mean age was 44 years (±10.3) and this was similar between both groups of participants. HIV-uninfected participants had a higher mean BMI compared to the HIV-infected participants (27 Vs 23; p<0.001). The overall sero-prevalence of hepatitis C was 3% and this was similar in both groups. Chronic hepatitis B was five times more prevalent among the HIV-infected partic-ipants (14% vs 3%; p<0.001). HIV-uninfected participants were more likely to have used herbal drug medicines (49% vs25%; p<0.001), however there were no differences in the use of alcohol and tobacco products. HIV-infected participants had a higher mean serum alanine transaminase (24mmol/l vs 21mmol/l; p0.011).

Discussion
Overall all the indirect serum bio-markers (API, AAR, APRI, FIB-4 and GPR) had poor diagnostic value in predicting the presence of SLF having AUROC values less than 0.7 even when stratified by HIV status. Indirect serum bio-markers that demonstrated better diagnostic value overall included APRI and FIB-4 and similar among the HIV-infected participants. None of the indirect serum bio-markers significantly performed well among the HIV-uninfected participants. Our findings are similar to other studies in the region conducted among patients with chronic hepatitis B infection that also found low ability to predict presence of SLF using APRI but good performance of GPR 18,22 . The selection of study participants for the latter studies excluded those with conditions known to alter the serum GGT levels, such as alcohol consumption, co-infection with HIV/AIDS and HCV.
Our inclusion of these categories of participants may have affected the accuracy of GPR 23 . In addition, other studies seem to suggest that there is poor correlation between GPR and Fibroscan® for evaluation of SLF among HBV/HIV patients in sub Saharan Africa and this may further explained its poor diagnostic performance among our HIV-infected participants 23 . Our findings were also similar to studies conducted among western populations focused on alcoholic liver disease that showed APRI to have poor diagnostic value predicting SLF 24 .
In contrast, other studies still within the region conducted among treatment naïve patients with chronic hepatitis B infection and patients with chronic hepatitis C infection showed APRI, FIB-4 and GPR to have moderate to good diagnostic value in predicting SLF 11 . We had low sample sizes for patients with HBV and HCV mono-infection and this could have possibly affected our estimates. In addition, we screened for HCV using antibody testing whose presence does not necessarily indicate active infection.
We found a high negative predictive value and low positive predictive value for all the indirect serum bio-markers similar to other studies 11,13,18 . For all the serum bio-markers the inter-observer agreement with Fibroscan® for the diagnosis of significant liver fibrosis was poor and for some almost by chance.

Limitations
Our study had a number of limitations that included, restraints in finances that could not allow for liver biopsy to be used as gold standard. We also did not take fasting LSM that could have led to an over estimation of the burden of SLF. Our study recruited participants from a tertiary health care unit which may not be representative of the general population. We had a limited total number of persons identified with cirrhosis on Fibroscan® (cutoff LSM≥13KPa) and were unable to evaluate the performance of the indirect serum bio-markers for diagnosis of liver cirrhosis to compare with other studies in the region. Screening for hepatitis C was done using serology which often gives false positive results. Despite these limitations, our study evaluates a number of indirect serum bio-markers in the background of an interplay of several risk factors (that often never exist in isolation) for SLF including HIV infection, which is prevalent in the region. Our study adds to the existent evidence on the disparity of the performance of indirect serum bio-markers when applied in our setting for the evaluation of SLF.

Conclusion
The indirect serum bio-markers (API, AAR, APRI, FIB-4 and GPR) had unsatisfactory performance in identifying participants with SLF among HIV-infected and uninfected adults in urban Uganda. Given their poor diagnostic value and varied performance, we recommend Fibros-can® be prioritized as a more accurate non-invasive technique for screening of SLF. We also recommend further studies evaluating the value of repeated testing using indirect serum bio-markers as part of a screening algorithm and/or alternate applications such as in disease prognostication.