The Metformin Trial (META) is a phase III randomised double-blind placebo-controlled trial aimed at determining the effectiveness of metformin in preventing diabetes among PLHIV who have prediabetes. The five study sites for this trial are located in the Dar es Salaam area, and include three public regional referral hospitals (Mwananyamala, Temeke, Amana), a public district hospital (Mnazi Moja) and a private not-for-profit hospital (Shree Hindu Mandal). This is a cross-sectional study using baseline data from the META Phase III Trial cohort to evaluate liver fibrosis risk using FIB-4, MAF-5, NFS, and LiverRisk scores and to estimate the risk of incident cirrhosis within ten years using CORE. Clinical trial registration: This process evaluation is listed on the ClinicalTrials.gov registry (registration number: NCT06743698). The META trial is listed on the International Standard Randomised Controlled Trial Number (ISRCTN) registry (registration number: ISRCTN77382043).

Inclusion criteria for the patients recruited into this study were: to be an adult, living with HIV, having been on ART for at least six months and considered stable with a viral load test in the previous 12 months that was below 1,000 copies/ml, have impaired fasting glucose according to the Oral Glucose Tolerance Test (OGTT) prediabetes criteria (≥ 7.8-11.1 mmol/L; where fasting blood glucose was ≥ 7.0 and < 11.0 mmol/L), and planning to remain in the area for more than 12 months. Patients were excluded if they were pregnant, had participated in the META Phase II study or had a known hypersensitivity to metformin or any excipients associated with the preparation, were known to have renal disease, had clinical evidence of liver disease or congestive heart failure requiring pharmacological treatment, had evidence of alcoholism or acute alcohol intoxication, had any form of acute metabolic acidosis including lactic acidosis or diabetic ketoacidosis or other acute conditions with the potential to alter renal function or to cause tissue hypoxia, or had any other acute conditions requiring hospital admission or emergency clinical intervention, including blood pressure > 180/110 mmHg, haemoglobin < 6.5 g/dL for women or haemoglobin < 7.0 g/dL for men (grade 3); white cell count < 1.5 × 10⁹ cells/mm³ (grade 3) and any baseline liver function derangements at grade 4 according to Division of AIDS (DAIDS) table for grading the severity of adult and pediatric adverse events criteria^[18]. Researchers provided the individuals invited to participate with information about the study and asked each to give their consent in writing before participation.

Participants who consented to be screened for the trial were asked to return to the clinic having fasted, i.e., not eaten or drunk anything except water for at least 8 h before their appointment. At screening, a morning venepuncture sample was collected on arrival at the clinic and used for measurement of fasting blood glucose, HbA1c point-of-care test, and rapid malaria point-of-care antigen test (CareStart-Malaria pf/PAN[HRP2/plDH] Ag Combo RDT, USA). Participants were given a 75 g anhydrous glucose solution dissolved in 300 mL non-carbonated water (Rapilose oral glucose tolerance test solution, Penlan Healthcare, UK), and a 2-h post-glucose-load capillary blood glucose measurement was taken via finger prick. To determine blood glucose levels, a point-of-care test (Hemocue Glucose 201 RT-Hemocue AB, Sweden) was performed immediately following venepuncture and finger prick. Baseline blood biochemistry and haematology testing was conducted on the morning venepuncture sample from those enrolled in the trial following screening.

We calculated the FIB-4, MAF-5, NFS, LiverRisk, and CORE scores according to published or derived formulas^[9,13-16]. Age was measured in years, aspartate aminotransferase (AST), alanine aminotransferase (ALT), and Gamma-glutamyl Transferase (GGT) were measured in IU/L, and platelet count was measured in 10⁹/L. For MAF-5 and NFS calculations, which use the presence of diagnosed diabetes, HbA1c, or fasting blood glucose thresholds, we used fasting blood glucose ≥ 5.6 mmol/L for MAF-5 and ≥ 7.0 mmol/L for NFS or HbA1c ≥ 5.7% (39 mmol/mol) for both scores^[13,14]. Waist circumference (WC) was measured in centimeters and body mass index (BMI) was calculated as the ratio of weight in kilograms and height in meters squared. Ten participants (0.6%) did not have requisite data to calculate the FIB-4 score, nor 326 (19.3%) for MAF-5, nor 15 (0.9%) for NFS, nor 14 (0.8%) for LiverRisk, nor 2 (0.1%) for CORE.

Following current international clinical guidelines, we categorized FIB-4 scores into risk strata based on age-specific thresholds^[10]. For participants aged younger than 65 years, scores were classified as low risk (< 1.30), intermediate risk (1.30-2.67), and high risk (> 2.67). For participants aged 65 years and older, the corresponding cut-offs were low risk (< 2.00), intermediate risk (2.00-2.67), and high risk (> 2.67). These categorical variables were then labelled as “low risk”, “intermediate risk”, and “high risk” for descriptive and inferential analyses. MAF-5 score categories are based on the published cut-offs of low fibrosis risk (MAF-5 < 0) and high fibrosis risk (MAF-5 ≥ 1), with an indeterminate risk profile being assigned to scores between these values. NFS categories follow validated cut-offs: < -1.455 = low risk, -1.455 to 0.676 = indeterminate risk, and > 0.676 = high risk of advanced fibrosis. Cutoffs for LiverRisk (N = 1,677) are: minimal-risk group (< 6), low-risk group (6 to < 10), medium-risk group (10 to < 15), and high-risk group (≥ 15)^[15,19,20]. No published cut-off exists for CORE; we used a 1% cut-off, which would likely be an approach with a high sensitivity to detect persons that will develop cirrhosis while maintaining a high specificity^[21]. Likelihood ratio chi-squared test and Fisher’s exact test were used to test significance between scores and sex, which can only be interpreted as exploratory, given the female-to-male imbalance among participants. Data processing and analyses for FIB-4, MAF-5, NFS, and LiverRisk were conducted using Stata (version 16.1); CORE analyses were performed in R (version 2022.12.0) using the “rms” package (version 8.1-1).

Ethical approvals for the study were obtained from the Liverpool School of Tropical Medicine Research Ethics Committee (20-089) and the National Institute for Medical Research Ethics committee in Tanzania (NIMR/HQ/R.8a/Vol. IX/3613), and subsequently from the UCL Research Ethics Committee. The study is compliant with the Declaration of Helsinki. Written informed consent for participation in the trial was obtained prior to any data collection. During the consent process, all participants were provided with written information about the study; which was also explained verbally, and they were informed that their participation was voluntary and that they may withdraw from participation at any time without penalty.

Assessment of liver fibrosis risk varied substantially across non-invasive tools, suggesting that the choice of scoring method alone could materially alter how many patients are flagged for further evaluation within HIV care programs. Among the 1,691 participants enrolled in the META Trial, the majority were female (75.0%) and nearly all identified as black (99.2%, Table 2). The mean age of participants was 49.4 years (SD: 9.4), and on average, individuals had been living with HIV for 12.8 years, accounting for approximately 26.4% of their lifetime. Metabolic indicators showed evidence of elevated risk: the median BMI was 26.3 kg/m² and the median WC was 91 cm, both consistent with overweight and central adiposity. N = 316 were missing data for WC. Glycaemic markers were generally within normal ranges, with HbA1c (N = 20 missing) and fasting blood glucose (N = 0 missing) reporting medians of 5.7% and 6.1 mmol/L, respectively. Liver enzymes [i.e., ALT (N = 1 missing), AST (N = 2 missing), and GGT (N = 1 missing)] were also mostly within normal limits (medians: 18, 24, and 24 IU/L, respectively) but showed wide ranges [interquartile range (IQR): 14-23, 20-29, and 17-34 IU/L, respectively] [Table 2], indicating that a subset may have underlying hepatic or metabolic dysfunction. Additionally, there were N = 6 and N = 9 missing values for albumin and platelet count, respectively.

Table 3 summarizes the distribution of NIT scores in the study population, reporting the mean (SD), median, and range, along with the IQR, for each score. Assessment of liver fibrosis risk using non-invasive scores revealed substantial variation in risk classification across tools [Table 4]. The MAF-5 score identified a much higher proportion of participants (50.26%) as high risk than any other scale: 3.0% for FIB-4, 8.3% for NFS, and 0.1% for LiverRisk. CORE estimated that 5.8% of the sample had a 1% risk or greater to develop cirrhosis or have a cirrhosis-related event within ten years. The agreement among the four liver risk scores was generally poor, with only a fair level of concordance observed between FIB-4 and NFS (κ = 0.29), while all other interrater comparisons showed slight to no agreement [Table 5 and Supplementary Table 1].

Sex-stratified analyses showed significant differences between females and males for the FIB-4, LiverRisk, and CORE scores, whereas MAF-5 and NFS did not vary by sex [Supplementary Table 2]. For FIB-4, females were more frequently classified as low risk compared with males (70% vs. 50%, P < 0.001). High-risk classifications were uncommon in both groups (3% in females; 4% in males). For MAF-5, there was no statistical difference by sex; 54.1% of males were categorized as high risk compared with 49.1% of females. LiverRisk demonstrated marked sex differences, with females more frequently classified as minimal risk compared with males (76% vs. 29%, P < 0.001). Comparatively, 11.8% of males and 3.8% of females had a 1% risk or higher of cirrhosis or related event incidence within ten years.

This study highlights the potential utility and limitations of non-invasive fibrosis scores for assessing liver disease risk among PLHIV with prediabetes in a high-burden, resource-limited setting in sub-Saharan Africa. While FIB-4 identified a majority of participants as low risk, scores such as MAF-5 and NFS revealed a larger proportion with suggested underlying advanced fibrosis, underscoring the importance of tool selection and interpretation. Better availability of modern fibrosis estimation methods such as vibration-controlled transient elastography is needed in similar low-resource settings. Age-adjusted thresholds, particularly for FIB-4, should be examined in aging HIV populations, and further validation is needed for African populations to avoid misclassification bias. For HIV programs in resource-limited settings, these findings suggest that careful selection of non-invasive fibrosis tools, and explicit decisions about acceptable trade-offs between sensitivity and specificity, may be as important as expanding access to advanced diagnostics. Moving forward, integrating liver health screening into HIV and metabolic healthcare alongside investigation of underlying risk factors and NIT utility in predicting LREs can help strengthen efforts to identify and manage MASLD/metabolic dysfunction-associated steatohepatitis (MASH) and other liver and metabolic complications in sub-Saharan Africa.