The liver plays a pivotal role in intermediary metabolism, biosynthesis, and pharmacotherapy. As such, chemical influences can result in detrimental effects. Anatomically, the liver is centrally placed, with loose capillary fenestrations, which makes it susceptible to chemical assault [1]. Epidemiological evidence shows that chemical-induced liver injury is a major cause of acute liver failure, fibrosis and cirrhosis [2,3]. In preclinical studies, drug-induced liver injury is a major reason why many interesting lead molecules and novel compounds fail to transcend beyond the laboratory. It is also the principal reason for withdrawal of market authorization of medicines [4]. Furthermore, with ever increasing prevalence of diabetes mellitus, obesity, and alcohol consumption, the burden of liver diseases is expected to increase [5]. Individuals with compromised liver may require dose adjustments or complete avoidance of some essential medicines which otherwise may have been appropriate to their medical condition

New chemicals products, synthetic or natural, are continually being screened to discover cures. Amidst the ethical concerns, in vivo models provide the most predictable outcome to the expected human experiences [6]. This places an extra responsibility on the researcher to develop ethically acceptable models that perfectly simulate the pathobiology of human disease. In experimental hepatic safety pharmacology, the Gold standard test is a costly procedure of examining the effects of new drugs on liver histoarchitecture, biosynthetic function as well as on effects on endoboitic and xenobiotic biotransforming enzymes. For credibility of findings, researchers produce photomicrographs for histopathological analysis [7]. Despite major advances in chemical detection and tissue staining techniques, haematoxyllin and eosin staining are widely employed especially in resource constrained settings [8]. As it has been reported in human liver fibrosis, accuracy in results inference is affected by sampling errors and variability in pathological interpretation [9]. In dissemination of results, researchers present photomicrographs which in their subjective opinion best represent their findings. The possibility of making a type 1 error in these instances could be underestimated.

Clinically, to address limitations with liver biopsies, many simple, direct or indirect, objective, non-invasive indexes were developed to estimate and diagnose liver damage [10-13]. However, these indexes are yet to be fully assimilated in preclinical hepatic toxicological studies. In hepatic fibrosis in human immunodeficiency virus and hepatitis C virus infections, Sterling et al., proposed that the ratio of magnitude of the product of age and aspartate transaminase levels to the product of platelet count and square root of alanine transaminase levels of a patient is indicative of stage of liver fibrosis [14]. Sterling et al., (2006) had argued that age, AST, ALT, and platelet are simple indicators of liver fibrosis.

A number of well controlled studies and scientific reviews have assessed and recommended the diagnostic performance of the FIB-4 index [15-18]. Since the FIB-4 index parameters are routinely measured in preclinical settings, we hypothesized that FIB-4 index could be a simple tool for preclinical hepatic toxicological studies. The advantage will be to reduce cost, reduce number of animals needed for experimentation, and provide a means of easily screening large number of chemicals for hepatic effects whilst minimizing sampling errors and variability in results interpretation of results.

Search strategy

The study relied on secondary data from published preclinical hepatic toxicological/safety pharmacological studies. Research databases including Google Scholar, Pubmed, Medline, Web of Science, and Researchgate were purposively searched to obtain in vivo preclinical data between the years of 2008 and 2019. Word combinations such as hepatoprotective, hepatotoxicity, hematology, rats, mice, rabbits, paracetamol, carbon tetrachloride, isoniazid, liver damage were used as search items to retrieve relevant publications. Abstracts and full texts were screened, and a decision made on each paper based on a predefined criteria. In addition, reference lists as well as citations of selected studies were searched to obtain other relevant papers. This was done until there was a clear saturation in search.

Types of study designs included

The initial search results were analyzed according to a predefined criteria:

• That the study should include a hematogram/leaucogram where thrombocyte count has been reported as mean ± SEM in tabular form not graphical.
• The study should include a hepatogram with liver enzymes, alanine transaminase (ALT) and aspartate transaminase (AST) reported as mean ± SEM in tabular form not graphical.

Types of studies excluded

• Preclinical and experimental toxicological/safety pharmacological studies with tabular presentations of hematograms or hepatograms were used. Studies that presented results graphically were not included because graphs may not have attached scales to help estimate the actual mean ± SEM.
• Preclinical toxicological/pharmacological studies which fell outside the study period of 2008 to 2019.

Treatment of data

The experimental design of each of the selected studies including doses, mean platelet counts, AST and ALT were extracted and transferred to Microsoft excel 2013 data sheet. All statistical analysis was done using Graph pad prism version 6. Pearson’s correlation coefficient (r) was used to estimate linear relationship between quantities. To determine the diagnostic value of FIB-4 in preclinical research, an evaluation was performed to determine the sensitivity and specificity based on Area Under the Receiver Operating Characteristic curve (AUROC) using APRI tool (AST to Platelet Ratio Index) computed from the studies as standard.

Testing for Linearity: Linear regression was used to analyze each of the selected studies. Pearson’s correlation coefficient (r) was used as a test for correlation. Runs test (Wald–Wolfowitz) was used to test for departure from linearity.

Theory

Therefore, 

Applying the equation y=mx + c, plot of mean platelet count against AST/√ALT will give AGE/FIB-4 as slope when the intercept c = 0

Estimating FIB-4 index for each dose level of the selected studies: Using the equation  and assuming the mean age of animals in each dose level as equal to 1, the FIB-4 was calculated for every dose level in all the studies using the mean platelet count, AST and ALT.

Estimating the interaction between the parameters: The relationship between platelet counts, AST and ALT was estimated using linear regression.

Assumptions

• The quantitative data and mean, was used to represent each dose level. To simplify calculation, the dispersion from the mean i.e. standard error of the mean was ignored.
• In order to use the formula, the age of all animals in the experimental design was made constant (k=1). This was because secondary data doesn’t include the age of animals in a group. Secondly, experimental induction of liver damage is rapid and occurs within a defined time period contrary to clinical liver damage which may be gradual and over a long period. This suggests that age may not be an important factor in a well-controlled and matched study.
• To test the sensitivity and specificity of FIB-4, we had to compare the FIB-4 value to the conclusions reached by the primary researchers at each dose level in the 25 studies. The individual dose levels were 148. To objectively accomplish this, the APRI tool was used to compute a value for each dose level, and this was subsequently matched with its corresponding FIB-4 value at the dose level.

Characteristics of selected studies

Twenty five (25) published preclinical studies with hematograms and hepatograms met the inclusion criteria. Eighty Four percent of the study (n= 21) used either Sprague – dawley or Wister rats. Twelve percent used rabbits as experimental animals and four percent used mice. Hepatotoxicity was induced with carbon tetrachloride, paracetamol, potassium dichromate, hydroxyapatite, isoniazid, methyl methanesulfonate, sodium fluoride or hepatotoxicant combinations such isoniazid plus rifampicin, cadmium plus paracetamol, mercury plus paracetamol, lead plus paracetamol (Table 1).

Linear analysis

In subjecting each experiment to linear regression analysis by plotting platelet count (dependent variable) against AST/√ALT (independent variable) a linear association was noted. To confirm whether or not the data had correlation with these parameters, Run’s test of randomness was employed. There was no significant deviation from linearity by the Run test for randomness for all the 25 studies (Table 1).

Estimating FIB-4 index

The FIB- 4 index

Using the formula FIB-4=  and assuming that the age of every animal in the study was constant i.e. 1, the FIB- 4 index was estimated for every dose level. There were 148 indexes obtained from the 25 studies. In 80% of the cases, the FIB-4 index of the toxicant control group was higher than FIB-4 index of naïve or unexposed group. As such, the ratio of the toxicant control FIB-4 to the naïve/vehicle FIB-4 was greater than 1 (Table 1).

In Eighty-five percent of studies (17 out of 20) FIB-4 index of hepatoprotective groups was similar to or less than that of untreated controls and different from the toxicant control. In all these instances the conclusion drawn from this study was in line with that of the primary authors/ researchers. For approximately twelve percent of the studies, the ratio of the FIB-4 toxicant to control was approximately equal to 1 i.e (0.8- 0.99). In 8 percent of studies (2/25) ratio of FIB-4 was below 0.5. (Table 1).

Consistently, the FIB-4 index was able to detect diverse hepatoactive chemicals presupposing that FIB-4 index was less chemical nature specific but rather more dose and duration dependent.

Relation between FIB-4 ratios and Pearson’s correlation coefficient

There was an association between the direction of the Pearson’s correlation coefficient (r) estimated from the linear regression and the strength and the size of the calculated ratio of FIB-4 (toxicant) to FIB-4 vehicle control. Consequently, a strong negative Pearson’s correlation coefficient correlated with a high calculated ratio. For the 20% of the study of which the estimated FIB-4 deviated from the expected, the Pearson’s correlation Coefficient (r) was weakly negative or positive (Table 1).

Sensitivity and specificity of FIB-4

To test the sensitivity and specificity of the FIB-4 index as a tool in preclinical toxicology, the APRI index was computed with the APRI tool for each dose level of the selected studies (n=148). This was then used as standard for determining the sensitivity of FIB-4 index in preclinical studies.

The Area Under the Receiver Operating Characteristics Curve (AUC – ROC/ AUROC) was 0.9698 ± 0.008989 (95% Cl 0.9522 to 0.9874 at p < 0.0001 ****). This translates to about 95-98% sensitivity and specificity as a diagnostic model (Figure 1a).

Using Area Under the Receiver Operating Characteristics Curve (AUC – ROC/ AUROC), the separability or discriminatory performance of FIB-4 index in predicting or separating hepatoprotective from hepatotoxic chemicals was between 51-67% (AUC – ROC/ AUROC 0.5959 ± 0.03899; 95% Cl 0.5194 to 0.6723; p= 0.01542) (Figure 1b).

Estimating the interaction between the parameters AST, ALT, Platelet

Liver enzymes: There was a positive linear autocorrelation between AST and √ALT. The Pearson’s correlation coefficient (r) was 0.5657. The Runs test for randomness was not statistically significant (0.2923). This indicates that in in vivo preclinical hepatic toxicology, an increase in ALT is associated with an increase AST (Figure 2).

Thrombocytes: In all the 25 selected studies, platelet count was an indicator of hepatic activity. Hepatotoxicity and liver damage consistently resulted in decreased platelet count. Treatment with hepatoprotective agents such as silamyrin reversed the decrease. The decrease in platelet count was independent on the kind of hepatotoxic agent used. This was confirmed as Area Under the Curve of the vehicle control (AUC 18840; mean 461.7 ± 36.03) was higher than that of the of toxicant control (AUC = 15931; mean 438.9 ± 52.99). The D'Agostino & Pearson omnibus normality test indicated that the platelet counts of vehicle control were normally distributed whereas that of the hepatotoxic groups was skewed (K2 of 2.741 vs 49.93 < 0.0001) (Figures 3a & 3b).

Platelet and Liver enzymes: In examining each of the results, it was apparent that an increase in the liver enzymes AST and ALT generally corresponded with a decrease in platelet count. There was a weak positive linear relationship between AST and platelet count. Using linear regression equation Y = 3.451*X - 0.0, the slope was 3.451 ± 0.2984, the Pearson’s coefficient (r) was 0.2226 at a p value of 0.0065 (Figure 4).

Platelet and √ALT: Similarly, using the equation Y = 41.98*X - 0.0, at a slope of 41.98 ± 2.474, there are a virtually little or no correlation between √ALT and platelet count. Furthermore, there was a significant degree of randomness within the data set using Run test. The Pearson coefficient (r) was 0.0913, the Runs test was 0.0021 (Figure 5).

The liver’s susceptibility to chemical assault results from its blood perfusion rate, anatomical position, complex biochemical processes, and its role in chemical disposition [1]. A number of experimental models have been developed to determine the safety of medicinal products over the years. Preclinical safety studies are very laborious, expensive, and may require the use of large number of animals. A sensitive but less costly model can aid in the screening for many hepatoactive substances. Using a clinical index of liver fibrosis, we had proposed that hepatoactivity could easily be quantified from simple measure of platelet, AST and ALT levels experimentally.

During the literature search, it was apparent that many preclinical hepatic toxicological studies did not include hematograms. In some studies, haematological analysis omitted platelet count. However, there is strong scientific literature supporting the indispensable role of the liver in hematopoiesis [18]. In fact, in developmental biology, the liver is recognized as a haematopoietic organ and, more importantly, several elucidated hepatic processes involve interactions with blood derived cells or factors [19]. As such, hepatic damage invariably reflects as haematological aberrations. In line with this, many well-studied hepatotoxicants mechanistically interact with hematostaic factors [20-22].

Hepatic stellate cells appear to initiate and coordinate the actions of other mediators of liver damage [23]. Sinusoidal and sub-endothelial hepatic stellate cells induce sequestration of circulating thrombocytes during early phase of hepatic chemical injury. As such, thrombocytopenia is common to the pathophysiology of many hepatic disorders [24-26]. This has also been demonstrated in preclinical animal models of experimental liver fibrosis. Subsequently, liver damage has been managed experimentally with some degree of success using platelets or platelet rich fractions [27-29]. From this study, an important biomarker of hepatic damage is a decrease in platelet count. Again, liver regeneration was characterized by reversal of thrombocytopenia. This was evident in studies that employed silymarin or antioxidants. Ideally, the best time for therapeutic intervention in curative preclinical hepatic toxicological study should be after detection of appreciable decrease in platelet count.

The thrombotic pathway links up with the inflammatory cascade during hepatic injury and these two processes may be induced by reactive oxygen species and oxidative stress [30]. Tissue necrosis factor from macrophages/monocytes acting through TNF receptor 1 in the early phases of acute inflammation activate hepatic stellate cell which regulates matrix metalloproteinase 9 (MMP-9) expression and extracellular matrix (ECM) deposition [30-32]. The damage to hepatocytes causes the release of liver enzymes.

The two hepatic enzymes used in computing FIB-4 index exhibit different kinetic profiles. ALT is exclusively cytosolic and has a half-life of 47 hours [34,35]. It is a widely accepted specific indicator of hepatic injury than AST in humans [34]. There are two isoforms of AST. The cytosolic isoenzyme which is released early upon damage to hepatocytes has a short half-life of 17.5 hours and the delayed release of mitochondria enzyme which has a half-life of approximately 87 hours [36]. In analyzing the relationship between platelet count and liver enzymes, there was a correlation, albeit weak, between a decrease in platelet count and an increase in AST levels during chemical assault. However, there was a very minimal correlation between the decrease in platelet count and the rise in alanine transaminase. From our analysis, APT may be more sensitive and auto correlate to changes in platelet count than alanine transaminase. Perhaps, this may be the basis for the use of APRI index as a tool for clinical diagnosis of fibrosis.

However, in most human hepatic disorders, except alcoholic hepatitis and Reye syndrome, ALT activity level is higher than that of AST and a very good indicator of hepatic injury [37]. Contrary to this, in examining preclinical hepatic studies, AST activity levels were higher than ALT in most of the studies. It is possible that most experimental hepatotoxicants mechanistically target the mitochondria or are used for prolong periods, causing mitochondrial damage.

The sensitivity and specificity of FIB-4 as a diagnostic tool in preclinical hepatic toxicology study based on AUROC was about 95-98%. According to Zhijian et al., a diagnostic tool with Area Under the Reciever Operator Curve greater than 90% is an excellent tool [38]. This clearly indicates that FIB-4 has the potential to be a good tool for detecting hepatoactive substances. We further analyzed to see if FIB-4 index could discriminate or separate hepatotoxic from hepatoprotective chemicals in these selected studies. It’s performance was averagely 60% (51-67). This finding was important because the data employed were not homogenous. Variations between laboratories, differences in study design, different animal species, nutrition status of animals especially pyridoxine levels, sensitivity of equipment could all influence the outcome. In fact, some of the studies, selected and included here, used dose regimen of hepatotoxicant that could not possibly have induced hepatic damage. Again, because we employed secondary data without dispersions from the mean; the discriminatory performance or separability of FIB-4 in the studies could have been affected.

Summary of findings and recommendations

• In preclinical studies, the relationship between FIB-4 index and its independent variables is linear.
• In general, the FIB-4 index of the naïve/sham/ vehicle control treated animals tends to be far lesser than that of toxicant treated animals. Hepatoprotective substance exhibited FIB-4 index approximate to or less than that of the naive/vehicle /sham treated animals.
• The ratio of FIB-4 index of toxicant treated to FIB-4 of the naïve/sham/ vehicle control treated animals is greater than 1 with a negative linear correlation.

Platelet count may also be an early indicator of hepatotoxicity in experimental models of liver damage as such in curative studies, treatment should be initiated only after a decrease in platelet count is detected.

FIB-4 index is a quick, sensitive, objective index that can be employed in preclinical hepatic toxicology studies as an auxiliary and complimentary tool to aid in detecting hepatoactive substances. It has tremendous potential as a diagnostic tool and monitoring biomarker in studying the prognosis of hepatic damage or disorder without necessarily having to euthanize animals for analysis at every stage of the study.

FIB-4: Fibrosis 4 index; AST: Aspartate Transaminase; ALT: Alanine Transaminase; APRI: AST to Platelet Ratio Index; APAP: Paracetamol; CCL4: Carbon Tetrachloride; INH: Isoniazid; RIF: Rifampicin; AUROC: Area Under the Receiver Operating Characteristic curve

Ethical Statement

Not Applicable.

Funding

Authors had no funding.

Data availability

All data have been supplied.

Acknowledgement

The author acknowledges the help and advice of the late Nana Kweku Adu Sasu with this research work.