Reply to: “Gene-environment “amplification” of liver stiffness requires more clinically meaningful analytic approaches”

To the Editor:

We thank Drs Zhang et al. for their interest in our study,¹ and for emphasizing both the clinical relevance, but also the biological plausibility of our findings. Together with the increasing evidence from population-based studies, we sought to shed light on how metabolic risk factors and alcohol interact with genetic susceptibility (PNPLA3, TM6SF2) on the degree of liver fibrosis as assessed by vibration-controlled transient elastography (VCTE)-based liver stiffness measurement (LSM). Understanding whom to test and when is key for more personalized clinical management, yet the incremental value of incorporating genetic risk profiles into routine patient assessment has been highlighted as insufficiently studied in recent EASL clinical practice guidelines for MASLD. Especially with the growing awareness of the synergistic effect of metabolic risk and alcohol on liver disease severity, understanding genetic susceptibility has the potential to facilitate patient management.^2,3 We briefly want to comment on the two points raised by the authors.

We agree that patient pathways in steatotic liver disease (SLD) are based on specific LSM cut-offs,^3,4 with a cut-off <8 kPa identifying the patient population in whom advanced fibrosis can be ruled out.[5], [6], [7] Nevertheless, one cannot highlight enough that these cut-offs are always based on probabilities and are associated with uncertainty. For illustration, let us assume we have two patients: One patient with an LSM of 7.9 kPa will be treated as healthy (black) and another one with 8.0 kPa as diseased (white). While such stratification may be useful and is needed when it comes to patient management, it introduces avoidable misclassification and information loss, which can obscure the biological truth (i.e. fibrosis severity) when modelled like this compared with when LSM is modelled as a continuous variable.

Nevertheless, we agree that it is important for physicians to know the probability of observing a certain phenotype (e.g. ≥8 kPa and ≥12 kPa) given specific exposures. As such, we now provide heatmaps of observed probabilities of ≥8 kPa and ≥12 kPa in both cohorts in individuals with and without certain metabolic risk factors and alcohol consumption, stratified across PNPLA3 G-allele carriers and noncarriers (Fig. 1). In line with previous findings, prevalence increased in carriers of the PNPLA3 G-allele, with the increase being more pronounced in those with insulin resistance, higher alcohol consumption, and obesity (only cohort II).

Fig. 1.

Prevalence of ≥8 kPa and ≥12 kPa in the Tertiary-care cohort and the At-risk cohort from.²

Prevalences are shown in relative % of the respective subgroups.

Second, we agree that LSM captures hepatic inflammation in addition to hepatic fibrosis. This is supported by studies in chronic hepatitis C showing a prompt decline of LSM during or immediately after treatment,⁸ as well as biopsy studies, albeit with heterogeneous results. However, transaminases poorly correlate with inflammation on a histological level.⁹ Because false-positive LSM has been mostly studied in the context of excessive alcohol intake, and given heterogeneous data regarding ALT/AST as indicators for unreliable LSM, current guidelines recommend repeating VCTE only if elevated LSM is observed in the presence of both excessive drinking and AST >70 U/L.

In our study, median ALT was 38 U/L (IQR: 25–63) in the Tertiary-care cohort (21% with ALT >70 U/L, 3.5% with >175 U/L corresponding to ∼5 × ULN) and 27 U/L (IQR: 21–37) in the At-risk cohort (5.7% with ALT >70 U/L, 0.3% with >5 × ULN). Median AST was 30 U/L (IQR: 23–42) in the Tertiary-care cohort (8.6% with AST >70 U/L, 1.5% with >5 × ULN) and 25 U/L (IQR: 21–31) in the At-risk cohort (2.6% with AST >70 U/L, 0.1% with >5 × ULN). A relevant influence on LSM is therefore unlikely. Excluding patients with AST or ALT >70 U/L did not change any results (table not included due to formatting restrictions). Nevertheless, one must remember that such an approach significantly changes the patient population studied, as both PNPLA3 and TM6SF2 per se lead to higher AST, ALT or hepatic inflammation.¹⁰

Overall, the data consistently indicate a clinically relevant influence of single nucleotide polymorphisms in PNPLA3 and TM6SF2 on liver fibrosis progression and liver disease phenotype. Acknowledging gene–environment interactions enhances our understanding of their effects and may guide implementation of genetic testing in clinical routine.

Financial support

No financial support received to produce this manuscript.

Authors' contributions

Drafting of the manuscript (all author), critical revision of the manuscript for important intellectual content (all author).

Conflicts of interest

S.G. received travel support from Ipsen, Roche, Galapagos, Gilead. H.L.S. has nothing to disclose. M.T. received speaker fees from Echosens, Madrigal, Takeda, and Novo Nordisk. Advisory fee from Boehringer Ingelheim, Astra Zeneca, Novo Nordisk and GSK. Research grant from GSK. Co-founder and board member for Evido. Board member for Alcohol & Society (non-governmental organisation). Funded by a grant from the Novo Nordisk Foundation (NNF20OC0059393). G.S. received speaker fees from DiaSorin. Please refer to the accompanying ICMJE disclosure forms for further details.

Supplementary data

The following are the Supplementary data to this article: