Reply to: “Prediction of liver-related mortality in a community setting”

Prediction of liver-related mortality in a community settingJournal of HepatologyVol. 78Issue 4PreviewWe would like to congratulate Innes et al. for their study assessing the usefulness of established risk scores as a predictor of development of cirrhosis-related complications.1 Because of their everyday relevance, the study also raises several questions. While development of cirrhosis-related complications constitutes an indisputable hard endpoint, the occurrence of liver-related death as a primary cause of death should also be considered. Given that the UK Biobank (UKB) is linked to national death registries, the survival status and the causes of death are readily available. Full-Text PDF Performance of routine risk scores for predicting cirrhosis-related morbidity in the communityJournal of HepatologyVol. 77Issue 2PreviewModels predicting an individual’s 10-year risk of cirrhosis complications have not been developed for a community setting. Our objectives were to assess the performance of existing risk scores – both with and without genetic data – for predicting cirrhosis complications in the community. Full-Text PDF Open Access Mind the prevention paradox To the Editor: We thank Schneider et al.[1]Schneider C.V. Gross S. Stmad P. Prediction of liver-related mortality in a community setting.J Hepatol. 2022; Abstract Full Text Full Text PDF Google Scholar for their interest in our study.[2]Innes H. Morling J.R. Buch S. Hamill V. Stickel F. Guha I.N. Performance of routine risk scores for predicting cirrhosis related morbidity in the community.J Hepatol. 2022; 77: 365-376Abstract Full Text Full Text PDF PubMed Scopus (11) Google Scholar Our objectives were two-fold. First, to assess if it is possible to estimate 10-year risk of cirrhosis morbidity using readily available scores. Second, we determined the extent to which these scores could be enhanced by incorporating information on germline genetic polymorphisms. Onset of cirrhosis morbidity represents a watershed point in the natural history of chronic liver disease.[3]D’Amico G. Garcia-Tsao G. Pagliaro L. Natural history and prognostic indicators of survival in cirrhosis: a systematic review of 118 studies.J Hepatol. 2006; 44: 217-231Abstract Full Text Full Text PDF PubMed Scopus (2016) Google Scholar It heralds a downturn in patient quality of life, and an increase in health system expenditure. It is also of major interest to “at risk” individuals, who want to know their chance of developing life-impairing overt liver disease. Our decision to set incident cirrhosis morbidity as the primary outcome event was influenced by these factors. Schneider et al.’s analysis of UK Biobank (UKB) data suggests a sizeable proportion of liver deaths occur in individuals with lower FIB4/APRI/CIRRUS values.[1]Schneider C.V. Gross S. Stmad P. Prediction of liver-related mortality in a community setting.J Hepatol. 2022; Abstract Full Text Full Text PDF Google Scholar This is an interesting observation. Their results are supported by previous studies.[4]Aberg F. Luukkonen P.K. But A. Salomaa V. Britton A. Petersen K.M. et al.Development and validation of a model to predict incidence chronic liver disease in the general population: the CLivD score.J Hepatol. 2022; 77: 302-311Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar,[5]Haagstrom H. Talback M. Andreasson A. Walldius G. Hammar N. Repeated FIB-4 measurements can help identify individuals at risk of severe liver disease.J Hepatol. 2020; 73: 1023-1029Abstract Full Text Full Text PDF PubMed Scopus (75) Google Scholar For example, in a study of FIB-4 values from a community cohort in Sweden, Haagstrom et al. reported that half of incident cases of severe liver morbidity occurred in patients with low FIB-4 (<1.35).[5]Haagstrom H. Talback M. Andreasson A. Walldius G. Hammar N. Repeated FIB-4 measurements can help identify individuals at risk of severe liver disease.J Hepatol. 2020; 73: 1023-1029Abstract Full Text Full Text PDF PubMed Scopus (75) Google Scholar At first glance, this seems paradoxical. That is, if these risk scores really are so good at estimating 10-year risk, why do so many events occur in lower risk patients? To answer this point, we must think about how 10-year risk is distributed in the population. The number of people who develop a liver event within a given risk group depends not just on the 10-year risk, but also on the number of people who fall into that risk group. Note that low risk is not zero risk; thus, if the number of low-risk individuals is sufficiently high, there will naturally be scope for many events to occur. Vice versa, if the number of high-risk patients is relatively small, then it is not surprising if the number of events occurring in this group is also small. However, this does not take away from the fact that high-risk individuals are de facto more likely to develop the outcome event than low-risk individuals. Thus, the occurrence of events in low-risk patients does not undermine our conclusion regarding the ability of FIB4/APRI to predict 10-year risk. From a broader perspective, the phenomenon Schneider et al. describe – i.e. cases from low-risk groups exceeding those from high-risk groups – is a corollary of the prevention paradox, which applies widely to many risk factors and scores.[6]Rose G. Sick individuals and sick populations.Int J Epidemiol. 2001; 30: 427-432Crossref PubMed Scopus (901) Google Scholar For example, older maternal age is strongly associated with a higher likelihood of giving birth to a Child with Down’s syndrome. Despite this, most children with Down’s syndrome are born to younger mothers.[6]Rose G. Sick individuals and sick populations.Int J Epidemiol. 2001; 30: 427-432Crossref PubMed Scopus (901) Google Scholar Does this mean therefore that maternal age is not a “useful” predictor of giving birth to a child with Down’s syndrome? Schneider et al. call for more research “to enable the detection of liver-sick patients”. This may entail efforts to improve the discriminative ability and calibration of existing risk models – e.g. as we tried to do in our study by integrating genetic data – or even the development of new risk models entirely to estimate 10-year risk. We support this. However, we should not expect that an improved risk model is going to suddenly present us with a different and more convenient risk distribution. In our view, the prevention paradox that Schneider et al. highlight, is not a poor reflection on the risk score, but a reflection on the risk distribution which the risk score is simply describing. Nevertheless, Schneider et al. are right to question how 10-year risk could be used in clinical practice. In our view, we need an open discussion between patients, clinicians and healthcare payers about the risk/trade-offs we are willing to tolerate. This will inform development of decision-rules, mapping risk to clinical action. These ideas are eloquently articulated by Rowe and D’Amico.[7]Rowe I.A. D’Amico G. Taking a risk-based approach to testing for liver disease in primary care, a step in the right direction.J Hepatol. 2022; 77: 293-295Abstract Full Text Full Text PDF PubMed Scopus (1) Google Scholar At the moment, the “cart is before the horse” – i.e. we arguably have the risk tools, but we have not built consensus around how to use them. Overall, we contend that considering 10-year risk and how it is distributed is vital. Understanding this will facilitate a more strategic and patient-centred approach to managing liver disease in the community. HI is supported by a viral hepatitis fellowship from the Medical Research Foundation (grant ID: C0825). The authors declare no conflicts of interest that pertain to this work. Please refer to the accompanying ICMJE disclosure forms for further details. Writing and editing manuscript: HI and ING. The following are the supplementary data to this article: Download .pdf (.21 MB) Help with pdf files Multimedia component 1