TIMP-1: is surrogate testing for extracellular matrix remodelling ready for clinical translation or only an Icelandic phenomenon?

This editorial refers to ‘Fibrosis as measured by the biomarker, tissue inhibitor metalloproteinase-1, predicts mortality in Age Gene Environment Susceptibility-Reykjavik (AGES-Reykjavik) Study’†, by G. LaRocca et al., on page 3423. The extracellular matrix is indispensable for proper cell–cell contact and organ structure.1 It is a highly dynamic structure with a constant turnover which is regulated by a sophisticated interplay of metalloproteinases, their tissue inhibitors, multiple environmental signals transmitted by chemokines or cytokines, and epigenetic transcriptional modification of involved cells, to name a few (Take home figure). Co-ordinated fibrogenesis is crucial for tissue repair following many different causes of injury and irreversible cell damage. However, extracellular matrix remodelling is also activated in many diverse inflammatory or malignant diseases.1 It is well established that adverse matrix remodelling in the cardiovascular system is causally linked to clinical disease phenotypes such as diastolic heart failure, progression of vascular dilation, and arteriosclerotic plaque instability.2 , 3 Mechanisms involved in tissue fibrosis. A complex interaction of factors such as age, genes, and environment activate pro-inflammatory pathways that lead to transdifferentiation of a variety of different cell types including endothelial and epithelial cells to myofibroblasts. An unbalanced homeostasis of TIMP-1 and MMP-9 finally leads to collagen accumulation and organ fibrosis. Considering the well-established contribution of extracellular matrix remodelling to cardiovascular and non-cardiovascular disease phenotypes, many components involved in fibrogenesis have been used as biomarkers for many diverse disorders including different carcinomas, aortic aneurysms, and rheumatic and inflammatory bowel diseases. This pathophysiologically driven diagnostic approach is indicative of a very important repair and/or remodelling process, but will inevitably lack organ and disease specificity. Considering the ubiquitous expression of the components of the extracellular matrix and the multiple triggers of activation of extracellular matrix remodelling, the signal to noise ratio for a distinct disease process will be low. Thus, in clinical practice, a positive extracellular matrix biomarker finding will share the strengths and limitations of C-reactive protein, e.g. being indicative of activation of a molecular pathway but not of a classical disease entity. While it may aid in risk assessment, in most circumstances it will not trigger any specific therapeutic strategies. In their study in this issue of the journal, LaRocca and co-workers examined a subcohort of 5511 survivors of the initial Reykjavik study, now 67 years or older, to test the prognostic significance of fibrotic processes in an elderly in-dwelling population.4 These subjects were re-investigated between 2002–2006 and comprise the Age Gene Environment Susceptbility-Rejkjavik (AGES) study.5 The Reykjavik Study and the AGES substudy are epidemiological studies on inhabitants of Iceland, who are well characterized for their vascular, neurocognitive, musculoskeletal, and metabolic phenotypes.5 A particular strength of this cohort is repetitive phenotype assessments during the subjects’ lifetime that allow investigation of pathophysiological changes and treatment responses in a longitudinal and cross-sectional analysis over the 30-year follow-up period. The results derived from this large epidemiological endeavour have generated impressive data on the genetics of complex disorders, shared phenotypes, and the role of common risk variables published in a multitude of original manuscripts. Briefly, the present analysis from the AGES population evaluated the prognostic significance of fibrosis using tissue inhibitor metalloproteinase-1 (TIMP-1) and matrix metalloproteinase-9 (MMP-9) as biochemical surrogates and compared the predictive power with the Framingham risk score and other established biomarkers. As a key finding, TIMP-1 emerged as the second strongest predictor of cardiovascular death and the third strongest predictor of cancer death and other causes of death. Along with TIMP-1, highly sensitive C-reactive protein (hsCRP) demonstrated the highest continuous net reclassification improvement (NRI) over the baseline model for 5-year and 10-year survival, with a more prominent decline of the NRI for hsCRP over time, suggesting that inflammation might be less relevant than fibrosis in this older population of Iceland. There was no significant association between all-cause death and cause-specific death with MMP-9, and the predictive role of renal failure was disappointingly low. Despite this optimistic interpretation of findings on TIMP-1, there are some findings that merit further discussion. First, death in the AGES cohort was mostly attributed to cardiovascular disease, followed by cancer and other causes, accounting for 40.4, 24, and 35.6% of all deaths, respectively. Accordingly, the risk contribution of structural heart disease, atherosclerotic burden, and incident heart failure may not have been reflected appropriately without measurement of natriuretic peptides3 , 6–8 or cardiac troponins.8 For the latter, the introduction of highly sensitive cardiac troponin T (hsTnT) and troponin I (hsTnI) emerged as the strongest independent predictors of cardiovascular death in middle-aged9 as well as in older general populations.7 , 8 In keeping with this, previous findings from the same AGES cohort evaluating the predictive role of hsTnI (Abbott Architect STAT) confirmed the strong and independent value of hsTnI for prediction of 10-year risk of all-cause death, and incidence of cardiovascular disease and coronary heart disease after adjusting for traditional risk factors including age, diabetes, body mass index (BMI), hypertension medication, systolic blood pressure, smoking, log CRP, log creatinine, cholesterol, HDL cholesterol, log triglycerides, and statin medication.10 In the present analysis of the AGES cohort, however, it remains unclear whether hsTn assays or older conventional cTnT or cTnI assays were used, as the investigators did not provide details on assay generation or manufacturer (for cTnI), and used equivocal abbreviations for the assays. Irrespective of which cardiac troponin generation was used, the hazard ratios (HRs) for prediction of 10-year risk of all-cause and cardiovascular death in the multivariate Cox regression analysis were similar with TIMP-1, dampening the initial enthusiasm. Unfortunately, statistical methods that would have allowed a more refined discrimination and re-classification of risk including C-statistics, NRI, categorical NRI, clinical NRI, and integrated discrimination improvement (IDI) were not employed for this particular comparison between cTn and TIMP-1. Secondly, findings were not validated in an external population. The Icelandic population is genetically relatively homogeneous11 and is distinct from other populations due to geographic isolation, freedom from most infectious epidemics, cold climate, restricted diet and increased physical activity, relatively low neonatal mortality, and little air and water pollution, all being factors that can potentially influence epigenetics.5 Therefore, particularities of the AGES cohort can be regarded as an advantage to examine genetic and gene–environment interactions by reducing the need for multiple adjustments, but on the other hand can also be regarded as disadvantageous because generalization of findings to other genetically more heterogenous populations is difficult without independent external validation. The different performance of biomarkers across populations was nicely highlighted in the MORGAM population evaluating 30 biomarker in three different populations.12 In particular, the fully adjusted HR for TIMP-1 was significant in the FINRISK97 men and FINRISK97 women population, but demonstrated a neutral HR in the Belfast Prime men population.12 Thirdly, another issue that limits generalization of findings from the AGES cohort is the considerably older age, with the mean age of the participants of the AGES cohort being 76.8 years (range 68–98 years), whereas ages range between 49 and 73 years in most other general population studies.9 Existing risk scores including the Framingham Risk Score and the ESC SCORE have not been validated sufficiently in patients older than 65–75 years, and it is believed that patients >70 years are already at high risk (>20% 10-year cardiovascular risk).13 , 14 Accordingly, as the traditional risk prediction scores become less informative, older age can cause an overestimation of biomarker-based risk assessment.15 The BIOMARCARE study15 reported such an age-dependent effect regarding NRI of 10-year risk prediction by troponin I over the ESC SCORE. The biomarker-based risk information was particularly useful among individuals aged >65 years without any added benefit in individuals below the age of 45. Finally, there are several other reasons that might have hindered the broader use of TIMP-1 and MMP-9 in clinical routine. There is still no laboratory platform for automated measurement of TIMP-1, and potential pre-analytical errors of TIMP and MMP measurement in peripheral blood have been a matter of debate.16 Serum levels of matrix metalloproteinases are probably influenced by MMP release following degranulation of leucocytes and platelets during the ex vivo blood clotting. Despite increasing evidence for an added value of some biomarkers, e.g. hsTn and natriuretic peptides, the ESC guidelines on cardiovascular disease prevention still do not recommend routine measurement of circulating and urinary biomarkers in addition to the ESC SCORE. Whether TIMP-1 will become a candidate in the future is speculative. The present report from the AGES confers another piece of evidence in this very distinct older population. However, whether findings from this community-based study in Iceland can be generalized to other populations remains illusive in the absence of broad external validation. Nevertheless, this AGES cohort analysis confers an important piece of knowledge improving our understanding of ageing and ageing-related pathophysiological processes and a broader focus on different diseases and causes of death. Conflict of interest: H.A.K. has served as consultant for AstraZeneca and Bayer Vital, and received honoraria for lectures from AstraZeneca, Bayer Vital, Roche Diagnostics, and Daiichi Sankyo. E.G. serves as a consultant for Roche Diagnostics, AstraZeneca, Bayer Vital, and Brahms Deutschland, and received lecture fees from AstraZeneca, Roche Diagnostics, Bayer Vital, and Daiichi Sankyo.