Discussion
The ECM is a highly dynamic structure that is integral to myocardial structure and function which detrimentally remodels following cardiac injury leading to the altered turnover, replacing contractile tissue with collagen rich connective tissue and ultimately the development of myocardial fibrosis.5 Myocardial fibrosis is characterised by adverse remodelling which contributes to systolic and diastolic HF.5 28 PINP, PICP and PIIINP are released into the circulation during conversion and deposition of procollagen to collagen and are upregulated during myocardial fibrosis and associated with adverse HF outcomes.5 7 15 28 Mechanistically, higher upregulation of collagen would challenge a CRT’s ability to reverse remodel and for the patient to respond. Umar et al10 supported this hypothesis observing significantly lower baseline PINP expression predicted echocardiographic response. Dong et al26 did observe lower baseline PIIINP predicted echocardiographic response on univariate analysis, but not multivariable analysis. In contrast, Garcia-Balao et al9 observed higher baseline expression of PICP in responders and PICP:CITP ratio (type I collagen turnover) of ≥14.4 had greater than twofold increased chance of predicting functional response, driven by PICP. Critically, echocardiographic and clinical/functional response criteria correlate poorly,18 so could not be contrasted. Importantly, Lopez-Andres et al,8 the largest study included in the review, did not observe upregulation of collagen synthesis predicting echocardiographic non-response, which does challenge the Umar et al10 and Dong et al26 observations; however, the cohort characteristics and study designs were different. The observations of collagen synthesis following CRT implantation conflict with each other. Umar et al10 reported a significant increase in PINP and decrease in PIIINP expression in responders at 6 months; both would mechanistically be expected to be lower at follow-up. In contrast Garcia-Bolao et al9 observed PICP levels decreased for responders and increased for non-responders at 1 year, which would be expected, but is based on a functional response definition. In contrast to collagen synthesis, degradation biomarkers (ICTP or CITP) did not predict CRT response.8–10 Furthermore, no significant change in ICTP or CITP expression was observed at follow-up across all three studies.8–10 Alteration in collagen synthesis rate is observed to be more powerful at predicting response than collagen degradation. Different patterns of collagen synthesis biomarkers predicting response have been observed; lower expression predicted LV reverse remodelling,10 26 whereas higher rates predicted functional response.9 The variation in these patterns is explained by the different response definitions and cohort characteristics. The study cohort for Umar et al10 had a higher proportion of men and ischaemic cardiomyopathy than Garcia-Bolao et al.9 The heterogeneities between these studies make drawing conclusions difficult. Lopez-Andres et al8 also challenge any observations due to size of cohort and no prediction value to collagen turnover observed. Overall, collagen synthesis is observed to be important in predicting CRT response, especially LV reverse remodelling, with results replicated in two studies that lower rates predict LV reverse remodelling.10 26
MMP-1, MMP-2 and MMP-9 perform a critical role in myocardial collagen degradation and have been identified as being important prognostic markers in HF.11 13 27 Predictive value for CRT non-response (death or LVEF ≤35% at 18 months) was only demonstrated in baseline MMP-1 expression ≤3 ug/l8 supporting an observation by Jordan et al11 that lower MMP-1 inferred worse HF prognosis. MMP-2 had large variations observed between the included studies,8 24 but was not demonstrated to predict response. MMP-9 was only observed in one included study showing no predictive value9; however, recently Dini et al13 demonstrated raised levels (>238 ng/mL) and predicted worse HF outcomes. MMP activity was not considered in any of these studies as a predictor but would be important to consider in the future. Current evidence suggests that MMPs, especially MMP-2 and MMP-9, have not yet had their potential fully evaluated.
TIMP-1 regulates the endogenous proteolytic MMP system involving discordant inhibition and in chronic inflammatory states stimulating collagen synthesis and myocardial fibrosis.5 24 Tolosana et al24 observed a significant baseline difference in RvsNR expression with lower TIMP-1 in responders. Tolosana et al24 demonstrated that baseline TIMP-1 (≥248 ug/L) predicted CRT non-response. Trucco et al29 in long-term follow-up of the same cohort demonstrated that the same threshold independently predicted mortality at 60±34 months (sensitivity 80% and specificity 71%). Tolosana et al24 also demonstrated that statistically significant lower TIMP-1 is found in participants that do LV reverse remodel (LVESV reduction ≥10%). Umar et al10 and Garcia-Bolao et al9 observed no difference statistically at baseline. Variation between the reported literature in the magnitude of association of TIMP-1 exists; however, Tolosana et al24 offers a well-designed prospective observational study which is powered giving strength to the conclusions drawn.
Gal-3 stimulates fibroblasts to release TIMPs and MMPs that regulate collagen turnover, resulting in myocardial fibrosis.14 Elevated levels are independent predictors of adverse outcomes in HF.14 Evaluation of Gal-3 as a predictor of response was limited, as RvsNR was not reported in either of the two studies.8 25 Truong et al25 demonstrated peripheral baseline Gal-3 ≥25.9 ug/L had specificity for predicting CRT response. Lopez-Andres et al8 observed Gal-3 baseline expression ≥30 ng/L had nearly threefold increased risk of death or hospitalisation for worsening HF following CRT. Though not demonstrated to be a strong predictor, the evidence suggests that Gal-3 is a good biomarker for predicting poor outcomes in HF and needs further evaluation.
The greatest challenge for research into CRT response and one this review demonstrated is lack of an accepted response definition. Differing definitions rarely correlate,18 which our review clearly demonstrates. Echocardiographic and clinical/functional definitions correlate very poorly and should never be compared or applied in a composite definition18; LV reverse remodelling should be considered separately.18 30
Study limitations
Heterogeneity among included studies was widespread despite a rigorous eligibility and screening criteria. The variations in study design, cohort characteristics and response definitions made pooling data in a meta-analysis impractical. CRT implantation techniques and indications have evolved over the last 15 years and offer another source of heterogeneity. Furthermore differences in laboratory techniques account for some variation among biomarker results. These limitations are particularly important to consider in future research studies.