TRED‐HF was a randomized controlled trial with a follow‐on single‐arm cross‐over phase that examined the safety and feasibility of therapy withdrawal in patients with recovered DCM over 6 months. The primary endpoint was relapse of heart failure defined by (i) a reduction in left ventricular (LV) ejection fraction >10% and to <50%, (ii) >10% increase in LV end‐diastolic volume and to above the normal range, (iii) a twofold rise in N‐terminal pro‐B‐type natriuretic peptide and to >400 ng/L, or (iv) evidence of heart failure. LV mass, LV and right ventricular (RV) global longitudinal strain (GLS) and extracellular volume were measured using cardiovascular magnetic resonance at baseline and follow‐up (6 months or relapse) for 48 patients. LV cell and extracellular matrix masses were derived. The effect of withdrawing therapy, stratified by relapse and genotype, was investigated in the randomized and follow‐on phases. In the randomized comparison, withdrawing therapy led to an increase in mean LV mass [5.4 g/m2; 95% confidence interval (CI) 1.3–9.5] and cell mass (4.2 g/m2; 95% CI 0.5–8.0) and a reduction in LV (3.5; 95% CI 1.6–5.5) and RV (2.4; 95% CI 0.1–4.7) GLS. In a non‐randomized comparison of all patients (n = 47) who had therapy withdrawn in either phase, there was an increase in LV mass (6.2 g/m2; 95% CI 3.6–8.9; P = 0.0001), cell mass (4.0 g/m2; 95% CI 1.8–6.2; P = 0.0007) and matrix mass (1.7 g/m2; 95% CI 0.7–2.6; P = 0.001) and a reduction in LV GLS (2.7; 95% CI 1.5–4.0; P = 0.0001). Amongst those who had therapy withdrawn and did not relapse, similar changes were observed (n = 28; LV mass: 5.1 g/m2, 95% CI 1.5–8.8, P = 0.007; cell mass: 3.7 g/m2, 95% CI 0.3–7.0, P = 0.03; matrix mass: 1.7 g/m2, 95% CI 0.4–3.0, P = 0.02; LV GLS: 1.7, 95% CI 0.1–3.2, P = 0.04). Patients with TTN variants (n = 10) who had therapy withdrawn had a greater increase in LV matrix mass (mean effect of TTN: 2.6 g/m2; 95% CI 0.4–4.8; P = 0.02).