Introduction: We assessed the correlation between childhood maltreatment (CM) and severity of depression in an elderly unipolar Treatment-Resistant Depression (TRD) sample. Methods: Patients were enrolled from a longitudinal cohort (FACE-DR) of the French Network of Expert TRD Centres. Results: Our sample included 96 patients (33% of the overall cohort) aged 60 years or above, with a mean age of 67.2 (SD = 5.7). The majority of the patients were female (62.5%). The Montgomery and Asberg Depression Rating Scale (MADRS) and Quick Inventory Depression Scale-Self Report (QIDS-SR) mean scores were high, 28.2 (SD = 7.49) [MADRS score range: 0–60; moderate severity≥20, high severity≥35] and 16.5 (SD = 4.94) [IDS-SR score range: 0–27; moderate severity≥11, high severity≥16], respectively. Mean self-esteem scores were 22.47 (SD = 6.26) [range 0–30]. In an age- and sex-adjusted model, we found a positive correlation between childhood trauma (CTQ scores) and depressive symptom severity [MADRS (β = 0.274; p = 0.07) and QIDS-SR (β = 0.302; p = 0.005) scores]. We detected a statistically significant correlation between physical abuse and depressive symptom severity [MADRS (β = 0.304; p = 0.03) and QIDS-SR (β = 0.362; p = 0.005) scores]. We did not observe any significant correlation between other types of trauma and depressive symptom severity. We showed that self-esteem (Rosenberg scale) mediated the effect of physical abuse (PA) on the intensity of depressive symptoms [MADRS: b = 0.318, 95% BCa C.I. [0.07, 0.62]; QIDS-SR: b = 0.177, 95% BCa C.I. [0.04, 0.37]]. Preacher & Kelly’s Kappa Squared values of 19.1% (k2 = 0.191) and 16% (k2 = 0.16), respectively for the two scales, indicate a moderate effect. Conclusion: To our knowledge, this is the first study conducted in a geriatric TRD population documenting an association between childhood trauma (mainly relating to PA) and the intensity of depressive symptoms.< Réduire