This study aimed to investigate the role of baseline levels of peripheral inflammation when testing the efficacy of antidepressant augmentation with minocycline in patients with treatment-resistant depression. We conducted a 4-week, placebo-controlled, randomised clinical trial of minocycline (200 mg/day) added to antidepressant treatment in 39 patients selected for elevated levels of serum C-reactive protein (CRP ≥ 1 mg/L), n = 18 randomised to minocycline (M) and n = 21 to placebo (P). The main outcome was the change in Hamilton Depression Rating Scale (HAM-D-17) score from baseline to week 4, expressed both as mean and as full or partial response, in the overall sample and after further stratification for baseline CRP≥3 mg/L. Secondary outcomes included changes in other clinical and inflammatory measures. Changes in HAM-D-17 scores and the proportion of partial responders did not differ between study arms. After stratification for CRP levels <3 mg/L (CRP−) or ≥3 mg/L (CRP+), CRP+/M patients showed the largest changes in HAM-D-17 scores (mean ± SD = 12.00 ± 6.45) compared with CRP-/M (2.42 ± 3.20, p < 0.001), CRP+/P (3.50 ± 4.34, p = 0.003) and CRP−/P (2.11 ± 3.26, p = 0.006) patients, and the largest proportion (83.3%, p = 0.04) of partial treatment response at week 4. The threshold point for baseline CRP to distinguish responders from non-responders to minocycline was 2.8 mg/L. Responders to minocycline had higher baseline IL-6 concentrations than non-responders (p = 0.03); IFNγ was significantly reduced after treatment with minocycline compared with placebo (p = 0.03). Our data show some evidence of efficacy of add-on treatment with minocycline in MDD patients but only in those with low-grade inflammation defined as CRP ≥3 mg/L.