Abstract Objectives To analyse whether integrase inhibitor (InSTI)-based regimens achieve better immunological recovery than NNRTI- or boosted PI (bPI)-based regimens as initial ART. Methods In a retrospective analysis, we selected patients who initiated ART with two NRTIs plus an InSTI, an NNRTI or a bPI and maintained both the same ‘third drug’ and an HIV-RNA <50 copies/mL in ≥95% of determinations once undetectable viral load had been achieved. We compared CD4+ count, %CD4+ and CD4+/CD8+ ratio recovery over 2 years. Data were analysed using mixed-effects regression models for repeated measures. Results Of the 836 patients included, 208, 481 and 147 initiated with InSTI, NNRTI and bPI, respectively. For CD4+, %CD4+ and CD4+/CD8+ two main slopes were identified: from month 0 to month 6, with the highest increments; and from month 6 to month 24, with smaller increases every semester. Although the patients on InSTI achieved undetectable viral load faster, for CD4+ and %CD4+ there were no differences in the slopes of change according to the third drug either for the first phase (P = 0.137 and P = 0.393, respectively) or from month 6 onwards (P = 0.834 and P = 0.159, respectively). The increase in CD4+/CD8+ was slightly higher for bPI compared with InSTI (difference of 0.0119, 95% CI 0.0020–0.0205; P = 0.018), but clinically negligible. From month 6 onwards, no differences were found between treatment groups (P = 0.176). Conclusions Immune restoration measured as CD4+ count, %CD4+ and CD4+/CD8+ increases was independent of the third antiretroviral drug class used when given with two NRTIs.