Partial safety factors — the load and resistance factors embedded in limit state design codes — are the primary mechanism through which structural design codes translate probabilistic reliability targets into deterministic design practice. Their calibration against explicit reliability targets using the First-Order Reliability Method (FORM) is a mathematically rigorous process that remains poorly documented and rarely applied in the context of bridge design codes in developing regions. This paper presents a comprehensive statistical calibration framework for partial safety factors applicable to bridge design codes, grounded in the Hasofer-Lind-Rackwitz-Fiessler (HL-RF) FORM algorithm. The framework is applied to six bridge structure types — reinforced concrete slab, pre-stressed concrete box girder, steel composite, reinforced concrete arch, steel truss, and cable-stayed — with material statistical parameters characterised from African and international laboratory databases. For each bridge type, the resistance model uncertainty, dead load bias, and live load statistics are quantified and used to compute the target reliability index beta under the ultimate limit state (ULS). Calibrated partial factors gamma_G (dead load) and gamma_Q (live load) are derived that achieve a target reliability index of beta = 4.3 for a 50-year reference period, consistent with the EN 1990 Annex B recommendation for consequence class CC2 bridges. Results reveal that the EN 1990 default factors overestimate required safety for steel composite and truss bridges — suggesting material efficiency gains of 5 to 9% — and underestimate required safety for reinforced concrete arch bridges in tropical climates, which require gamma_G = 1.38 and gamma_Q = 1.60 to achieve the target beta. The sensitivity