Optimum detection of the Burkholderia cepacia complex (BCC) from sputum of patients with cystic fibrosis (CF) is essential in preventing patient-to-patient transmission of this organism. The aim of this study was to develop an improved PCR assay with reference to sensitivity for the direct detection of BCC organisms from CF sputum employing the recA locus. The sensitivity results of three recA PCR assays were compared using various combinations of previously published primers. These included (i) a single-round approach using the primer set BCR1/BCR2, yielding a 1036-bp product, (ii) a single-round approach using the primer set BCR1/Mr, yielding a 465-bp product, and (iii) a semi-nested PCR (SN-PCR) approach using the primer set BCR1/BCR2 followed by BCR1/Mr. The sensitivity of these assays were determined by spiking B. cepacia-free sputum with known numbers of four strains of BCC, namely, genomovar II [B. multivorans] (C1576), genomovar IIIa (C5424, C6433) and genomovar IIIb (C1394). Following optimization, the chosen assay was performed on 14 patients. Employment of the singleround assay with BCR1/BCR2 was the least sensitive with a detection threshold of 107 cfu/g sputum for GIIIa and GIIIb, and 108 cfu/g sputum for GII. Sensitivity was improved by targeting the smaller amplification region of the recA locus (465 bp) employing the BCR1/Mr primer pair, in combination with a single-round approach, whereby the detection threshold was improved by 1 log for each genomovar. Employment of the semi-nested assay demonstrated optimum sensitivity, whereby the detection threshold increased to 101 and 102 cfu/g sputum for genomovar IIIa/IIIb and genomovar II, respectively. Subsequent genomovar characterisation can be performed by sequencing of the PCR amplicon without the need for culture which may be beneficial in patients in the initial stages of colonisation or who are transiently colonised and who may be culture-negative for BCC.