The ability to detect tuberculosis (TB) continues to be a global health care priority. This paper describes the development and preliminary assessment of the clinical accuracy of a heterogeneous immunoassay that integrates a serum pretreatment process with readout by surface-enhanced Raman scattering (SERS) for the low-level detection of mannose-capped lipoarabinomannan (ManLAM). ManLAM is a major virulence factor in the infectious pathology of Mycobacterium tuberculosis (Mtb) that has been found in the serum and other body fluids of infected patients. The effectiveness of ManLAM as a TB diagnostic marker, however, remains unproven for reasons not yet well understood. As reported herein, we have found that (1) ManLAM complexes with proteins and possibly other components in serum; (2) these complexes have a strongly detrimental impact on the ability to detect ManLAM using an immunoassay; (3) a simple pretreatment step can disrupt this complexation; and (4) disruption by pretreatment improves detection by 250×. We also describe the results from a preliminary assessment on the utility of serum pretreatment by running immunoassays on archived specimens from 24 TB-positive patients and 10 healthy controls. ManLAM was measurable in 21 of the 24 TB-positive specimens, but not in any of the 10 control specimens. These findings, albeit for a very small specimen set, translate to a clinical sensitivity of 87.5% and a clinical specificity of 100%. Together, these results both provide much needed evidence for the clinical utility of ManLAM as a TB marker, and demonstrate the potential utility of our overall approach to serve as a new strategy for the development of diagnostic tests for this disease.