Urine contains some minute amount of albumin. i doubt your nanoparticles are adsorbed or may be masked with the albumin. this might not be the case when you use buffer. Try adding albumin to your buffer and check the reading. if the latter is the case, you might draw a line!! 

Sensitivity of the assay depends on signal generated from label used in the system. If HRP is used as label the assay sensitivity changes by changing the substrate used in the system. The order of sensitivity is given below when colorimetric, fluorometric and chemiluminometric substrates are used. Colorimetric assays

Urea is also a pretty good protein denaturant. Different antibodies may vary in their sensitivity to this denaturation, and it's effect on antigen binding. If you have additional antibody versions available, you may want to look for one that works better. I think it fair to say that you are going to lose some sensitivity in urine relative to serum or buffer. Tweak the system till you can live with the sensitivity.

I'm not aware that anyone had tried before, but MST (MicroScale Thermophoresis) is a technique to quantifiy interactions between any kind of biomolecules in buffer but also in complex bioloquids (sera, cell extracts, etc.), so it should work in urine as well. This means, you could directly determine the binding affinity of your antibody-antigen interaction in buffer and urine (or different mixtures). In the next step you could try the interaction in presence of urea (using similar concentrations as typically found in urine) and you can also directly test the interaction in presence of albumin (or other substance you suspect could interfere with the binding). With this set of data, you should get a quite good understanding of what is going on.

I agree with Tulsidas on his comment regarding the variation of sensitivity based on the substrate. Specially in metalloimmunoassys the sensitivity might have to be sacrificed for the precision of the assay. The solution to this might be the use of label free technologies as QCM biosensors (http://www.dddmag.com/news/2012/05/quartz-crystal-microbalance-technology-provides-biologically-relevant-data). Using these technologies the full binding kinetics (association and dissociation constants; ka and kd) and affinity (KD) is easily derived with the same reliability and precision as for standard biochemical assay. The other advantage is the possibility to regenerate the surface. This allows you to carry out interaction studies on the same surface using different experimental conditions, for instance using buffer or complex biological fluids. There are sensor surfaces available that show very low unspecific binding. 

Thanks for all the feedback.  I've yet to do a systematic study to answer this question, but I can say that sensitivity is also decreased in artificial urine (i.e. a mixture of salts at pH 6) but not to the same extent.  As a result, urea, or perhaps the combination and amount of salts and the slightly different pH play important roles.  I've heard of high ionic strengths interfering with antibody-antigen bonding when the antibody-antigen bonding involves electrostatics to a large degree.  Regarding sensitivity, I'm actually not at all interested in increasing the sensitivity or lowering the LOD of the assay.  Rather, I'm currently attempting to shift/expand the dose-response to much higher target concentrations.