When this came up before, the consensus was that it didn't apply to Nikon DSLR sensors. I believe the "noise is worse at part-stop ISO values" theory may have been true (at some stage) for other brand cameras, because different technology was used.
That's because there is nothing to find. People who have done detailed measurements of sensor performance have found that for the most part, it works as you would intuitively expect: each 1/3-stop increase in ISO lowers the quality (SNR, dynamic range) slightly. Sometimes at the lowest ISO's there is a little variation in this, but the key word is little, and it isn't the same for every model so there is no general rule.
My advice is to follow Brian's advice: Don't worry about it!