I know this is an old post but I'm a noob here and I thought I'd try to break it down a little bit for my and everyone's understanding because i find this interesting. Not sure if this will be interesting at all to most but maybe some people will appreciate this from someone with a basic understanding of HPLC. I'm not trying to criticize the lab operator, just trying to explore possibilities of how the data may not always be 100% representative of the analyte.
-------------------------------------------------------------------------------------------------------
I cant exactly use the raw data to recalculate the results because I don't have the calibration data such as a calibration curve (which i think is only one 20ng injection of test and var each so it's not a curve at all). It would be nice to have the integration of that 20ng test peak. Usually, a series of a standard in serial concentrations are injected to plot the concentration of a sample vs the signal so that the sample signal can be compared to the curve to get the concentration. Ideally, you only count data that's in the range of the calibration curve, if the signal is too high and outside the curve you should dilute the sample further or its called 'data extrapolation'. In this case it seems all the data is in a sense 'extrapolated' rather than interpolated.
In this case we have the HPLC to separate compounds and the UV spec to measure the signal (amount) of those individual compounds. The peaks are overlapping as different colors because an ion is fragmenting the molecule apart to produce a unique 'fingerprint' of parent/daughter compounds of what you are analyzing. That's why the areas need to be added back together to get the total quantitation. This area is compared to the area of the known 20ng test sample to estimate the mass of the steroid samples in the 100-fold ethyl acetate dilution. Whats important if you use this method instead of building the curve is that the signals are comparable to the standard, which I'm not sure about. That's why you build a curve. My guess is he did a direct inject of 20 ng test and used it as a single external standard. That worries me because the pills were all extracted (chloroform, washed with h2o x2) and it's better to use an internal standard (spike) so you can know the recovery of the extraction. These issues are mostly irrelevant in this case where we have a 10 fold discrepancy in the expected result from every sample 6-12 rather than only 10-20%. I'm assuming the T4-T3 was also off by one order of magnitude +/- 25%.
---------------------------------------------------------------------------------------------------------
The scientist is obviously very experienced with HPLC-UV (much more than I). The data is great, look at how close the duplicates are to each other. BUT, it wouldn't surprise me if a dilution mistake was made. Usually, if things are off by nice even factors like 2, 5, 10, etc. it's a dilution error. I mean, even right there he admits in sample 6B he got a 5 fold higher concentration than was expected 'for some reason'. I'd like to know if he double checked his sample prep and quantitation for #6-12. Bottom line is that it's easier to make a miscalculation when pipetting microliter aliquots and doing several dilutions, than shoving 300 g of a powder in a liter of oil and letting it sit on a magnetic stir hotplate. I've done plenty of both. Then again one's a phd and the other is probably a dropout and at the mercy of untested raws.
