Sample Preparation and Matrix Effects

As discussed earlier, preparing a liquid sample of a hair specimen is the first step in hair analysis. Enzymatic digestion of samples for testing in biological assays, coupled with RIA assays of the hair digests for the drugs cocaine, opiates, PCP, methamphetamine/methylenedioxymethamphetamine (MDMA), and cannabinoids, have been cleared by the FDA and used by this laboratory since 1985 (7,21).

Although solvent extraction methods might be expected to produce a sample more amenable to immunoassay than the digestion method, these methods can present serious challenges. A solvent extract of a hair sample will not contain keratin, for example, but it will contain a significant and likely variable amount of lipid. When the solvent is evaporated, which it must be because only a small amount is tolerated in any immunoassay, the lipid is left to be partially solubilized or suspended in an aqueous medium added to the dried extracts. Detergents can be added to the extract to aid in the solubilization of the lipid, but this must be carefully monitored and controlled to avoid damaging the antibodies or enzymes in the subsequent immunoassay. Too much detergent can affect primary or secondary antibody binding, or can cause detachment of antibody bound to solid phase such as in microtiter wells. Variations in amounts of lipid among different hair samples, and in micelle formation when reconstituting samples in aqueous medium, can cause great variability among negative samples. Such reconstitution issues can even cause large variability among replicate samples of the same hair specimen. Nonspecific matrix problems such as these can be a serious limitation to solvent extraction in achieving the analytical goals of the screening test, because they impact precision, accuracy, and sensitivity.

As an illustration of the matrix effects in an assay of enzymatically digested samples, Fig. 1 shows the distribution of a population of 100 different

LOD = 91.9 B/Bo x 100 for the negative population shown. Mean of the negatives spiked at 5 ng cutoff = 52.4 B/Bo x 100 1 S.D. of the spiked negatives = 2.1 B/Bo x 100 3 S.D. Range at Cut-off: 46.1 - 58.7 B/Bo x 100

Populations are separated by more than 30 B/Bo x 100 units

Jj llll

108 105 102 33 36 33 30 ST 84 81 78 75 72 63 66 63 60 57


42 33 36

Fig. 1. Matrix effects among different samples and their impact on precision at the cutoff concentration in a screening assay for cocaine in hair.

Table 1

Intra-Assay Precision of a Radioimmunoassay (RIA) for Cocaine

Table 1

Intra-Assay Precision of a Radioimmunoassay (RIA) for Cocaine

% of Cutoff



Minus 50

Minus 25


Plus 25

Plus 50

Cocaine (ng/10 mg hair)







RIA Response (% B/Bo)






















hair samples with no drug and with drug at the cut-off concentration in a cocaine RIA used in the authors' laboratory. The distribution of 100 negatives is shown in the histogram nearest the y-axis, and the distribution of these same negatives spiked with cocaine at the cut-off concentration is shown to the right in the figure. If there is a great variability in the responses of the negative samples (termed the B0, which is the amount of binding in the absence of nonradio-active drug), this variability will likely also occur at the cut-off, creating greater uncertainty in the correct identification of samples containing the cut-off concentration of cocaine. In this assay, the mean of the negatives shown was 99.1% B/B0, with a standard deviation (SD) of 2.4. (% B/B0 is the response of the unknown divided by the negative or B0 reference, expressed as a percent). The spread of such a population of samples is a result not just of matrix effects, but also of the many factors that affect precision. One can estimate the contribution of matrix differences among different samples to this spread by comparing the precision of replicates of the same sample at zero and at the cut-off concentration of drug. In this case, the mean of 20 replicates of the same negative sample had a mean of 98.4% B/B0 and a SD of 1.74 (Table 1), indicating that the matrix effects were quite small, because the precision among the 20 replicates of the same samples had nearly the same amount of error.

Figure 1 also illustrates another desirable feature of a screening assay (i.e., a clear separation between the negative population and the population at or beyond the cut-off). In this example, the lower edge of the negative 3 SD distribution of the zero-drug samples is a full 30% B/B0 units above the upper edge of the 3 SD distribution of the samples at the cut-off. Note that at the cut-off there will always be one-half of the samples falling above and one-half falling below the cut-off. A sizable separation between the negatives (zero drug) and the cut-off must not be achieved, however, at the expense of operating in the optimal region of the assay. An assay usually has a working range for quan-titation purposes of one to two orders of magnitude at best, with the optimum precision in the steeper part of the curve (in the case of a competitive RIA or EIA). Although assays using only a cut-off calibrator do not require a full dose-response curve, knowing the nature of such a curve is helpful in determining the optimum point for the cut-off. Placement of the cut-off in the most linear region of the curve facilitates achieving maximal precision at the cut-off and at points 25% and 50% above and below the cut-off.

Achieving acceptable statistical precision for samples containing drug at ±25% and ±50% of the cut-off has been a challenge for hair-screening assays. Controlling matrix effects, especially at the levels of sensitivity required, is likely the largest single factor in doing so.

Dealing With Drugs

Dealing With Drugs

Get All The Support And Guidance You Need To Be A Success At Dealing With Drugs. This Book Is One Of The Most Valuable Resources In The World When It Comes To A Parents Guide To The Drug Talk.

Get My Free Ebook

Post a comment