Sampling
Drugs submitted to forensic science laboratories can vary enormously in the manner of presentation. One exhibit might be a quantity of ecstasy tablets that are well made and essentially clean to handle, and the next might be packs of diamorphine or cocaine that had been concealed internally by a drug user. The initial approach to each scenario depends on such matters as:
• health and safety
• linking of packaging material
• fingerprint analysis
• other evidence (e.g. DNA).
It is also true that, because of time constraints and costs, not all of the items submitted will be analysed and therefore a sampling policy must be established. The number of items tested is determined by a number of factors, such as the number of items seized, their physical appearance and the need to satisfy a court of law. Exhibits of controlled substances may be received by forensic laboratories in large numbers, particularly where trafficking shipments have been intercepted (Fig. 1), and a representative sampling plan must be established. The benefits of a sampling plan are to:
• reduce the number of analytical determinations
• reduce the overall workload
• decrease exposure to controlled substances
• reduce handling of biologically contaminated evidence.
With any sampling plan, an initial visual examin ation of all the units in the population is conducted. If all the units are the same in appearance, the population can be considered homogeneous and a sampling plan can be implemented (Clarke and Clark 1990; Frank et al. 1991; Tzidony and Ravreby 1992; Colon et al. 1993; Aitken 1999). The most frequently used sampling procedures are as follows: 1. Take a sample (n) equal to the square root of the population size (N), that is n=N (e.g. from a population of 400 take a sample of 20). 2. Take a sample equal to 10% of the population size, that is n 0.1N (e.g. from a population of 400 take a sample of 40).
Figure 1 Example of a large-scale drug seizure illustrating the potential workload that can arise for a drug laboratory and why a sampling plan may be necessary. (Photograph: US Drug Enforcement Administration.)
Although experience has shown that the square root method produces reliable results, it does not provide a statistical foundation to the sampling problem. Another popular method of sampling is to use a hypergeometric distribution. This procedure involves statistical tables that can be used to determine how many samples should be taken for analysis. Table 3.3 is an example of such a table. If after testing these samples are positive, an inference can be made for the whole population. For example, if 1000 suspect tablets are seized and sent to the laboratory, if 28 tablets are tested and if the tests are positive for MDMA, there is a 95% probability that 90% of the tablets contain MDMA.
In relation to sampling from single items, it is important that a correct sampling procedure be put in place. The principal reason for a sampling procedure is to produce a correct and meaningful chemical analysis. Most methods qualitative and quantitative– used in forensic science laboratories for the examination of drugs require very small aliquots of the material, so it is vital that these small aliquots be entirely representative of the bulk from which they are drawn. Because of various legal requirements it may not always be possible to render a sample into a homogeneous state before sampling. This could be the case if the sample has to be tested separately by another laboratory or if the sample in its original ‘heterogeneous’ state is an important piece of physical evidence.
As a rule, however, items should be in a homogeneous state prior to sampling, which is especially important in quantitative analysis. This may involve a simple grinding of a small amount of powder (or tablet) in a mortar and pestle, or may involve taking numerous samples from a larger quantity of powder and grinding them together.
