This month, we’ll examine some of the types of controls frequently employed in molecular diagnostic assays—both their varieties and their utilities. Controls are by nature the materials or reactions incorporated in or added to a reaction or one of a group of reactions, which should provide a predetermined result. When the expected result for a control is obtained, an aspect or aspects of the test are confirmed as working within some specification; more importantly, when a control does not give the expected result, the test in question is shown to not meet expected performance. By clever use of multiple types of controls, it’s possible for the laboratorian to not only confirm appropriate test performance, but in the event of a test failure, to quickly identify which of many possible steps has not performed correctly. Such information is invaluable in troubleshooting to identify the root cause(s) of assay failure and corrective actions to be taken to re-establish proper performance.
When considering controls, it is important to examine exactly which steps of the MDx process they evaluate, and which they do not. Failure to do so can lead to erroneous conclusions as to the meaning of apparently passing (or failing) controls. A rule of thumb would be that the control material which most closely approximates a real target, in all features, will give the most complete validation. Any differences between nature or format of control material and real assay target should be thought of as potential places for the control to give misleading results if the differences are not consciously considered.
Different assay methods and platforms can have different types of controls. For simplicity let’s start by considering the case of a hypothetical, real-time probe-based quantitative PCR reaction for a viral target. Most of the controls we will consider in this context are generalizable to other assay types and platforms; we’ll follow up on this with a few other platform or assay type-specific examples.
Negative controls. Probably the easiest to obtain and interpret, as the name implies, they are a sample which does not contain the assay target, and which should provide a negative / no amplification signal. Simplest examples are use of dH2O as a PCR reaction template; when used in this form, downstream of sample extraction, this is best considered a “negative amplification control.” An expected, negative result here demonstrates there’s no contamination or spurious amplification occurring arising from the basic PCR “master mix.” A more informative negative control would, however, consist of a true blank sample matrix, processed through the sample extraction method, in parallel with the test samples. An expected negative result here would also help to validate for the test run in question that contamination or spurious amplification is not arising from something in the sample matrix or arising during the extraction process. An even more complete (but somewhat harder to produce and validate) negative control might consist of a target negative human tissue culture sample, processed through extraction. While this would add a level of certainty that endogenous host (human) nucleic acids don’t cause unexpected positive signals, it can be reasonably argued that this shouldn’t be a variable attribute, and if non-positivity of the assay in the presence of host DNA has been effectively shown in the initial assay validation, it does not need to be proven again each assay run. Either of these latter two options would be best considered as a “negative process control.” Note that some assays may employ both a negative amplification control and a negative process control; doing so would allow for almost immediate identification of extraction step contaminations.
Positive controls. These come in a wider range of forms than negative controls. In the context of our current example, starting from the controls evaluating the least steps in the assay to those evaluating the most, these can include:
Internal controls. These are most commonly in the form of a predetermined template and matching primers (and probe, in our present model) which are included in the reaction “master mix” for all reactions. By necessity, this requires an assay system capable of multiplexing at least to the level of two targets (assay target and internal control). While care must be taken in the assay design and validation to ensure that such a control does not appreciably reduce sensitivity of the assay for its target, and that the internal control- specific primers do not somehow lead to spurious false positive results for the assay target, the internal control can be one of the most useful controls available. A positive internal control signal not only demonstrates proper activity of the polymerase, shared master mix components, appropriate cycling conditions, and function of the detector system; it also uniquely has the ability to highlight sample-specific PCR inhibition. Arising from the carryover of any of a number of potentially PCR-disrupting substances such as heme, chelating agents, or SPS during extraction, PCR inhibition is not infrequently encountered in the clinical laboratory and, unless tested for, can lead to false negative results. The importance of this form of control cannot be overstated, particularly if the specimens are of certain types prone to inhibition (e.g., stool samples).
Where multiplexing is not possible in the platform used, a slightly less useful alternative to the internal control is to split a test sample into paired aliquots post-extraction, running one as the specimen test, and spiking a PCR positive control template (see below) into the paired reaction. While this fails to evaluate the actual individual test reaction and brings risks of template contamination, it still maintains the capacity to evaluate a specimen for inhibitory substances through failure of the spiked sample to amplify. (A template / primer / probe set not identical with the assay target can also be used for this function, and avoids the template contamination risk.)
The actual CT (CP) value of the internal control (or paired reaction cousin) can also be useful. Significant variation of this from its normal value toward higher values can indicate the presence of incomplete inhibition in a test sample, or a partial loss of function in an assay. It’s worth noting that this can apply to any real-time assay format, even those for which the primary target is qualitative in nature, and would indicate a loss in Limit of Detection (LoD).
PCR positive controls. The next level up in our example system would be a PCR positive control, analogous to our negative amplification control. Usually, this consists of a simple DNA template material for the assay target, added to its own dedicated PCR reaction run in parallel with the clinical samples on a test. A positive test result in this reaction validates the function not only of the shared master mix components tested by an internal control, but also uniquely shows the function of target-specific primers and probes. Upstream processes such as extraction are not tested by this form of control.
Quantitative standards. Most frequently these represent a special case of PCR positive control, and consist of a set of PCR positive controls of known concentrations. Taken as a set, they form the basis for a standard curve, as described in our prior installment on quantitative assays. Occasionally, one finds these formulated as intact assay targets (e.g., titred virus, in our current example) in sample matrix; this allows them to also serve as an extraction control (see below), and allow the quantitation standard to take into direct account possible variations in extraction efficiency across a range of target concentrations.
Extraction controls. These are a standardized sample of real, or closely simulated, assay target. For our current case of a viral detection assay, this could be titred live virus; for a more easily handled substitute this can be replaced by chemically inactivated and stabilized virus. A further variation on this can exist in the form of assay-specific target nucleic acids (not complete organisms), artificially encapsidated in something similar to a live organism. Both of these sorts of materials are available from commercial suppliers of assay standard materials, for many common assay targets. Regardless of the exact form used, a positive extraction control signal validates extraction processes, and all downstream steps (amplification, detection, etc.). Depending on how an assay workflow is designed, the line between extraction controls and process controls can be blurred and functions combined.
External controls. Usually in the form of real, known positive (or negative) clinical samples, to be treated in parallel with test specimens, external controls represent the most complete control type which can be used. The difficulty in obtaining or using them lies in the term “known”; effectively, these samples must have been reliably evaluated by an alternate gold standard reference method to the assay. The availability and ease of such an alternate reference method, and reliability / uniformity of suitable external control material, can vary greatly across platforms and assay types in question.
The above represent some of the most frequently encountered control types in a quantitative real-time PCR for pathogen detection. Most of these control types exist identically in other assay types and settings with minor changes. Some of the more common other types of controls seen in alternate assay settings include:
Hybridization controls. In an assay with an array-based detection system, a defined labeled oligonucleotide may be included in the hybridization buffer. Complementary in sequence to one of the array spots, this control allows for evaluation of hybridization, wash, and detection steps of the assay separate from all upstream processes such as extraction and amplification.
RNA specific controls. The intrinsic low chemical stability of RNAs and ubiquitous nature of RNases make RNA-based assays particularly susceptible to template degradation, leading to false negative results. The types of controls discussed above cannot address whether such sample decay has occurred prior to a specimen being received; for some particular types of RNA samples, this question may be assessed post-extraction through instrumental analysis of parameters such as the RIN (RNA Integrity Number, beyond the scope of our present inquiry). Other controls can, however, be used to confirm that whatever RNAs were intact at the time of specimen receipt are being successfully recovered and detected. These can take the form of RNA-based extraction controls, RNA-based process controls, and/or RNA-based external controls—simply RNA analogues of the DNA versions discussed above.
In the common context of an RT-PCR-based assay where the first post-extraction step is reverse transcription of RNA to DNA, it is not essential that all system controls be RNA-based, as long as at least one form of included control validates the reverse transcription (RT) step. Because of the real potential for RNA control materials to decay and lead to apparent assay failures, some assay developers prefer to use more stable DNA-based control materials for RT-PCR assays. This simplifies assay kit handling and storage conditions and can avoid spurious control failures. At a minimum, at least one RNA-based control (extraction control, process control, or amplification control) is essential to confirm performance of the RT step, and possibly preceding steps depending on control form. The formulation of this RNA control relates to where it can be used. For example, an in vitro transcript RNA species can be a good RT-PCR positive amplification control when added to samples post-extraction (or at least, after extraction steps designed to inactivate RNases, usually, the initial lysis step). Adding this same “naked” RNA species into raw samples pre-extraction would be a poor control, as it’s likely to be rapidly degraded. Encapsidated RNAs or live or inactivated RNAs (such as RNA bacteriophage) can, however, be added directly to raw samples and serve as extraction or process controls.
Sequence controls. For assays such as genetic polymorphism (SNP) analysis, control templates for each of the analyzed genotypes are generally run to assure appropriate classification and discrimination between very similar targets.
The above has summarized (and hopefully explained the utilities of) several of the more common forms of molecular assay controls. For a given specific assay or platform there may be variations on or specific additions to the types discussed here, but in all cases an understanding of the exact nature of the control—what it tests and what it assumes—will help in evaluating its results. Readers whose interest has been piqued by this topic and looking for a much more technical and detailed discussion of MDx controls from a regulatory perspective are recommended to look up CLSI guidelines specific to the assay class of interest.
