Aside from the accurate measurement of BCR–ABL1 transcript numbers within each sample, a control gene is also quantified routinely. The term housekeeping gene is not preferred. Control gene measurement is crucial in demonstrating RNA quantity and quality and ensuring specimen acceptability by highlighting any potential RNA degradation. It allows for the BCR–ABL1 transcript numbers to be reported as a ratio and also helps to establish the sensitivity of each assay. All things being equal, the sensitivity of an assay increases with the number of measurable control gene transcripts [24] (Fig. 7.3 and Table 7.1).

Consensus recommendations have been published with regard to the essential characteristics of good candidates of a control gene [7, 16, 23, 25–27]. These include (i) similar level of expression between different individuals and between different cell types, including tumour versus non-tumour cells; (ii) expression levels at similar levels to the target gene (in this case, BCR–ABL1 at diagnosis); (iii) similar rate of stability and degradation as the target gene (i.e. constancy of the ratio of target to control); and (iv) lack of similar sequences (such as pseudo-genes in genomic DNA) that may be amplified with primers used in the proposed assay.

In the context of BCR–ABL1 measurement, BCR, ABL1 and GUSB (β-glucuronidase) are the most commonly used control genes. BCR has similar expression levels and stability to BCR–ABL1 [28] and was the control gene used in the molecular assays that accompanied the IRIS study – the first phase III study using imatinib in CML treatment [29]. ABL1 is the recommended control gene of the EAC and is the most commonly used of the three [30]. Laboratories may use primers that amplify exons 2 and 3 of ABL1 in qPCR reactions, which may lead to a loss of linearity at high levels of BCR–ABL1 (Fig. 7.4). This issue becomes important when one wishes to calculate the velocity of BCR–ABL1 reduction in the initial months following commencement of TKI treatment [31]. However, for most specimens, the underestimation of BCR–ABL1/ABL1 at the high end of the scale is of no clinical consequence. BCR–ABL1 transcript numbers are usually significantly lower than ABL1, especially for the important cut-off of MMR at 0.1 % IS. Similarly, GUSB transcripts have been demonstrated to have similar stability to BCR–ABL1, and it has the advantage of not being involved in the formation of the BCR–ABL1 rearrangement, and some authors would favour using this as their control gene [23].

One important analytical variable affecting the measured transcript numbers is the determination of the standard curve, as generated for each RT-qPCR assay through accurately quantified calibration standards. The limited availability of stable biological material able to act as a reference standard to calibrate local assays has hampered both the development of international external quality assurance programmes and further enhancements of assay comparability. An ideal reference material should be available in adequate quantities to allow easy access, be stable in transport and storage, homogenous across batches, and yet replicate the characteristics of the primary test material as closely as possible under assay conditions. No material currently meet all these criteria as far as BCR–ABL1 testing is concerned. In the absence of an internationally recognised calibration standard, laboratories produce standards individually or in collaboration with others, using serially diluted plasmid mixtures of known quantity.

The WHO International Genetic Reference Panel first released a reference material in 2009, consisting of freeze-dried K562 cells positive for the e14a2 transcript diluted in BCR–ABL1 negative HL-60 cells in four different dilutions. Each is assigned a reference BCR–ABL1 IS value after repeated testing [32]. However, the small quantity of the material available is released mainly to the manufacturers of secondary reference materials. Although plasmids cannot simulate the relevant biological test material at the RT stage, it is relatively easy to produce in larger quantities, and recent interest has increased in using this in substitution for cells. The performance characteristics of a certified plasmid reference material had recently been described [20]. This plasmid, pIRMM0099, contains sequences of BCR–ABL1 e14a2, BCR, ABL1 and GUSB. Six tenfold serial dilutions on a panel called ERM-AD623a-f had been independently quantified and assigned values by digital PCR before being tested in 63 laboratories and are currently distributed by the Institute for Reference Materials and Measurements in Belgium. An alternative to plasmids involves the use of BCR–ABL1 RNA. One proprietary product developed by Asuragen is called Armored RNA Quant or ARQ. RNA is usually labile and subjected to degradation from ubiquitous RNAses present in the environment and biological materials. This may be minimised by packaging RNA molecules inside protective protein envelops. The use of stabilised RNA as a reference material allows the efficiency of the RT step to be included in the overall quality assessment. Preliminary results of a validation project have been recently published [33].

The lack of internationally recognised suitable standard materials not only leads to difficulties for laboratories to assign values to their standards, but has also hampered direct comparability of results between different laboratories. Variations in reagents, machine platforms, control genes, primer/probe sequences and calibration standards lead to different values being reported by different laboratories for the same specimen [34]. One study demonstrated that, even when common primer/probe sets and plasmid standards were used with optimised methods, significant variation in reported BCR–ABL1 values still occurred when testing the same sample amongst 37 European laboratories [35]. Thus, longitudinal comparisons of treatment response within the same individual were not possible when results came from different laboratories. Furthermore, correlation of survival outcomes with achievement of particular molecular treatment milestones would not be possible without inter-laboratory and inter-patient comparability.

A number of efforts have been made to standardise pre-analytical and analytical variables in order to minimise assay variability. These include the use of published consensus primer and probes sets from the EAC for targets and control genes [16, 30], as well as a set of consensus recommendations intended to harmonise other aspects of BCR–ABL1 RT-qPCR methodology [23] (Fig. 7.2). One key proposal involved the development of laboratory specific conversion factors that allowed results to be adjusted and reported on a common International Scale (IS), enabling valid comparisons across individuals and laboratories [36, 37].

The initial steps of standardised RT-qPCR reporting first began with the IRIS study, the first to report molecular results following treatment with a TKI. Baseline samples from the same 30 patients were measured independently by three laboratories in London, Seattle and Adelaide, and the median value was taken to be the standardised diagnostic baseline and given a value of 100 %. Reduction in BCR–ABL1 values for subsequent results from all IRIS patients was measured as log₁₀ reductions from this value and expressed as a ratio of BCR–ABL1: BCR [29]. Achievement of major molecular response (MMR) was designated for patients who achieved a 3-log reduction in transcript numbers from this standardised baseline and is equivalent to a BCR–ABL1 value of 0.1 %. Subsequent major international standardisation efforts involved sample exchanges between a number of central laboratories and peripheral laboratories who performed assays in parallel. Peripheral laboratories were expected to have optimised their methods and minimised assay variations prior to participation in the project, which led to the determination of a conversion factor, permitting participating laboratories to report results on the International Scale [36, 37]. Through this process, concordance between the central laboratory and a participating laboratory on whether a sample has BCR–ABL1 ≤ 0.1 % (MMR) reached 91 % at best, although only about half of the participating laboratories were deemed to meet the desirable performance characteristics [36]. Inherent variability of the assay or suboptimal method development was responsible for discordance in the remaining laboratories.

Currently, conversion factors only enable e13a2 and e14a2 results to be adjusted for reporting on the IS, and a mechanism to report atypical transcripts such as e1a2 and e19a2 using a similar scale does not yet exist.

RT-qPCR testing is currently recommended for all patients at diagnosis. The type of transcript expressed should be documented at diagnosis using an assay that can identify atypical transcripts in cases that do not express the more common e13a2 and e14a2. Thereafter, RT-qPCR should be performed every 3 months until MMR is achieved. Loss of MMR or CCyR is rare for patients with BCR–ABL1 ≤ 0.1 % [2], and RT-qPCR monitoring every 3–6 months for these patients is adequate [3]. Management decisions based on molecular monitoring should be made only when the inherent variability of an optimised RT-qPCR assay had been considered. The technical variability is potentially equivalent to a 2-fold variation of the reported result for a sample with 0.1 % BCR–ABL1. At the higher level of 10 % BCR–ABL1, the CV is slightly lower (~18 %), and at 0.0032 % the variation approaches ~5-fold (both unpublished observations within our laboratory). Although these CVs may be wide when compared to more commonly used diagnostic assays such as haemoglobin or serum sodium, this degree of variability is expected given the dynamic range of the assay and the clinical context (i.e. a very small change in the expression of BCR–ABL1 does not have the same physiological significance as an out of range serum sodium). Changes in a patient’s BCR–ABL1 value should be interpreted within this context to determine whether a change is within the measurement reliability of the assay or a true biological change [7]. A repeat assay should be performed when the significance of a BCR–ABL1 increase is unclear.

Several groups have explored the optimal cut-off in the BCR–ABL1 increase most likely to predict for an adverse outcome, though a consensus value had not been reached. The different conclusions are related to differences in the performance characteristics of individual laboratory molecular assays, and although a common value cannot be agreed upon, clinically relevant rises in BCR–ABL1 in many instances have been demonstrated to approximate the precision limit of the assay in the investigating laboratory. Furthermore, loss of MMR in the context of a rising BCR–ABL1 level increases the clinical significance of the rise [38]. For instance, Press et al. demonstrated that in their cohort of 90 patients, a rise in the BCR–ABL1/G6PDH ratio of >0.5 log₁₀(~3.2-fold) was associated with a loss of CCyR. Using a lower cut-off in the BCR–ABL1 rise led to a higher false-positive rate, as the contemporaneous variability in RT-qPCR in their laboratory due to technical variations was 0.46 log [39]. The same group subsequently concluded that a 2.6-fold increase in BCR–ABL1 was the optimal cut-off for predicting the development of a concomitant kinase domain mutation and reported that a 2.6-fold change was the analytical precision limit for that particular analysis. In contrast, the Hammersmith group found that a rise of BCR–ABL1/ABL1 was predictive of loss of CCyR and progression to advanced phase, providing that the resultant BCR–ABL1 was >0.05 % [40]. We would regard a rise in the BCR–ABL1 of >2-fold in our laboratory as significant and in line with current recommendations would proceed to mutation analysis if this results in a loss of MMR [38]. Other aspects that can lead to therapeutic failure should also be addressed, such as compliance or concomitant drug interactions (see mutation analysis below). Patients with a >2-fold rise, but not associated with loss of MMR, should be monitored more closely.

Patients who fail to achieve time-dependent molecular targets are at higher risk of disease transformation and death. According to the current recommendations of two most widely used CML treatment guidelines, both the ELN and the NCCN would regard the failure to achieve BCR–ABL1 ≤ 10 %, ≤1 % and ≤0.1 % at 3, 6 and 12 months, respectively, after a CML-CP patient commences treatment with a TKI to be associated with inferior outcomes [3, 41] (Fig. 7.5). The evidence with regard to the 3-month time point is particularly strong, and patients with BCR–ABL1 > 10 % at this time point have inferior overall and progression-free survival as well as inferior achievement of molecular responses, regardless of the TKI chosen for treatment [42]. The level of uncertainty with regards to BCR–ABL1 measurement is of particular importance when an individual patient’s molecular result is measured against recommended milestone responses. In particular, even though the guidelines are specific as to the cut-offs that segregate high-risk patients from good-risk patients, in reality the association between the BCR–ABL1 ratio and the degree of risk exists as a continuum. For instance, a patient with BCR–ABL1 of 1 % 3 months after starting TKI therapy clearly has a lower risk of disease progression compared to a patient with a result of 10.5 % at 3 months. However, it is unclear if a patient with 9.5 % has a clearly different level of risk to a patient with 10.5 %, yet the response level and risk are assessed as different using an absolute cut-off. Taking into account the underlying inherent assay variability, a reported BCR–ABL1 of 9.5 % may actually reflect a true value that ranges between 6 and 12 % (given a CV of 18 %). Consequently, many would urge caution when making clinical decisions based on one single BCR–ABL1 result, and a patient’s history should be considered when management decisions are made. Furthermore, the ELN cautioned that a single measurement at 3 months is insufficient for decisions regarding a change of treatment. The ELN recommends repeat testing at up to monthly for patients with BCR–ABL1 > 10 % at 3 months and changing treatment for patients who are still >10 % at 6 months [3].

The speed at which a patient achieves milestone responses had been demonstrated to add valuable prognostic information. This is encapsulated in the concept of the “BCR–ABL1 halving time” which the Adelaide group has used to further refine prognosis in patients with BCR–ABL1 > 10 % at 3 months. This takes into account both the magnitude of the fall in BCR–ABL1 transcript numbers and duration over which this occurred. The initial fall in BCR–ABL1 transcripts after TKI commencement is linear when BCR is used as the control gene, and a slope can be calculated using these two parameters. From their cohort of 500+ imatinib-treated patients, the number of days necessary for the BCR–ABL1 to reduce by 50 % can be calculated providing both results at baseline and 3 months are available. Patients with BCR–ABL1 > 10 % at 3 months and a halving time of >76 days are at particularly high risk of treatment failure [43]. The German CML study group arrived at the same conclusion that the velocity of BCR–ABL1 transcript elimination in the early months of treatment is correlated with treatment outcomes and concluded that failure to achieve a half-log decrease in transcript numbers is associated with inferior progression-free and overall survival [31].

The ELN and NCCN guidelines also set down time-dependent cytogenetic responses which correlate with clinical outcomes. If cytogenetic results are unavailable (either for technical or clinical reasons), molecular responses can be used as a surrogate. A BCR–ABL1 ratio of <10 % is generally considered to be equivalent to major cytogenetic response (Ph + metaphases 1–35 %; MCyR), whereas patients with BCR–ABL1 < 1 % have almost certainly achieved complete cytogenetic response (0 % Ph + metaphases; CCyR) [7, 44–46].

Given the stated CVs and measurement uncertainty with molecular assays, some have argued that cytogenetics should be the preferred method for monitoring treatment response in CML, especially because of the perceived greater certainty of cytogenetic responses. This is an erroneous assumption, however, as cytogenetics has actually been demonstrated to be associated with an even greater level of measurement uncertainty, especially with regard to cut-offs important for clinical decision making [47]. Given this context, molecular and cytogenetic data should be used in combination, especially when the limited dynamic range of cytogenetic analysis is considered.