NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Booth RA, Ansari MT, Tricco AC, et al. Assessment of Thiopurine Methyltransferase Activity in Patients Prescribed Azathioprine or Other Thiopurine-Based Drugs. Rockville (MD): Agency for Healthcare Research and Quality (US); 2010 Dec. (Evidence Reports/Technology Assessments, No. 196.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Assessment of Thiopurine Methyltransferase Activity in Patients Prescribed Azathioprine or Other Thiopurine-Based Drugs

Assessment of Thiopurine Methyltransferase Activity in Patients Prescribed Azathioprine or Other Thiopurine-Based Drugs.

Show details


This review was nominated by the American Association for Clinical Chemistry (AACC), and commissioned by the Agency for Healthcare Research and Quality (AHRQ), to examine testing for thiopurine methyltransferase (TPMT) enzymatic activity (phenotype), and allelic polymorphism determination (genotype) in chronic autoimmune disease, with an overall view potentially to optimize use of thiopurine medications. The objectives were to examine the analytical aspects of TPMT status determination (genotype or phenotype); and resultant change in patient management and clinical outcomes of thiopurine toxicity in light of pretreatment knowledge of TPMT status; and costs associated with TPMT testing, along with costs of adverse events arising from thiopurine toxicities. Inclusion of evidence to meet these objectives was restricted to chronic autoimmune disease populations.

TPMT Status Determination

TPMT status may be determined through either genetic or TPMT enzymatic activity analysis. Genetic analysis involves detection of variant alleles coding for TPMT enzymes with reduced enzymatic activity, while enzymatic assays are able to determine directly the activity of the TPMT enzyme. Both genotyping and phenotyping can be determined from routine blood specimens, as white blood cells providing the required genetic material and red blood cells as the source of the TPMT enzyme. From a clinical laboratory perspective, determination of TPMT status encompasses preanalytical, analytical and postanalytical phases of testing.

Preanalytical testing requirements are often overlooked while focusing on analytical performance. However, the majority of errors in laboratory medicine occur during the preanalytical phase.220 The preanalytical phase of testing covers from the moment a specimen is collected to the point that it is analyzed, and includes sample collection method, anticoagulant used, transportation conditions, time between specimen collection and analysis, storage, specimen preparation, and preanalysis storage time and conditions. If appropriate preanalytical conditions are not met, then significant error can be introduced. These sources of error are often not recognized, as they are not identified by routine quality control monitoring of the assay. We therefore included examination of relevant preanalytical requirements, and potential confounders.

The analytic phase of testing involves the actual analysis of enzymatic activity or detection of variant TPMT alleles. Each method has individual analytical performance characteristics including precision, reproducibility, and diagnostic sensitivity and specificity, which are reviewed here.

Postanalytical requirements generally involve procedures to report results.

Preanalytic Requirements and Sample Stability

Thirteen studies examined relationships between storage conditions and TPMT enzymatic activity, with mixed results.108,115,116,118,121,124,127–129,131,134,136,137 Six studies reported TPMT to be stable at room temperature in anticoagulated whole blood for periods up to seven days,108,118,121,124,127,134 whereas in one study, reported only as an abstract, TPMT activity decreased by 25 percent over 24 hours.137 Similarly, in red blood cell (RBC) lysate stored at −20°C for 3 months, two studies reported TPMT activity to be stable,108,124 while one study reported a seven percent decrease.128 Four studies reported that TPMT activity was stable in heparinised whole blood,127,134, EDTA whole blood124, or an unspecified anticoagulant136 at 4°C. When RBC lysate was stored at −80°C, TPMT enzymatic activity was reported to be stable for up to 25 days, whereas a 15 percent decrease in activity was measured after 16 months.128,129,131

One explanation for these disparate results is that only one study was actually designed to evaluate the effect of storage on TPMT activity.137 The available data suggests that TPMT activity is stable in EDTA and heparin anticoagulated whole blood for up to 7 days at room temperature or 4°C. RBC lysate is stable for 3 months at −20 °C. Longer storage should be at −80°C, although in the range of 15 percent of TPMT enzymatic activity may be lost after 16 months. Repeated freeze-thaw cycles were reported to decrease the results by up to 16 percent, although the drop was not statistically significant.115

No studies were identified that addressed any preanalytical requirements for TPMT allelic polymorphism determination. However, since preanalytical requirements are common for genetic testing, previously published guidelines can be used. The Clinical and Laboratory Standards Institute (CLSI) has published excellent guidelines covering all preanalytical requirement for collection, transportation, preparation and storage of specimens for genetic testing.221

Six of the seven laboratories asked to participate in the survey returned responses, three from Canada and three from the United Kingdom. Among the responding, yearly volumes ranged from 50 to 1500 allelic determinations and 600 to 19,000 enzymatic TPMT determinations.

TPMT Variation Amongst Patient Populations

Gender, age, and race. All studies reported no gender difference for TPMT. One study of a small sample size reported a difference between TPMT enzyme activity of whites and mixed race that was not statistically significant.129 Of ten studies, a single report showed a significant difference (p less than 0.001) in TPMT enzymatic activity between 192 children (12.0 U/mL RBCs (range 0.6 to 25.4 U/mL RBC)) and 959 adults (12.9 U/mL RBCs (range 0.2 to 24.6 U/mL RBCs)).111 However this difference was small and not clinically relevant. Two studies observed no significant differences in TPMT activity across races, including blacks, whites, mixed races, and Japanese.109,129 However, more races with appropriate sample sizes should be included to confirm the lack of racial differences. One large study published in July 2004 in the journal Pharmacogenetics was excluded from the review on population. It analyzed 1200 healthy German individuals and demonstrated a statistically significant difference in TPMT activity between males and females. They also showed a statistial difference between smokers vs non-smokers; male and female smokers. However, clinically the differences are likely unimportant.

Coadministered drugs. Fifteen drugs (5-aminosalycilate, sulfasalazine, mesalazine, azathioprine, mesalamine, ac-5-aminosalicylate, syringic acid, prednisone, prednisolone, 6-methylprednisolone, cyclophosphamide, methotrexate, trimethoprimsulphamethoxazole, SKF 525-A, 3,4-dimethoxy-5-hydroxybenzoic acid, trimethoprim, vincristine, dexamethasone, L-asparaginase) have been evaluated in ten studies.102,104,105,110,114,117,119,120,128,131 Only six of the studies were conducted in vivo,104,117,119,120,131 in which no clinically relevant interactions were demonstrated.

Hematocrit. Three studies investigated the effect of hematocrit on TPMT enzymatic activity.104,121,126 Two reported a positive correlation of hematocrit with TPMT enzyme activity121,126 and one104 observed no difference when comparing high and low hematocrit levels. Although two studies did demonstrate a correlation of hematocrit with TPMT activity, the effect was small (less than 7 percent in the normal hematocrit range) and likely not clinically relevant.126 Standardizing TPMT measurement to grams of hemoglobin or milliliters of packed RBCs should correct for any significant effect of hematocrit on TPMT measurement.

Morbidities. Two studies assessed the effect of concomitant diseases on TPMT activity.104,106 Inflammatory bowel disease (ulcerative colitis, Crohn disease, or indeterminate colitis), autoimmune hepatitis, multiple sclerosis, myasthenia gravis, pemphigus and chronic renal failure were shown to influence TPMT activity. Although the differences between disease groups showed statistical significance, the differences were minor and not clinically relevant, with the exception of patients requiring dialysis. Patients with renal failure showed elevated TPMT enzymatic levels prior to hemodialysis, which dropped by approximately 50 percent following hemodialysis to levels comparable to normal individuals’. The mechanism responsible for the elevated TPMT activity prehemodialysis is unclear, but may involve unidentified TPMT activating uremic compounds.106 Although there are no comparative studies of harms in dialysis population directly evaluating TPMT testing pre- and postdialysis, the available evidence suggests that dialysis patients should be measured postdialysis, as the levels most closely match those that would otherwise be seen in them as healthier individuals. Measurement of TPMT activity prior to dialysis may result in falsely identifying a low/absent or intermediate metabolizer as a normal metabolizer, potentially placing them at increased risk of drug toxicity. The remaining disease states studied to date are organ transplant and acute lymphoblastic leukemia, which were not included in this review.

Analytic Performance

The enzymatic measurement of TPMT was originally developed by Weinshilboum et al134 and has since undergone only minor modifications. In brief, RBCs are concentrated by centrifugation, washed, resuspended and lysed to release the TPMT enzyme. The lysate is added to a buffered solution of radioactively labeled S-adenosyl-L-[14C]methionine and substrate 6-mercaptopurine (6-MP). TPMT methylates 6-MP to form radioactively labeled 6-methylMP which can then be measured. Modifications include use of 6-thioguanine monophosphate (6-TG) as substrate, or nonradioactive detection by high performance liquid chromatography (HPLC). The rate of product formation is dependent upon TPMT enzymatic activity, and is independent of the detection method - radiochemical or HPLC. Enzymatic assays using either 6-MP or 6-TG, regardless of the detection method (radiochemical or HPLC) were reasonably precise, with inter-assay coefficients of variance (CVs) of less than 10 percent in all cases. With an analytical coefficient of variance less than 50 percent of the biological variability, the amount of variation added to the true test variability is 11.8 percent.222 In comparison with other enzymatic assays, the currently achievable intra-laboratory CV for TPMT enzymatic analysis is better than the minimal acceptable performance for routine enzymatic analysis specified by the U.S. Department of Health and Human Services: Clinical Laboratory Improvement Amendments of 1988 (e.g. total creatinine kinase (CK) below 30 percent, or aspartate aminotransferase (AST) below 20 percent).223

Three studies that partially addressed the reproducibility and accuracy of variant TPMT allelic polymorphism detection reported 100 percent concordance between denaturing HPLC and restriction fragment length polymorphism (RFLP) genotyping tests. The dichotomous nature of genetic results, reported as either present or absent, does not allow for traditional precision and accuracy determination as is done for enzymatic determination. However, a number of guidelines are available that address the complex issues. Recently, the Centers for Disease Control released a Morbidity and Mortality Weekly Report detailing good laboratory practices for molecular genetic testing. It reviewed the need for adequate quality control of genetic testing and put forward recommendations for laboratories performing molecular genetic testing.224 The Clinical and Laboratory Standards Institute (CLSI) has also published guidelines for Molecular Diagnostic Methods for Genetic Disease225 and Validation and Verification of Multiplex Nucleic Acid Assays.226 It is recommended that any laboratories performing allelic polymorphism detection of TPMT review the above guidelines to ensure the accuracy of their results.

Diagnostic Sensitivity and Specificity

Enzymatic analysis was selected as the reference standard, as this method should identify all patients with reduced or absent TPMT activity, regardless of the mechanism. To date at least 30 mutant alleles have been identified within the coding region14,27–29,213,227–231. Others have been identified in the 3’ untranslated and promoter regions.29 These mutations likely do not directly affect the activity of the TPMT enzyme molecule, but may influence the quantity of enzyme present and thereby indirectly decrease the overall in vivo TPMT enzymatic activity.

Reporting units and ranges of enzymatic activity are not standardized, so in our analyses the activity cutoff values stated in each article were used to assign patients to one of three groups: low/absent; intermediate; or normal/high. None of the studies were specifically designed to determine the diagnostic accuracy of genotyping in comparison to enzymatic activity. Thus, not surprisingly, using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) tool35 a substantial (37 percent) of the studies were rated poor quality.

Diagnostic groups were organized in our analyses according to TPMT allelic variant(s) tested rather than by specific point mutation, to correspond to reporting in clinical practice. The relative abundance of specific allelic variants in a population has a direct impact on the diagnostic sensitivity and specificity of genetic testing, and therefore should be considered when developing genetic testing strategies. Sahasranaman et al. reviewed the relative frequency of the four most common alleles, *2A, *3A, *3B, and *3C.227 Pooled data suggests that the most common allele in Caucasians is *3A, with a mean frequency of 3.89 percent in a general population of 5076 (range 2.1 to 8.6 percent), while the most common in Africans is *3C, with a mean frequency of 4.7 percent in a population of 884 (range 2.4 – 7.6 percent). Pooled frequencies in a general population of 356 Asians and South Asians were lower than those seen in Caucasians and Africans, with a mean frequency of 1.0 percent (range zero to 2.3 percent) for *3C, and 0.17 percent (range zero to one) for *3A.

A total of 16 studies were included in the quantitative syntheses, assessing diagnostic sensitivity and specificity.1,46,49–53,56,59,70,93,157,158,161,162,167 From seven studies, the pooled sensitivity of genotyping for homozygosity or heterozygosity of the common TPMT *2, *3A, *3B, and *3C alleles, to correctly identify patients with absent to intermediate TPMT enzymatic activity, was 70.7 percent (95 percent confidence interval (CI) 37.9 to 90.5 percent) (meta-analysis 1). The pooled specificity of noncarrier genotype to correctly identify those with normal or high enzymatic activity approached 100 percent. With other combinations of alleles, pooled specificities of genotyping remained close to 100 percent, but sensitivity did not improve convincingly, as inadequate relevant evidence led to substantial imprecision in estimates [sensitivity ranged from 70.70 to 82.10 percent (95 percent CI, lower bound range 37.90 to 54.00 percent; upper bound range 84.60 to 96.90 percent]. Studies by Okada et al,53 and von Ahsen et al52 reported markedly lower sensitivities compared with the other studies. Okada et al analyzed a Japanese cohort, previously shown to have a low frequency of the common alleles, which suggest that other relatively common unidentified alleles may be present in the Japanese population. von Ahsen et al examined a German Caucasian cohort, and remarked upon the lower sensitivity observed, relative to previous reports. No plausible explanation could be identified for the heterogeneity in effect estimates.

Few individuals exhibited homozygous TPMT variant alleles. The pooled sensitivity of a homozygous TPMT genotype to correctly identify patients with low to absent enzymatic activity was based on two small studies with two percent of 341 patients identified as homozygous for variant allele. The pooled sensitivity was 87.10 percent (95 percent CI 44.30 to 98.30 percent). The pooled specificities of the noncarrier and heterozygous carrier states to correctly identify those without low or absent TPMT enzymatic activity were determined for the different combinations of tested allelic variants (meta-analysis 2). The specificities were high, approaching 100 percent.

For a screening test, genotyping appears to have moderate sensitivity to detect those with subnormal (i.e. intermediate plus low plus absent) enzymatic activities while possibly high sensitivity of 87 percent to identify only those with low to absent activities. However, the available evidence is imprecise and of uncertain validity given that 37 percent of studies were rated as poor during risk of bias categorization. These limitations in diagnostic sensitivity of genotyping are not unexpected as genetic analysis of TPMT most often targets only the common polymorphisms and will fail to identify new or rare mutations. Furthermore, the commonly employed genotypic tests while able to identify a specific SNP, are unable to determine the allelic location of it. Therefore, a patient typed as a heterozygote for TPMT*3A (i.e. wild type/*3A) may have been misdiagnosed as such while actually being a compound heterozygote TPMT*3B/*3C for the observed TPMT activity.232,233

As discussed later, with a dearth of relevant primary literature, it remains unclear how incidence rates of thiopurine related adverse events may be affected by pretreatment genotyping. Therefore, this evidence should not be interpreted to conclude that prior genotyping is not effective in reducing thiopurine related drug toxicity in the treatment of chronic autoimmune diseases, especially when data associated with homozygosity were scant.

Currently, there is insufficient data to determine the optimum combination of TPMT alleles that must be tested in order to identify patients with reduced or absent enzymatic activity. There is also a lack of well powered, good quality studies comparing the diagnostic accuracy and relative effectiveness of the two methods of genotyping and phenotyping to determine TPMT status.

Postanalytic Requirements

This review did not identify any relevant studies that addressed postanalytic requirements in terms of reporting units, common reference intervals, or result reporting for either enzymatic testing or allelic polymorphism measurement. In general, results should be communicated to ordering practitioners as soon as possible after testing is completed. However, as this test is normally used prior to administration of thiopurine drugs, there is no critical requirement to contact practitioners directly to communicate abnormal results. Ideally, reports should include both lower and upper reference limits, information on how the reference interval was determined, and an indication of overlap that is seen between normal and intermediate metabolizers. At a minimum, information on reference interval determination and overlap between intermediate and normal activities should be available on request.

Enzymatic assays are currently reported using one of two commonly used units, nmol/h/g Hb and Unit/mL RBC. One Unit is defined as generation of 1 nmol of product per hour (6-methylMP or 6-methylTG). Thus, the key difference is the standardization of the product generation, per gram of hemoglobin in RBC lysate, or per milliliter of packed RBCs. Hence, conversion between the two is neither easy, nor exact. It is recommended that a single preferred unit of measure be identified and used in the future to simplify reporting and interpretation, and to allow comparisons between laboratories.

Interpretation of enzymatic testing results is highly dependent on the stated reference interval provided by the performing laboratory. Individuals with very low or absent TPMT enzymatic activity (homozygous abnormal) are relatively easily identified, as they are clearly separated from patients with normal activity. Those patients with intermediate activity, however, are more difficult to identify as theirs’ often overlap with the enzymatic activity of normal metabolizers. Therefore, determination of the lower reference limit for normal metabolizers is most important. Currently, there is no universally agreed upon lower limit of normal for TPMT activity, however many studies used similar lower limits. Standardization of analytical methods and reporting units will aid in identifying a universal lower limit of normal, and should be a future goal.

Clinical Laboratory Survey

A survey of laboratories was conducted to gather information regarding current clinically available TPMT analysis. Following identification of potential laboratories, two organizations and seven laboratories were contacted to determine their willingness either to complete a questionnaire or to disseminate the questionnaire to other relevant laboratories. Six of seven laboratories invited to participate returned a completed questionnaire. Preanalytical requirements, acceptable specimen type, storage times and conditions, were in keeping with the results of this review. Stated analytical precision of enzymatic analysis by the surveyed laboratories, as expected, was similar to that reported within published articles and ranged from 3 to 10 percent within runs, and from 5 to 20 percent between runs. Among the surveyed laboratories, reported concordance between enzymatic analysis genotyping ranged from 60 percent to 100 percent. Although the range is in keeping with the published reports, no data was provided to directly support these conclusions.

Knowledge of TPMT Status to Guide Therapy

A single fair quality randomized trial in 333 patients demonstrated that over a four month observation period, pretreatment genotyping did not significantly alter prescribing practice compared with no pretesting. There was no significant difference between the two groups in terms of starting doses and mean AZA prescribed dose at the end of the study period. Despite prior knowledge of noncarrier TPMT status in the tested group of patients, most patients were administered starting doses lower than 2mg/kg/day, similar to the nontested group. This was because physicians were free to practice as per routine, and which they did just as cautiously as the nontested group despite prior knowledge obtained from genotyping. Knowledge of heterozygous status, however, did result in prescription of lower starting doses compared with noncarriers in the group tested before therapy. There is limited applicability of this evidence because there was just one homozygous carrier in the whole sample of mostly IBD patients. The finding conforms with an earlier national survey in the United Kingdom in which the uptake of prior TPMT testing differed substantially by clinical specialty – there was a higher uptake of TPMT enzyme-level testing by dermatologists, compared with gastroenterologists and rheumatologists, and this might explain why the group that underwent prior genotyping still ended up receiving doses of azathioprine similar to the nongenotyped control group.234 Since most patients had inflammatory bowel disease, it appears that gastroenterologists, specifically, tend to exercise a cautious prescribing approach over one primarily guided by prior knowledge of the TPMT status.

When compared with no pretesting for TPMT status, testing did not demonstrate a significant difference in the odds of mortality and serious adverse events in 333 randomized patients. The evidence was rated as insufficient given a medium risk of bias and strong possibility of type II error. The applicability of the evidence was deemed limited as there was just one homozygous carrier of TPMT variant allele in the entire sample of mostly IBD patients, the followup period was just 4 months and eligibility criteria of the trial excluded patients who would most likely have experienced adverse events. Evidence was also rated insufficient for the outcomes of health-related quality of life and myelotoxicity as no data were available.

Evidence from one RCT with low event rate showed no significant advantage of prior genotyping with respect to the intermediate outcomes of neutropenia and pancreatitis. The applicability of this evidence is quite limited because there was just one homozygous carrier in the whole sample of 333 patients, the followup period was just 4 months and eligibility criteria of the trial excluded patients who would most likely have experienced adverse events. Also, type II error cannot be ruled out. As TPMT status determination may not identify all individuals at increased risk of drug toxicity,235 direct and conclusive evidence of the utility of pretesting in terms of drug related harms reduction is wanting for evidence-based guidelines on thiopurine therapy. For the outcome of liver toxicity, significantly higher odds were observed in the group that underwent prior TPMT genotyping, odds ratio 2.54 (1.08, 5.97)]. There was no significant difference in starting or mean doses received between the tested and nontested group. Although this finding merits further investigation, but it appears to be a type I error and unrelated to the intervention of pretreatment genotyping.

Various recent guidelines, as well as the product monograph for azathioprine, have advocated determination of TPMT status prior to treatment with thiopurine drugs.30,243 The proposition that knowledge of TPMT status prior to therapy would lead to decreased rates of dose-dependent toxicity is rational and based on evidence of strong genotypic and phenotypic associations in observational studies of limited validity. Compared with non-carriers and heterozygous carriers, homozygous are considered to be most at risk of developing neutropenia. However, from an evidence-based perspective, guideline recommendations of pretreatment TPMT testing are premature for several reasons. First and foremost, the direct evidence base for these recommendations is lacking – especially the crucial evidence that TPMT pretesting before thiopurine therapy decreases myelotoxicity specific mortality. Also, given just one homozygous carrier in the only available direct evidence investigating usefulness of pretreatment, evidence is equally lacking for this particular subgroup of patients. Second, patients on thiopurine drugs are required to undergo complete blood count monitoring on a regular basis in an attempt to prevent severe myelotoxicity by early detection. Third, azathioprine and 6-MP had been used successfully for a number of years prior to the availability of TPMT testing and management (i.e. testing or not before therapy) varies across clinical specialties. Fourth, thiopurine related toxicities are also partially explained by mutations in other enzymes, drug interactions, intercurrent infections, and immune mediated drug reactions. Fifth, direct evidence of effectiveness of pretesting in the specific subpopulation of patients homozygous or compound heterozygous for the TPMT variant alleles is lacking the most, albeit not surprisingly, because of the low prevalence of homozygosity. Furthermore, the use of TPMT status to guide treatment has the potential to reduce the efficacy of thiopurine drugs if physicians are overzealous in reduction of thiopurine dosage. Indeed, the 2004 guidelines from the British Society of Gastroenterology recognized this and stated, “It cannot yet be recommended as a prerequisite to therapy, because decades of experience has shown clinical [azothioprine] to be safe in [ulcerative colitis] or [Crohn’s disease]”.244 As far as the utility of pretesting for TPMT status before thiopurine treatment is concerned, our review is indeterminate because of insufficient evidence and calls for urgent further research. This is at odds with previously published economic evaluations recommending testing. However, those evaluations have been criticized for incorporating clinical data from retrospective studies and expert opinion instead of prospective empiric evidence – the latter, as our review shows, are lacking236

Association of TPMT Status With Thiopurine Toxicity

In the presence of insufficient direct evidence of prior knowledge of TPMT status to guide thiopurine therapy, possible associations between TPMT status and the clinical outcomes of mortality, infections, hospitalization, withdrawals due to adverse events, serious adverse events and health-related quality of life, as well as the surrogate outcomes of myelotoxicity, liver toxicity, and pancreatitis were examined.

Toxicity of thiopurine drugs is thought to be mediated primarily through their pharmacologically active metabolites, 6-tGNs, which can be considered a dose-dependent toxicity. Incorporation of 6-tGN into DNA triggers cell-cycle arrest and apoptosis through the mismatch repair pathway.8 Recent evidence has also shown that thiopurine drugs can induce apoptosis in T-cells through modulation of Rac1 activation upon CD28 costimulation. Therefore accumulation of 6-tGNs can clearly induce various degrees of myelosuppression. Dose-independent toxicity, until recently was not understood, and appears to be either immune mediated or due to metabolites previously thought to be inactive. Both hepatotoxicity and pancreatitis are thought to be caused through dose-independent toxicity. Immune mediated reactions with AZA include hepatitis, pancreatitis, rash etc. and usually occur with 4 weeks of initiation of therapy. In some patients, this reaction can be overcome by switching to 6-MP, implying a role for the imidazole moiety in toxicity.237 The mechanisms of hepatotoxicity have been studies in most depth. A link between hepatotoxicity and increased levels of 6-methlyMP ribonucleotide (6-MMPR) has been suggested. Seidman et al. identified a link between TPMT activity and levels of 6-MMPR, suggesting that TPMT may play a role in hepatotoxicity.238 However, in this case higher TPMT activity may be more relevant to induction of toxicity than lower activity (see Figure 1). In a subset of patients with subtherapeutic levels of 6-tGNs, dose escalation of 6-MP did not increase 6-tGN levels, however levels of 6-MMPR did increase. In one study, 24 percent of patients with elevated 6-MMPR levels showed higher rates of hepatotoxicity.119 Mardini et al also found that elevated levels of 6-MMPR correlated with hepatotoxicity.239 While others have found no relation between levels of 6-MMP and hepatotoxicity.240,241

Thirty-four studies were identified that provided relevant data on allelic variants and 16 studies were identified that provided data on TPMT enzymatic activity in relation to clinical outcomes. There is insufficient evidence examining association of TPMT status, as determined by either allelic determination or enzymatic activity, with the outcomes of mortality, hospitalization rates, serious adverse events, health related quality of life and neutropenia precluding meaningful conclusions. Furthermore, insufficient data were available for infection and thrombocytopenia by enzymatic analysis; however limited data showing no effect were available for allelic determination. The majority of the studies were of cross-sectional design and fair quality.

The available evidence confirms our previous understanding that there is strong association between the outcome of leukopenia and presence of variant TPMT alleles or subnormal enzymatic activity, and dose response relationships with both allelic variants and TPMT enzymatic activity.227 The strongest association was for homozygous carriers or low to absent enzymatic activity, compared with noncarrier or normal enzymatic activity patient groups. There is also some indication that lower levels of TPMT enzymatic activity may be associated with the composite outcome of myelotoxicity, defined as decreased levels of at least two hematopoietic cell lines. For most other outcomes, there was no significant association with either a presence of a TPMT allelic variant or subnormal enzymatic activity. Given the small number of studies involving few patients with events, type II error cannot be ruled out for most outcomes other than hepatotoxicity and pancreatitis. For these two outcomes, our findings of no association of these outcomes with either low enzymatic activity or presence of TPMT carrier states are consistent with extant literature.72,82,163

Costs of Determining TPMT Status Versus Costs of Treating Drug-Associated Complications

Global interest in costs of TPMT phenotyping and genotyping is reflected in studies from around the world, published between 1995 and 2010.

Most cost estimates were based on theoretical populations, although one study was based on a bullous pemphigoid patient with AZA toxicity.210 Across all studies, there was some consensus on the cost of genotype and phenotype testing, although the cost perspective was often not reported. The one study reporting costs from a societal perspective showed higher costs than the others.208 Heterogeneous estimates of the total cost likely arose from differing methodological choices. For example, the cost of treating AZA-associated complications was estimated to be between $1,325 and $5,877 in USD. This four-fold difference has the potential to result in disparate total cost estimates. The average cost of TPMT phenotyping was approximately half of the average cost of TPMT genotyping, but these costs may not be generalizable to all TPMT tests. These costs will have to be taken into consideration, along with the relative sensitivity and specificity of TPMT genotyping and TPMT phenotyping, when deciding upon which test to use.

Strengths and Limitations

This is the first comprehensive systematic review answering the question whether testing TPMT status prior to thiopurine therapy changes management and thiopurine toxicity outcomes. It is also the first review of the analytical performance characteristics of enzymatic measurement of TPMT activity and determination of TPMT allelic polymorphisms. These key questions were developed from a conceptual framework of the topic with input from the Technical Expert Panel. This panel included clinical, genetics, biochemistry and systematic review methodology experts. We contacted authors and obtained additional data that we incorporated in meta-analyses. A survey of laboratories, although not part of the original work plan, further added to our understanding of laboratory practices related to TPMT testing.

Despite our rigorous methodological approach, this review has several limitations. From a clinical perspective, the most important equipoise about the utility of prior TPMT testing remains insufficiently answered due to a dearth of comparative effectiveness literature on TPMT testing and its limited applicability. Evidence relating to other key questions originated in observational studies of poor to fair quality, so is of limited strength. The genetic associations established in this review, while confirming previous literature, are still of limited reliability.242 Lastly, we pooled diverse studies, with the assumption that most thiopurine toxicity is determined genetically and biochemically. There was insufficient primary evidence to identify important effect modifiers or to carry out separate subgroup meta-analyses on studies with lower risk of bias.

Recommendations and Future Research

There is insufficient evidence examining the effectiveness of TPMT pretreatment enzymatic or genetic testing, to minimize thiopurine related toxicity in patients with chronic autoimmune diseases. As a priority, well powered, good quality, randomized controlled studies need to be conducted, in diverse and representative patient populations, to compare the effectiveness of TPMT genotyping and phenotyping with one another, and with no TPMT testing. These studies should be large enough to include a sizable number of patients homozygous for the variant alleles and should be pragmatic in conduct, mimicking routine clinical practice. Outcomes would include both treatment efficacy and harms associated with thiopurine therapy. Another objective would be to establish the optimum initial dose adjustment for a given TPMT status. These studies should ensure that outcomes are truly assessed without prior knowledge of results of TPMT testing and administered drug dose, by employing appropriate blinding procedures. The recently concluded pragmatic TARGET study by Newman and associates was under-powered to detect differences in clinically important outcomes, largely because it faced recruitment problems. In future such recruitment problems may be mitigated by educating the public and clinicians that the evidence base for pretreatment TPMT testing is lacking and that it is unclear whether pretreatment testing does more good (i.e. reduction in thiopurine related toxicity) than harm (i.e. reduction in thiopurine efficacy because of overzealous dose reductions based on prior testing).

Until such experimental high quality evidence becomes available, alternative evidence may be sought in prospectively designed observational studies that estimate health related quality of life, drug prescription patterns, and myelotoxicity related mortality as important outcomes associated with and with no pretreatment TPMT testing. With availability of empiric evidence from such studies, decision-analytic modeling that comprehensively consider alternative strategies such as regular blood cell count and liver enzyme testing, metabolite monitoring, and dose adjustments for concomitant medications that impact the TPMT enzymatic pathway can help guide practice until evidence becomes available from well powered pragmatic trials. Subsequent models might also need to consider new information as technologies develop and knowledge evolves.

TPMT genotyping should test for the most common TPMT polymorphisms in the population of interest. There is little direct evidence identifying the optimum set of alleles to be tested, and this may need to be established for specific populations if TPMT genotyping turns out to be effective in future studies.

TPMT activity analyses are reported on one of two bases: per milliliter of packed red blood cells; or per gram of hemoglobin. These are not readily or exactly comparable. Common reporting units are needed, as well as cutoffs for low/absent, intermediate, normal TPMT enzymatic activity, and high enzymatic activities.

Future studies should clearly report numbers of uninterpretable or equivocal test results.


(see Revised Conclusions)

This is the first comprehensive systematic review answering the question whether testing TPMT status prior to thiopurine therapy changes management and thiopurine toxicity outcomes, including leukopenia and myelotoxicity. There is currently insufficient evidence regarding effectiveness of determining TPMT status prior to thiopurine treatment in terms of improvement in clinical outcomes and incident myelotoxicity in comparison with routine monitoring of full blood counts and adverse events. It is also unclear whether pretesting guides appropriate prescribing. Indirect evidence confirmed previously known strong associations between lower levels of TPMT enzymatic activity and the presence of TPMT variant alleles with thiopurine related leukopenia.

Sufficient preanalytical data are available to recommend preferred specimen collection, stability and storage conditions for determination of TPMT status. There was no clinically significant effect of age, gender, various coadministered drugs, or most comorbid conditions (with the exception of renal failure and dialysis). The currently available methods for determination of TPMT enzymatic activity show good precision, with coefficients of variation generally below 10 percent. Based upon limited evidence, the reproducibility of TPMT allelic polymorphism determination is acceptable. However, the sensitivity of genetic testing to identify patients with low and/or intermediate TPMT enzymatic activity cannot be precisely estimated. Thus, if knowledge of TPMT status is desired, and if recent RBC transfusion is excluded, the available evidence suggests that enzymatic assay should be preferred over the determination of allelic polymorphism.

Despite widespread interest, precise costs associated with TPMT phenotyping are unknown. More research has been conducted examining TPMT genotyping but the cost estimates are heterogeneous, likely due to different methodological choices.


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...