Haematological adverse events associated with tyrosine kinase inhibitors in chronic myeloid leukaemia: A network meta-analysis
Abstract
Aims
Despite their overall favourable safety profile, tyrosine kinase inhibitors (TKIs) are related to severe adverse events including haematological toxicities such as anaemia, leucopenia, neutropenia and thrombocytopenia. We designed a systematic review and network meta-analysis of randomised controlled trials to compare safety among TKIs (bosutinib, dasatinib, imatinib, nilotinib, ponatinib and radotinib) used by patients diagnosed with chronic myeloid leukaemia.
Methods
We obtained data from the PubMed, Scopus, Web of Science, and SciELO databases. The Bayesian approach was used for direct and indirect comparisons, and the treatments were ranked by the surface under the cumulative ranking curve (SUCRA).
Results
Seventeen studies were included in the network meta-analysis. Our data show that dasatinib was generally considered worse than the other TKIs, with SUCRA values for 140 mg dasatinib of 90.3% for anaemia, 87.4% for leucopenia, 90.6% for neutropenia and 97.2% for thrombocytopenia. In addition, nilotinib was shown to be safer, with SUCRA values for 600 and 800 mg doses of 21.9 and 35.8% for anaemia, 23.8 and 14.6% for leucopenia, 33.0 and 17.7% for neutropenia, and 28.7 and 32.6% for thrombocytopenia, respectively.
Conclusion
Dasatinib appeared as the least safe drug for chronic myeloid leukaemia, probably because it binds to multiple key kinase targets, being more prone to cause serious haematological adverse events. Nilotinib demonstrated a safer profile, mostly due to its selective binding capacity.
Graphical Abstract
What is already known about this subject
- Tyrosine kinase inhibitors are indicated for the management of chronic myeloid leukemia (CML), and haematological grades 3–4 adverse events have been reported during treatment.
- Haematological toxicities may require temporary treatment interruption or dose modification, or sometimes even lead to treatment discontinuation.
- To the best of our knowledge there are no studies comparing the tyrosine kinase inhibitors with respect to haematological adverse events.
What this study adds
- The present systematic review found that dasatinib presents the least safe profile considering haematological events for patients with CML.
- Nilotinib is a promising alternative for CML treatment.
1 INTRODUCTION
Tyrosine kinase inhibitors (TKIs) represent a milestone in chronic myeloid leukaemia (CML) treatment. Although these drugs are effective and have an overall favourable safety profile with mild to moderate adverse events (AEs), they can also cause severe events including haematological AEs, such as anaemia, leucopenia, neutropenia and thrombocytopenia, which occur due to myelosuppression. Moreover, the use of TKIs may lead to dermatological and gastrointestinal AEs, hepatic and pancreatic disorders, musculoskeletal symptoms, fluid retention, and pulmonary and cardiovascular toxicity.1-3
Imatinib was the first TKI licensed for CML treatment, becoming the gold standard for newly diagnosed patients in all phases of the disease. Five new-generation TKIs were later approved—dasatinib, nilotinib, bosutinib, ponatinib and radotinib—all exerting their antileukaemic effect by means of binding to different tyrosine kinases. These differences in the mechanisms of action justify differences in their efficacy and safety.3
Since CML management generally requires TKI use for the rest of the patient's life—except for those who achieve deep molecular response and who are susceptible to treatment-free remission, it is essential to evaluate drug's tolerability, in addition to their effectiveness and safety. The development of intolerance to a particular TKI results in treatment interruption and switching to another TKI; thus, the clinical decision regarding treatment selection must take into account the safety profile of each drug, along with the comorbidities that the patient presents.1, 3, 4 The purpose of the present study was to systematically review the haematological safety profile of all the TKIs used in CML treatment.
2 METHODS
The present systematic review is part of a project evaluating the efficacy and safety of TKIs in patients with CML (PROSPERO registration number: CRD42017065864). An article on the overall efficacy and safety of these drugs at 12 months was previously published.5 The systematic review and meta-analyses were performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses—network meta-analysis extension and Cochrane Collaboration recommendations.6-8 Two researchers performed all steps, independently. In the case of disagreement, a third researcher was consulted.
2.1 Search strategy and eligibility criteria
We performed systematic searches in PubMed, Scopus, Web of Science and SciELO electronic databases, with no data or language restrictions (last updated in February 2018). Search results were not limited by any filtering tool, nor by year or country of publication. We also manually searched for records in the reference lists of review articles and included studies, on trial's registration databases (clinicaltrials.gov), and on Google Scholar. The complete search strategy is depicted in Table S1. Following article recovery from the databases, 2 authors independently read each title and abstract. Relevant articles were subsequently read in full, and those fulfilling the inclusion criteria were included for data extraction.
The inclusion criteria were as follows: (i) studies involving adult patients (age >18 years) diagnosed with CML without any other oncological comorbidities; (ii) trials evaluating any TKI (i.e. imatinib, dasatinib, nilotinib, bosutinib, radotinib or ponatinib, at any dose or regimen) compared head-to-head with another TKI or to the same TKI at a different daily dose or to placebo; (iii) trials assessing haematological safety outcomes (incidence of serious AEs, SAEs); and (iv) studies designed as a randomised controlled trial (RCT).
Other types of studies were excluded (e.g. observational studies, reviews, pharmacokinetic trials, nonrandomised trials, dose titration or dose-escalating trials), in addition to those evaluating other drugs, clinical conditions or outcomes, or those written in a non-Roman alphabet.
2.2 Data extraction and quality assessment
The data extracted from the included articles were: (i) general study information (authors' names, year of publication, country, sample size, study design); (ii) patient baseline characteristics (age, sex, previous treatment and clinical condition); (iii) interventions (dosage, regimen, and treatment duration); (iv) haematological safety outcomes (incidence of SAEs grades 3–4). Grade 3 AEs are considered severe, usually involving multiple and disruptive symptoms, and may require hospitalisation and surgery. Grade 4 AEs can be life-threatening and result in disabling or catastrophic consequences.9
The methodological quality of the articles included in the present systematic review was evaluated using the Jadad Scale and the Cochrane Collaboration's tool for assessing the risk of bias.10, 11
2.3 Statistical analysis
Network meta-analysis combines direct and indirect evidence, making possible the comparison of treatments that would not be viable using the traditional pairwise meta-analysis.12, 13 We conducted Bayesian random effects network meta-analysis assuming common heterogeneity parameters, using the Markov chain Monte Carlo simulation method. We used noninformative priors for the analysis. A consistency model was built for each outcome, and the relative effect size for each treatment is reported as the odds ratio (OR) with its 95% credible interval (CrI). The goodness-of-fit of the model was assessed using residual deviance. No adjustment for multiple comparison was performed. In addition, we constructed rank probabilities with the aim of estimating which drug was probably the best and the worst for each outcome. To better represent the rank results, we performed surface under the cumulative ranking curve analysis (SUCRA).13-16
The geometry of the networks followed the complexity level of the primary studies. To ensure that there was no divergence between direct and indirect comparisons, and to estimate the robustness of the network, we performed inconsistency and node-splitting analyses (P-values <.05 indicate inconsistencies). To better explore the results and test their robustness, we anticipated that sensitivity analysis could be performed.17, 18 Subgroup analyses considering drugs used as primary or secondary therapy were performed. Additional analyses grouping the different dose levels to generate comparisons between drugs were conducted, as well as analyses separating the studies by treatment time. Analyses were performed using the ADDIS version 1.17.6 software (Aggregate Data Drug Information System; http://drugis.org/addis)19 and Microsoft Excel.
2.4 Nomenclature of targets and ligands
Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY,20 and are permanently archived in the Concise Guide to PHARMACOLOGY 2017/18.21
3 RESULTS
After searching in the electronic databases and removing duplicates, 4114 records were screened by title and abstract. Of these, 62 were read in full, and 27 were included in the systematic review. Another 3 records were manually added (Table S2),22-24 resulting in a total of 30 articles.22-51 These articles referred to 16 different RCTs (Figure 1), 13 of which were performed in patients with no previous CML treatment (naïve patients) and another 4 trials included patients who had received prior treatment. One trial was split into 2 publications.32, 47 To avoid data duplication, only 1 article from each trial was included in the meta-analysis (i.e. the most recent trial referring to a longer follow-up period), the only exception was the case mentioned in the previous sentence. The main characteristics of the 17 included trials are presented in Table 1.23-25, 28, 29, 31-33, 35, 38, 40, 44-47, 50, 51
| Study | Trial design | Trial name | Duration (y) | Previous treatment | Drug | Daily dose (mg) | Number of patients | Age (y) | % men |
|---|---|---|---|---|---|---|---|---|---|
| Baccarini, 2014 | OL RCT | TOPS | 3.5 | No prior CML treatment | Imatinib | 400 | 157 | 45 (18–75) | 53 |
| Imatinib | 800b | 319 | 48 (18–75) | 57 | |||||
| Cortes, 2016 | OL RCT | DASISION | 5 | No prior CML treatment | Dasatinib | 100 | 259 | 46 (18–84) | 56 |
| Imatinib | 400 | 260 | 49 (18–78) | 63 | |||||
| Cortes, 2016 (2) | OL RCT | LASOR | 2 | After at least 3–18 months on imatinib 400 mg once daily | Nilotinib | 800b | 96 | 46 (32–56) | 54 |
| Imatinib | 600 | 95 | 44 (33–56) | 58 | |||||
| Cortes, 2018 | OL RCT | BFORE | 1 | No prior CML treatment | Bosutinib | 400 | 268 | 52 (18–84) | 58 |
| Imatinib | 400 | 268 | 53 (19–84) | 56 | |||||
| Deininger, 2014a | OL RCT | - | 1 | No prior CML treatment | Imatinib | 400 | 72 | 50 (23–80) | 45 |
| Imatinib | 800 | 73 | 52 (19–82) | 47 | |||||
| Gambacorti, 2014 | OL RCT | BELA | 2.5 | No prior CML treatment | Bosutinib | 500 | 250 | 48 (19–91) | 60 |
| Imatinib | 400 | 252 | 47 (19–89) | 54 | |||||
| Hjorth-Hansen, 2014 | OL RCT | NordCML006 | 3 | No prior CML treatment | Dasatinib | 100 | 22 | 53 (29–71) | 32 |
| Imatinib | 400 | 24 | 58 (38–78) | 63 | |||||
| Hochhaus, 2016 | OL RCT | ENESTnd | 5 | No prior CML treatment | Nilotinib | 600b | 282 | 47 (18–85) | 56 |
| Imatinib | 800b | 281 | 47 (18–81) | 62 | |||||
| Imatinib | 400 | 283 | 46 (18–80) | 56 | |||||
| Hughes, 2017 | OL RCT | ENESTcmr | 4 | After at least 2 years on imatinib | Nilotinib | 800b | 104 | 46 (23–82) | 68 |
| Imatinib | 400 or 600 | 103 | 52 (19–76) | 63 | |||||
| Kalmanti, 2015 | OL RCT | CML - Study IV | 10 | No prior CL treatment | Imatinib | 400 | 406 | 53 | NR |
| Imatinib | 800 | 422 | |||||||
| Imatinib +interferon-α | 400 | 434 | |||||||
| Imatinib after interferon-α failure | 400 | 131 | |||||||
| Imatinib + cytarabine | 400 | 158 | |||||||
| Kantarjian, 2009 | OL RCT | START-R | 2 | After imatinib resistance | Dasatinib | 140b | 101 | 51 (24–85) | 52 |
| Imatinib | 800b | 49 | 51 (24–80) | 45 | |||||
| Kwak, 2017 | OL RCT | RERISE | 1 | No prior CML treatment | Radotinib | 600 | 79 | 45 (20–75) | 66 |
| Radotinib | 800b | 81 | 43 (18–84) | 58 | |||||
| Imatinib | 400b | 81 | 45 (18–83) | 64 | |||||
| Lipton, 2016 | OL RCT | EPIC | NR | No prior CML treatment | Ponatinib | 45 | 155 | 55 (18–89) | 63 |
| Imatinib | 400 | 152 | 52 (18–86) | 61 | |||||
| Petzer, 2010 | OL RCT | ISTAHIT | 0.5 | No prior CML treatment | Imatinib | 400 | 113 | 46 (20–68) | 42 |
| Imatinib | 800b | 114 | 46 (18–76) | 47 | |||||
| Radich, 2012a | OL RCT | - | 5 | No prior CML treatment | Imatinib | 400 | 123 | 50 (19–89) | 58 |
| Dasatinib | 100 | 123 | 47 (18–90) | 60 | |||||
| Shah, 2010 | OL RCT | - | 2 | After imatinib resistance or intolerance | Dasatinib | 100 | 167 | 56 (20–78) | 50 |
| Dasatinib | 100b | 168 | 55 (21–84) | 51 | |||||
| Dasatinib | 140 | 167 | 54 (20–84) | 42 | |||||
| Dasatinib | 140b | 168 | 55 (18–83) | 46 | |||||
| Wang, 2015 | OL RCT | ENESTchina | 2 | No prior CML treatment | Nilotinib | 600b | 134 | 41 (18–76) | 68 |
| Imatinib | 400 | 133 | 39 (19–74) | 61 |
- CML, chronic myeloid leukemia; INF, interferon; NR, not reported; OL, open-label; RCT, randomised controlled trial.
- Age is given as the median (range).
- All drugs were administered orally.
- The articles assigned.
- a to the evaluation of different comparisons with different populations that were registered under the same NCT number (ClinicalTrials.gov identifier).
- Regimens assigned.
- b were equally divided into 2 doses per day (twice daily frequency). The remaining regimens refer to once daily frequency.
All studies were head-to-head and encompassed the drugs bosutinib (n = 2), dasatinib (n = 5), imatinib (n = 16), nilotinib (n = 4), ponatinib (n = 1) and radotinib (n = 1) at different doses. No placebo-controlled trials were found. All studies included only patients in the chronic phase of the disease. In 1 trial, imatinib was combined with interferon-α and with cytarabine.38 Only in 1 of the 17 RCTs was imatinib not the main comparator.50 Thirteen studies (76.5%) included >100 patients in each arm. The quality of the studies was considered low to moderate when assessed by the Jadad scale. Fourteen studies (82.4%) received a score of 2 of a maximum 5, and 3 studies scored 3. All studies were described as randomised, but only 3 (16.7%) reported the method of randomisation. None of the studies were double-blinded. The individual scores of each trial can be found in Table S3. The results of the risk of bias tool show that all articles possessed a low risk when considering detection, attrition, and bias reporting. Despite the fact that all RCTs were open-label, the blinding did not interfere with the assessment of outcomes, since all the evaluated AEs were objectives (monitored by biochemical examinations). Overall, a high risk of bias was obtained for the domain other bias, since the majority of trials were sponsored by a pharmaceutical company (Figures S1, S2).
The most-reported grades 3–4 haematological adverse events among the trials were anaemia (n = 16 studies), leucopenia (n = 11), neutropenia (n = 14) and thrombocytopenia (n = 17). For each 1 of these AEs, a network meta-analysis was performed considering all patients. The network plots are depicted in Figure 2.
Subgroup meta-analyses with patients in primary CML therapy (naïve) were also obtained (Figures S3, S4; Table S4). Given the small number of studies reporting data on secondary CML therapy (n = 4), statistical analyses for this subgroup were not possible. The SUCRA results for the analyses of different dose levels of each drug are demonstrated in Table S5. Furthermore, subgroup meta-analyses separating outcomes for time were feasible for anaemia, neutropenia and thrombocytopenia outcomes only for studies with 12-month outcomes (Tables S6–S8), due to the small number of studies with outcomes at 6, 24, 30, 36, 42, 48, 60 and 120 months.
The analysis revealed that 140 mg dasatinib caused significantly more cases of anaemia than 400 mg imatinib (OR 0.30 with 95% [CrI 0.07–0.86]), 600 mg imatinib (OR 0.04 with 95% CrI [0.00–0.61]), or 600 mg nilotinib (OR 0.18 with 95% CrI [0.03–0.81]; see Table 2A),being considered the less safe therapeutic alternative for this outcome (90% probability of being the least safe in SUCRA (Table 3). Dasatinib at a lower dose (100 mg) also resulted in more cases of anaemia than 600 mg imatinib (OR 0.05 with 95% CrI [0.00–0.87]). Imatinib (600 mg) was considered the safest option for this outcome (5% in SUCRA), followed by 600 mg nilotinib (22% in SUCRA). Considering only the RCTs that assessed primary CML therapy, 800 mg imatinib was related to more cases of anaemia as compared with the same drug at a lower dose (400 mg; OR 0.43 with 95% CrI [0.17–0.88]) and with 600 mg nilotinib (OR 0.25 with 95% CrI [0.06–0.93]). In this scenario, both doses of nilotinib (600 and 800 mg) presented the lowest probabilities of causing anaemia, with 18 and 22% in SUCRA, respectively (see Supporting Material for complete analysis, Figure 5A).
| (A) Anaemia (grade 3–4) | |||||||||||
| Bosutinib 400 mg | 1.23 (0.17, 8.96) | 1.72 (0.32, 9.90) | 2.63 (0.46, 20.16) | 0.78 (0.18, 3.15) | 0.09 (0.00, 2.03) | 1.60 (0.34, 8.53) | 0.47 (0.08, 2.85) | 0.63 (0.11, 4.55) | - | 1.32 (0.14, 11.30) | 0.59 (0.06, 5.92) |
| 0.53 (0.06, 4.89) | Bosutinib 500 mg | 1.38 (0.27, 8.14) | 2.08 (0.39, 16.71) | 0.63 (0.15, 2.62) | 0.08 (0.00, 1.66) | 1.28 (0.27, 6.93) | 0.38 (0.07, 2.32) | 0.51 (0.09, 3.75) | - | 1.09 (0.12, 9.14) | 0.48 (0.05, 4.59) |
| 3.97 (0.29, 69.13) | 7.43 (0.46, 128.45) | Dasatinib 100 mg | 1.50 (0.59, 5.10) | 0.45 (0.17, 1.11) | 0.05 (0.00, 0.87) | 0.94 (0.31, 2.85) | 0.27 (0.06, 1.11) | 0.38 (0.09, 1.78) | - | 0.77 (0.10, 4.97) | 0.35 (0.04, 2.51) |
| 4.19 (0.42, 47.89) | 7.63 (0.68, 84.17) | 1.03 (0.25, 4.02) | Dasatinib 140 mg | 0.30 (0.07, 0.86) | 0.04 (0.00, 0.61) | 0.62 (0.17.1.83) | 0.18 (0.03, 0.81) | 0.25 (0.04, 1.25) | - | 0.51 (0.05, 3.44) | 0.22 (0.02, 1.72) |
| 1.37 (0.31, 6.02) | 2.57 (0.49, 12.61) | 0.34 (0.03, 2.97) | 0.32 (0.05, 1.92) | Imatinib 400 mg | 0.12 (0.00, 1.84) | 2.04 (1.04, 4.72) | 0.60 (0.21, 1.83) | 0.82 (0.27, 2.97) | - | 1.69 (0.31, 9.06) | 0.76 (0.12, 4.70) |
| 0.44 (0.02, 5.14) | 0.87 (0.04, 9.64) | 0.12 (0.00, 1.75) | 0.12 (0.00, 1.33) | 0.32 (0.03, 2.28) | Imatinib 600 mg | 17.32 (1.08, 602.24) | 4.79 (0.32, 161.94) | 6.49 (0.63, 195.64) | - | 14.39 (0.60, 679.51) | 6.23 (0.23, 294.46) |
| 2.74 (0.48, 15.68) | 5.17 (0.73, 31.94) | 0.66 (0.07, 5.35) | 0.63 (0.12, 3.01) | 2.00 (0.80, 4.72) | 6.18 (0.73, 84.53) | Imatinib 800 mg | 0.29 (0.08, 1.01) | 0.39 (0.10, 1.61) | - | 0.83 (0.12, 4.85) | 0.36 (0.05, 2.40) |
| 0.50 (0.07, 2.64) | 0.89 (0.12, 5.85) | 0.12 (0.01, 1.25) | 0.11 (0.01, 0.81) | 0.36 (0.12, 0.95) | 1.05 (0.14, 12.81) | 0.18 (0.04, 0.67) | Nilotinib 600 mg | 1.35 (0.40, 5.16) | - | 2.83 (0.36, 20.05) | 1.25 (0.15, 9.72) |
| 0.38 (0.03, 2.16) | 0.70 (0.06, 4.32) | 0.09 (0.00, 0.90) | 0.08 (0.01, 0.61) | 0.27 (0.05, 0.80) | 0.81 (0.13, 4.63) | 0.14 (0.02, 0.51) | 0.73 (0.14, 2.53) | Nilotinib 800 mg | - | 2.07 (0.25, 14.86) | 0.91 (0.10, 7.53) |
| - | - | - | - | - | - | - | - | - | Ponatinib 45 mg | - | - |
| - | - | - | - | - | - | - | - | - | - | Radotinib 600 mg | 0.45 (0.07, 2.45) |
| - | - | - | - | - | - | - | - | - | - | - | Radotinib 800 mg |
| (B) Leukopenia (grade 3–4) | |||||||||||
- All treatments are listed in alphabetical order. Comparisons between treatments should be read from left to right, and the estimated ratio is in the cell in common between the column-defining treatment and the row-defining treatment. Values are presented as odds ratio (OR) with 95% credible interval (CrI). For anaemia (grade 3–4; upper triangle, values in green) an OR <1 favours the event for the row-defining treatment while an OR >1 favours the column-defining treatment. For leucopenia (grade 3–4; lower triangle, values in white) an OR <1 favours the event for the column-defining treatment while an OR >1 favours the row-defining treatment. Significant results between the interventions are highlighted in italic and bold.
| Drug | Serious adverse event (grades 3–4) | |||
|---|---|---|---|---|
| Anaemia | Leukopenia | Neutropenia | Thrombocytopenia | |
| Bosutinib 400 mg | 52% | 47% | 22% | 71% |
| Bosutinib 500 mg | 60% | 27% | 21% | 34% |
| Dasatinib 100 mg | 74% | 84% | 74% | 79% |
| Dasatinib 140 mg | 90% | 87% | 91% | 97% |
| Imatinib 400 mg | 39% | 59% | 56% | 34% |
| Imatinib 600 mg | 5% | 24% | - | 20% |
| Imatinib 800 mg | 74% | 79% | 70% | 63% |
| Nilotinib 600 mg | 22% | 24% | 33% | 29% |
| Nilotinib 800 mg | 36% | 15% | 18% | 33% |
| Ponatinib 45 mg | - | - | 21% | 65% |
| Radotinib 600 mg | 63% | - | 65% | 28% |
| Radotinib 800 mg | 35% | - | 80% | 47% |
- Surface under the cumulative ranking curve values can range from 0 to 100%. Higher rates represent a less safe option (negative outcome).
For the outcome of leucopenia, both doses of dasatinib (100 and 140 mg) and imatinib (400 and 800 mg) were considered less safe options (SUCRA values ranging from 84 to 87%; Table 3; and Figure S5B), while 600 or 800 mg nilotinib were superior alternatives (SUCRA values 15–24%). Similarly, in the subgroup analyses of primary CML therapy, nilotinib was also considered the safest drug (SUCRA values 19–25%), being statistically superior to imatinib (OR 2.60 with 95% CrI [1.08–6.77] for nilotinib 600 mg vs imatinib 400 mg; OR 5.22 with 95% CrI [1.58–17.77] for nilotinib 600 mg vs imatinib 800 mg; OR 0.15 with 95% CrI [0.04–0.70] for nilotinib 800 mg vs imatinib 800 mg; see Table 2B).
When considering the outcome of neutropenia, 140 mg dasatinib was ranked last (91% in SUCRA; Table 3), being statistically different from nilotinib (600 and 800 mg), ponatinib (45 mg), and bosutinib (400 and 500 mg). Ponatinib (45 mg) and bosutinib (400 and 500 mg) showed the lowest probabilities of causing neutropenia (< 22%), both in the initial and subgroup patient analyses (Figure S5C). In the subgroup analyses, bosutinib (400 and 500 mg), nilotinib (800 mg) and ponatinib (45 mg) were statistically safer than imatinib (800 mg; OR 0.25 with 95% CrI [0.08–0.79], OR 0.25 with 95% CrI [0.08–0.87], OR 0.28 with 95% CrI [0.10–0.90] and OR 0.21 with 95% CrI [0.04–0.91]), respectively (Table 4A).
| (A) Neutropenia (grade 3–4) | |||||||||||
| Bosutinib 400 mg | 0.98 (0.23, 4.54) | 3.44 (0.84., 10.81) | 4.49 (1.25, 20.61) | 2.17 (0.76, 6.74) | - | 2.91 (0.83, 11.00) | 1.34 (0.39, 5.68) | 0.95 (0.24, 4.06) | 0.95 (0.14, 4.83) | 2.69 (0.57, 14.55) | 3.71 (0.78, 20.29) |
| 0.38 (0.05, 2.68) | Bosutinib 500 mg | 3.37 (0.84, 10.96) | 4.49 (1.23, 21.32) | 2.23 (0.77, 6.80) | - | 2.96 (0.84, 11.57) | 1.33 (0.38, 5.39) | 0.93 (0.24, 3.85) | 0.92 (0.13, 4.84) | 2.63 (0.60, 14.69) | 3.69 (0.81, 20.41) |
| 1.16 (0.21, 6.57) | 3.09 (0.62, 18.52) | Dasatinib 100 mg | 1.44 (0.72, 4.03) | 0.68 (0.39, 1.42) | - | 0.92 (0.42, 2.35) | 0.43 (0.17, 1.29) | 0.30 (0.11, 0.96) | 0.29 (0.05, 1.28) | 0.85 (0.24, 3.57) | 1.17 (0.34, 5.09) |
| 2.73 (0.45, 21.37) | 7.39 (1.25, 61.39) | 2.39 (0.83, 8.13) | Dasatinib 140 mg | 0.49 (0.18, 1.12) | - | 0.67 (0.24, 1.51) | 0.28 (0.09, 0.93) | 0.20 (0.06, 0.68) | 0.19 (0.03, 0.89) | 0.55 (0.13, 2.46) | 0.80 (0.17, 3.47) |
| 0.41 (0.09, 1.69) | 1.08 (0.26, 4.85) | 0.35 (0.13, 0.84) | 0.15 (0.04, 0.45) | Imatinib 400 mg | - | 1.34 (0.65, 2.72) | 0.61 (0.30, 1.36) | 0.43 (0.18, 1.01) | 0.41 (0.08, 1.49) | 1.21 (0.39, 4.15) | 1.67 (0.54, 5.87) |
| 0.23 (0.01, 2.16) | 0.62 (0.04, 5.94) | 0.20 (0.02, 1.50) | 0.08 (0.00, 0.65) | 0.57 (0.06, 3.44) | Imatinib 600 mg | - | - | - | - | - | - |
| 0.73 (0.14, 3.72) | 1.97 (0.40, 10.72) | 0.64 (0.21, 1.86) | 0.27 (0.07, 0.84) | 1.82 (0.89, 3.94) | 3.19 (0.46, 38.49) | Imatinib 800 mg | 0.44 (0.17, 1.37) | 0.31 (0.11, 0.97) | 0.30 (0.05, 1.35) | 0.87 (0.24, 3.73) | 1.21 (0.33, 5.46) |
| 0.35 (0.05, 1.94) | 0.95 (0.15, 5.41) | 0.30 (0.07, 1.08) | 0.13 (0.02, 0.54) | 0.87 (0.28, 2.25) | 1.52 (0.22, 14.47) | 0.48 (0.12, 1.52) | Nilotinib 600 mg | 0.73 (0.27, 1.74) | 0.64 (0.11, 2.84) | 2.00 (0.49, 8.31) | 3.04 (0.69, 11.32) |
| 0.37 (0.04, 2.01) | 0.99 (0.10, 5.33) | 0.32 (0.05, 1.18) | 0.13 (0.01, 0.57) | 0.91 (0.18, 2.58) | 1.58 (0.32, 7.67) | 0.50 (0.08, 1.69) | 1.05 (0.22, 3.27) | Nilotinib 800 mg | 0.92 (0.15, 4.43) | 2.70 (0.68, 12.73) | 4.32 (0.94, 17.77) |
| 0.81 (0.09, 7.18) | 2.17 (0.26, 19.65) | 0.71 (0.10, 4.17) | 0.30 (0.03, 1.89) | 1.99 (0.41, 9.96) | 3.48 (0.32, 65.77) | 1.11 (0.18, 6.06) | 2.31 (0.36, 16.90) | 2.22 (0.36, 23.40) | Ponatinib 45 mg | 3.08 (0.53, 23.25) | 4.29 (0.75, 31.27) |
| 0.32 (0.03, 2.71) | 0.860 (0.10, 7.22) | 0.27 (0.04, 1.69) | 0.11 (0.01, 0.80) | 0.78 (0.16, 3.85) | 1.39 (0.13, 21.86) | 0.43 (0.07, 2.39) | 0.90 (0.14, 6.35) | 0.86 (0.14, 9.16) | 0.39 (0.04, 3.59) | Radotinib 600 mg | 1.52 (0.41, 4.58) |
| 0.51 (0.06, 4.36) | 1.35 (0.16, 11.18) | 0.44 (0.07, 2.57) | 0.18 (0.02, 1.22) | 1.25 (0.24, 5.98) | 2.22 (0.20, 36.92) | 0.69 (0.12, 3.61) | 1.44 (0.23, 9.81) | 1.38 (0.22, 13.95) | 0.62 (0.06, 5.85) | 1.58 (0.32, 7.65) | Radotinib 800 mg |
| (B) Thrombocytopenia (grade 3–4) | |||||||||||
- All treatments are listed in alphabetical order. Comparisons between treatments should be read from left to right and the estimated ratio is in the cell in common between the column-defining treatment and the row-defining treatment. Values are presented as odds ratio (OR) with 95% credible interval (CrI). For neutropenia (grade 3–4; upper triangle, values in green) an OR <1 favours the event for the row-defining treatment while an OR >1 favours the column-defining treatment. For thrombocytopenia (grade 3–4; lower triangle, values in white) an OR <1 favours the event for the column-defining treatment while an OR >1 favours the row-defining treatment. Significant results between the interventions are highlighted in italic and bold.
For the outcome thrombocytopenia, similar to the other haematological adverse events, 140 mg dasatinib was the less safe option (97% probability in SUCRA; Table 3; Figure S5D). Lower doses of imatinib (400–600 mg) and 600 mg radotinib presented the lowest probabilities of causing this event (20–27% in SUCRA). Considering only the RCTs that assessed primary CML therapy, 400 mg imatinib was demonstrated to be safer than 400 mg bosutinib (OR 0.40 with 95% CrI [0.17–0.90]), 100 mg dasatinib (OR 0.50 with 95% CrI [0.27–0.84]), and 800 mg imatinib (OR 0.47 with 95% CrI [0.32–0.71]). In this scenario, 600 mg nilotinib presented a 27% probability of being the least safe drug (see Table 4B for complete consistency analysis).
Sensitivity analysis was also performed in all networks considering drug regimen, study design, and methodological quality; however, no additional differences were found from the original analysis (data not shown). All networks were found to be robust for the assessed outcomes. With respect to the inconsistency analysis and node-splitting technique, no substantial differences (P < .05) in magnitude or direction were identified between the results of the direct and indirect effects.
4 DISCUSSION
TKI drugs are commonly used to prevent cell proliferation in oncology patients with an overall satisfactory efficacy profile5, 52; however, this process can also result in myelosuppression, which manifests in patients as anaemia, leucopenia, neutropenia and thrombocytopenia.53 These haematological complications can lead to other AEs such as fatigue in patients with anaemia, opportunistic infections in those with leucopenia or neutropenia, and risk of bleeding associated with thrombocytopenia.54
Typically, myelosuppression occurs during the initiation of treatment, when a hypocellular marrow frame is observed; however, this phenomenon is rapidly reversed.3, 55 During long-term treatment, haematological AEs may also be found at high rates, mostly in advanced stage cancer. Cytopenia is a manifestation of the reduction in residual bone marrow reserve. With CML treatment success, healthy Philadelphia chromosome-negative bone marrow is unable to normalise peripheral blood counts after suppression of the abnormal clone, rather than the direct toxicity of the drug on hematopoiesis.1, 54, 56, 57 In addition to the c-Abl pathway, other avenues that play a key role in normal haematopoiesis are inhibited by TKIs, such as Src, KIT, and platelet-derived growth factor receptor (PDGFR) pathways, which further contributes to myelosuppression.53, 55, 58, 59
Our analysis shows that cytopenias were typically more frequent with the use of dasatinib in both doses (100 and 140 mg), despite 140 mg not being a usual dose in clinical practice. This drug has a broad spectrum of inhibition of tyrosine kinases (low specificity), including Bcr-Abl, c-KIT, the ephrin receptor (EphA2), Src family kinases (such as, Fyn, Yes, Src and Lyn), and PDGFR kinases, which are associated with the development of haematological toxicities .54, 60-64
Anaemia may be a consequence of c-KIT inhibition by TKIs. This receptor is considered essential for normal haematopoiesis and erythroid cell development.54, 60, 64-68 The possible main causes of leucopenia and neutropenia are the binding of drugs to a number of key kinase targets of the immune system, including Lck, Lyn, Btk and Src.69 Four families of TKs are associated with T-cell receptor (TCR) signalling, among which Src kinases (SFKs), such as Fyn and especially Lck, play a critical role in the development and activation of T cells.70-75 It is known that dasatinib inhibits TCR signalling, the expression of activation markers, and cytokine production and proliferation; however, this drug does not compromise the viability or induction of the proximal TCR signalling apparatus.73, 74 Furthermore, the effect of dasatinib is different depending on the T-cell type, with the inhibition being more prominent in CD4+ than in CD8+ cells, and more expressive in naïve than in memory T-cell subsets.70-72 In addition, the effects of dasatinib on T cells are dose-dependent and transient. Beyond that, some studies have reported that the reduction in the level of B cells is more evident than that of T cells; however, since the effect of dasatinib is lower during the mature stages of B lymphopoiesis, there are no significant changes in serum immunoglobulin levels.69, 71 Dasatinib also blocks the IgE-dependent activation of basophils and compromises the innate immune system.69, 71 Fortunately, leucopenia and neutropenia can be rapidly reversed with drug withdrawal or dose adjustment, which should be decided in clinical practice according to the patient profile and drug access.70 Similarly, imatinib inhibits Lck, which is in agreement with the results found in our study for leucopenia; however, the activity of this drug against Lck is lower as compared with dasatinib, i.e. the required concentration of imatinib to cause immunomodulatory effects is significantly higher.70, 72-75 The causes of thrombocytopenia are mainly related to the inhibition of SFKs, which are expressed in megakaryocytes and platelets. SFKs initiate and propagate signalling from several platelet surface receptors, in addition to affecting dissemination and migration to the vascular-rich niche. Therefore, despite the increased number of mature megakaryocytes in the bone marrow, inhibition of SFKs results in thrombocytopenia by compromising migration and proplatelet formation.53, 55, 62 Our results show that bosutinib and dasatinib were associated more with thrombocytopenia, which is probably due to their mechanism of action against SFKs. Moreover, dasatinib may also affect PDGFRs and their downstream signalling molecules, which contribute to the compromise of platelet activation.58, 64, 76
Our findings regarding nilotinib (both doses) demonstrate that this drug was less frequently associated with haematological grades 3–4 AEs, justified by its relatively high specificity and very narrow target spectrum.61
Haematological toxicities are generally clinically significant and require temporary treatment interruption or dose adjustments, sometimes even leading to treatment discontinuation. The nonadherence to TKIs is associated with a lower probability of complete remission and a poor prognosis.54, 77-79 Considering this, orientation to patients regarding the benefits of adherence and expected AEs with therapy is crucial.1 Moreover, treatment choice should not be based only on the effectiveness of the treatment, but also on the management of AEs and long-term tolerability.3, 54 Supportive care may be useful in situations of long-term and recurrent cytopenia, especially in patients with anaemia.3, 54, 77 Furthermore, considering the vast list of available TKIs, a change in therapy may also be considered in patients with significant AEs. Beyond that, the choice of treatment for CML should take into account haematological cross-intolerance among TKIs.1 No placebo-controlled trials were included in our systematic review and meta-analysis, probably given the complexity of CML, difficulties related to trials performance and ethical issues.
We recognise certain limitations in our analyses. Firstly, given the limited number of studies for some drugs, including ponatinib and radotinib, further analysis and conclusions were not possible. Secondly, all included trials were designed as open-label due to the severity of the disease; however, since the measured outcomes were objective, no further biases were associated with the study results. Subgroup analysis for second- and third-line TKIs could not be carried out, despite evidence reporting that haematological toxicity occurs in a higher proportion of patients using second- and third-line TKIs due to intolerance or resistance to prior TKIs.1, 3, 80 Furthermore, studies including patients in the accelerated and blast phases of CML, which are phases related to treatment resistance—which is considered as a decisive factor in the therapeutic selection—were not found.
5 CONCLUSION
In summary, the available evidence suggests that dasatinib can be superior to other TKIs in causing grades 3–4 haematological AEs. Furthermore, the safety profile of nilotinib (at both doses) was demonstrated to be a promising alternative treatment to avoid the SAEs that may lead to discontinuation or switching of therapy.
ACKNOWLEDGEMENTS
The authors would like to thank the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) for the scholarship.
COMPETING INTERESTS
There are no competing interests to declare.
CONTRIBUTORS
M.M.F., F.S.T., R.P. and F.F.L., contributed to the concept and design. M.M.F. and F.S.T., L.P.L., acquired the data. M.M.F., F.S.T., L.P.L., I.R. and F.F.L., analysed the data. M.M.F., F.S.T., L.P.L. and R.P., interpreted the data. M.M.F., F.S.T., L.P.L. and I.R., drafted the manuscript. F.F.L. and R.P., revised the manuscript.
