Introduction

Treatment of chronic myeloid leukemia (CML) has been profoundly improved by the introduction of tyrosine kinase inhibitors (TKIs). Long-term survival with imatinib is excellent with a 5-and 8-year survival rate of 90% and 88%, respectively.1, 2 The life expectancy of patients who achieve complete cytogenetic remission is not different from that of the general population,3, 4 and is influenced mostly by comorbidities.5

The increased life expectancy requires closer long-term observation of potential side effects. The development of secondary malignancies is regarded as a common risk of antineoplastic therapies. An increased rate of secondary malignancies compared with the general population has been reported in patients with Hodgkin’s lymphoma,6, 7 chronic lymphocytic leukemia8, 9 and other lymphoproliferative diseases,10 as well as in polycythemia vera11 and essential thrombocythemia.12, 13

An increased rate of secondary malignancies has also been described in patients who had received allogeneic stem cell transplantation14, 15, 16, 17, 18 for various hematologic diseases. Exposure to radiotherapy,19, 20 chemotherapy and immunosuppression, either disease or treatment related,21 have been suggested as risk factors for secondary malignancies.

TKIs have also been discussed as risk factors for malignancies. Preclinical data demonstrated an interaction of imatinib with DNA repair mechanisms.22 In studies with rats, neoplastic changes occurred in kidneys, urinary bladder, urethra, preputial and clitoral glands, small intestine, parathyroid glands, adrenal glands and non-glandular stomach.23

Another TKI effect that may be relevant for the development of malignancies is the inhibition of T-lymphocytes and dendritic cells. It has been shown that imatinib inhibits the effector function of T-lymphocytes and impairs the differentiation of peripheral blood progenitor cells into dendritic cells.24

These effects may facilitate the development of lymphatic malignancies during long-term exposure to imatinib.

In CML, data on the incidence of secondary malignancies are contradictory (see Supplementary Table 1). An increased rate of prostate cancer was found in a French cohort of 189 CML patients treated with imatinib.25 However, data from the Novartis registries of more than 9500 patients did not confirm this observation, but they were not obtained from randomized trials.26

Analyses of patient cohorts from multiple phase I and II trials at the MD Anderson Cancer Center, who were treated with TKI for CML and other myeloproliferative neoplasms, showed a risk of secondary malignancies that was lower than expected in the general population.27 In line with this is the analysis from Poland of 221 CML patients under imatinib treatment (median of 61 months) with no increase of secondary malignancies.28

In contrast, two other studies demonstrated an higher incidence in CML patients: (1) an analysis of the US-American SEER database found a significantly higher observed/expected ratio of secondary malignancies in the imatinib era of 1.48 versus 1.06 in the pre-imatinib era;29 (2) a study that crosslinked the Swedish CML register to the Swedish Cancer registry found a standardized incidence ratio (SIR) of 1.52 for a patient cohort from the imatinib era;30 and (3) in a cohort of 1038 Czech and Slovakian CML patients treated with TKI, the age-adjusted incidence rate of secondary malignancies was found to be 1.5-fold higher than that of the general population, but the difference was not statistically significant.31

However, CML itself has been discussed as a risk factor for solid cancers and hematologic malignancies. The acquired translocation t(9;22) at diagnosis of CML and additional chromosomal changes/mutations as a sign of clonal evolution during the course of disease show the potential of genetic instability in CML. Therefore, progenitors may already have the capacity to enforce themselves as distinct cells with enhanced malignancy resulting in solid cancers/hematologic malignancies before or later than CML.32 Two epidemiological studies that analyzed cancer registries for patients with CML in Sweden31, 33 and patients with myeloproliferative neoplasms including CML in Denmark34 showed an increased risk of secondary malignancies in CML patients.

To further elucidate the risk for the development of secondary malignancies in chronic-phase CML patients under treatment, we analyzed data from the CML Study IV after a median treatment duration of >5 years.

Patients and methods

Patients

CML study IV is a randomized 5-arm trial that compares imatinib 400 mg vs imatinib 800 mg vs imatinib 400 mg in combination with interferon-α vs imatinib 400 mg in combination with low-dose cytarabine and vs imatinib 400 mg after interferon failure.

The study was conducted as previously published.1

Inclusion criteria allowed the history of primary cancer if the disease was in stable remission without impact on study procedures. A total of 102 malignancies were reported in 92 patients before the diagnosis of CML. If relapses occurred within 5 years after diagnosis of primary cancer, they were not considered for further analysis.

Median follow-up for all patients after diagnosis of CML was 67.5 months (range, 0.12–124 months). Analysis was done according to intention-to-treat principle, that is, for patients on primary imatinib and after switch of therapy based on failure or intolerance.

Statistical analysis

SIRs35 were calculated from the age-specific rates from the German reference population, obtained from the Robert Koch Institute.36 Patients <15 years of age were excluded from the study and the groups 15–19 years and >85 years of age were not considered as >30 patient years were observed. As usual, non-melanoma skin cancer was not considered.36 The 95% confidence intervals (CIs) were calculated with ‘Mid-P exact test’ using the modification of Miettinen37 as previously described.38 Overall survival and progression-free survival as defined by the ELN (European LeukemiaNet) criteria39 were calculated using the Kaplan–Meier method.40 Cumulative incidences of second malignancies were estimated under the presence of the competing risk of death.41 Unless otherwise specified, date of diagnosis was considered as starting point for all time-to-event analyses. If one type of malignancy occurred more than once in a patient it was counted as one case according to the IACR (International Association of Cancer Registries).42

If not specified otherwise, all computations were done with SAS 9.2 (SAS Institute, Cary, NC, USA) or R 3.0.1.43

Ethics

The protocol followed the Declaration of Helsinki and was approved by the ethics committee of the ‘Medizinische Fakultät Mannheim der Universität Heidelberg’ and by local ethics committees of participating centers. Written informed consent was obtained from all patients before they entered the study.

Results

Patient characteristics

From February 2002 to March 2012, 1551 CML patients in chronic phase were randomized; 1525 were evaluable. Patient characteristics are described in Table 1 and Figure 1.

Table 1 Patient characteristics of available patients and patients with secondary malignancies
Figure 1
figure 1

Consort statement of the CML study IV and occurrence of secondary malignancies per recruitment arm.

Secondary malignancies

In total, 67 secondary malignancies in 64 (4.2%) patients were found after a median follow-up of 67.5 months. Of these patients, 26 were female (41%). The median age of these 64 patients at diagnosis of CML was 65 years (range, 30–88 years), and the median age at diagnosis of the first secondary malignancy after diagnosis of CML was 66 years (range, 31–88 years). The median time from diagnosis of CML to secondary malignancy was 2.4 years (range, 0.1–8.3 years).

So far, cumulative incidences of secondary malignancies among the five therapy arms are similar; the 5-year cumulative incidence varied between 1.9 and 6.3% (see Figure 1).

Two patients with secondary malignancy had been switched to second-generation TKIs (dasatinib: 1, nilotinib: 1) because of imatinib failure 3 days and 3 years, respectively, before the diagnosis of the secondary malignancies. In addition, one patient received allogeneic stem cell transplantation (>4 years before diagnosis of secondary malignancy).

Twelve of the patients with primary cancer before CML diagnosis developed malignancies under TKI treatment. Six of these patients had metastases or recurrence of the first malignancy 5–19 years after diagnosis of the primary cancers (two patients with breast cancer and one patient with cancer of unknown origin, prostate rectal and renal cell cancer).

The types of malignancies were: prostate (n=9, 13%), colorectal (n=6, 9%), lung (n=6, 9%), non-Hodgkin’s lymphoma (NHL; n=7, 10%), malignant melanoma (n=5, 7%), skin tumors (basalioma n=4 and squamous cell carcinoma n=1, 7%), breast (n=5, 7%), pancreas (n=4, 6%), kidney (n=4, 6%), chronic lymphocytic leukemia (n=3, 4%), head and neck (n=2, 3%), biliary (n=2, 3%), sarcoma (n=2, 3%), and esophagus, stomach, liver, vulva, uterus, brain and cancer of unknown origin (each n=1, 1%, see Table 2).

Table 2 Incidence, treatment and outcome of secondary malignancies in the German CML IV study (cases: n=67)

Three patients had more than one malignancy while receiving TKI. One patient developed a leiomyosarcoma and later a liposarcoma, one patient had a NHL that recurred (recurrence after 7 years) and one patient had prostate cancer and developed a NHL.

Outcome of patients with secondary malignancies

Of the 64 patients, 8 were in complete cytogenetic remission, 31 in major molecular remission at the time of diagnosis of the secondary malignancy. Two had progression of CML before diagnosis of the secondary malignancy, and one of these regained a remission before diagnosis of secondary malignancy (Table 2).

After diagnosis of secondary malignancies, CML treatment was continued without modification in 36 patients (56%). After a median follow-up time of 46.8 months (range 0–105 months) from time of diagnosis of the secondary malignancy, 26 patients had died. Of these patients, 22 died from the secondary malignancy, 2 from other causes (cerebellar infarction, infection) and 2 from unknown causes. Progression of CML was not a cause of death in any case. With a 4-year-survival of 57% (95% CI 43–70%, median overall survival 6.5 years), the overall survival and progression-free survival was significantly reduced in patients who developed secondary malignancies (Figures 2a and b).

Figure 2
figure 2

Probability of survival with or without the appearance of secondary malignancies. (a) Overall survival from time of diagnosis of CML. (b) Progression-free survival from time of diagnosis of secondary malignancy.

Statistical analysis

Cumulative incidences

Cumulative incidence of secondary malignancies in patients >50 years of age was significantly higher than in patients 50 years old (P<0.001): at 6 years the cumulative incidence was 8.1% (95% CI 9.2–10.5%) and 0.8% (95% CI 0.3–1.9%), respectively (Figure 3).

Figure 3
figure 3

Cumulative incidence of all secondary malignancies according to age 50 vs >50 years.

No significant differences were found between males and females (cumulative incidence 4.6% (95% CI 3.2–6.5%) vs 5.2% (95% CI 3.4–7.5%)) and EUTOS (European Treatment and Outcome Study)44 high- vs low-risk patients at 6 years.

SIRs for secondary malignancies (without non-melanoma skin tumors) in the CML population in comparison with the general German population were 0.88 (95% CI 0.63–1.20) in men and 1.06 (95% CI 0.69–1.55) in women (38 and 24 patients observed vs 43.0 and 22.7 patients expected in the matched German population, Figures 4a and b).

Figure 4
figure 4

SIRs of secondary malignancies within CML study IV for men and women compared with normal population for different tumor types. (a): SIRs of men (b) and SIRs of women.

Cancer subtypes

Regarding the subtypes of secondary malignancies, the numbers for prostate cancer, colorectal cancer, breast cancer, malignant melanoma, pancreas and kidney cancer in CML patients were not statistically significantly different from expected numbers of the general population. The SIRs were between 0.49 (colorectal in male) and 3.33 (kidney cancer in female) (Figure 4).

The number of cases of NHL however was significantly higher in the CML IV cohort than the expected number in the matched German population. The SIR for male was 3.33 (95% CI 1.06–8.04) and 4.29 for female (95% CI 1.09–11.66) (see Table 3 and Figure 4b).

Table 3 Standardized incidence rates of secondary malignancies (excluding non-melanoma skin cancer)

Discussion

Overall, our data do not support an increased risk for secondary malignancies in CML patients treated with imatinib as the SIR of men and women were similar to that of general population. However, looking at subtypes of malignancies we found a significant increase of the SIR for NHL for both sexes.

These data are in contrast to analyses of population-based registries in Denmark34 and Sweden33 that found an increased risk for secondary malignancies in CML patients. The observation timeframes for both studies were between 1970 and 2007 and between 1977 and 2008, respectively, and therefore mostly from the pre-TKI era. Both studies did not report on the specific treatment, but one can conclude that most commonly hydroxyurea and interferon-α were given to the patients during most of the time period. Knowing that BCR-ABL itself is a mutant driver of malignancy, this could explain the discrepancy of the studies.

The data from the SEER database contrast the above observations as in their study secondary malignancies in the pre-imatinib era were less common than in the imatinib era.29

A recent analysis of 868 CML patients from the Swedish CML registry diagnosed between 2002 and 2011 that were crosslinked to the Swedish Cancer registry showed a 50% overall increased risk of second malignancies compared with the normal population.30 This is in line with the study by Rebora et al.33 A possible explanation of the differences to our analyses is that we have a very well-described patient population with very good remission rates under imatinib-based treatments.45 Therefore, the BCR-ABL effect as described above may play a less important role in the CML IV trial cohort and may contribute to lower incidence rates of secondary malignancies.

In line with our observation is the study by Voglova et al.31 The age-adjusted incidence rate of secondary malignancies in their cohort of 1038 Czech and Slovakian CML patients treated with TKI was 1.5-fold higher than the normal population, but the difference was not statistically significant.31

Subtypes

There are several studies showing an increasing risk for subtypes of different cancers under TKI treatment. Verma et al.27 reported on secondary malignancies in patients with different myeloproliferative neoplasms including CML. They found a smaller number of secondary neoplasms than expected but an increased risk of melanoma, kidney and endocrine cancers.

In our study the increased rate of prostate cancer, the most common malignancy we found, was not statistically significant. This corresponds to data from the Novartis registries of clinical trials and adverse event reports of more than 9500 patients and more than 1 20 000 patient years and is in contrast to Roy et al.25

The increased frequency of NHL in our study was statistically significant. It must be considered that two of the seven cases occurred in patients who had already developed a secondary malignancy: in one case the documented NHL was a recurrence after 7 years, and the other case was a NHL in a patient with a previously documented prostate cancer. Another reason for the increased number of NHL cases may be that three of the seven cases were low-grade lymphomas that are easily missed in the general population but found in a monitored study cohort. We could not demonstrate a sex difference in appearance of NHL as this was shown by Radivoyevitch et al.46

Prevalence

In addition, a high number of patients (92 out of 1525, 6.0%) with malignancies that were diagnosed before the CML diagnosis were randomized to our study. In an analysis of the SEER database, Brenner et al.47 found that 14% of patients with CML had a malignancy before CML was diagnosed. Usually, in official publications like from the German Robert Koch Institute, cancer prevalence is reported as a period prevalence, for example, 5-year prevalence instead of point prevalence. Thus, no number for comparison exists directly. However, the cancer prevalence in our patient population seems to be high and a potential influence on the pathogenesis of CML can be discussed.

The diagnosis of secondary malignancies had a significantly unfavorable impact on overall survival and progression-free survival compared with other study patients of our trial. Remarkably, the cause of death in all these patients was not related to CML as no progression was observed.

Observation data from other disease entities, for example, Hodgkin’s lymphoma, indicate a long latency time between time after start of exposure to a risk factor and risk of secondary malignancies. The relative risk increased from 2.2 after 5 years to 10.9 after 20 years.7 Peaks for the rate of secondary malignancies were 5 to 9 years after chemotherapy and remained raised for 25 years.6 Therefore, longer follow-up of CML patients is warranted. In summary, there is no consistent distribution of malignancies in the different reports.25, 26, 27, 31, 33, 34 The risk of secondary malignancies is increased in population-based studies of CML patients,27, 34 but not increased in case–control studies of CML patients who are treated with TKI.27, 28, 31 So far, it is impossible to dissect patient selection in the observed patient populations from the impact of CML treatment on the risk of secondary malignancies.

Therefore, it is speculative if secondary malignancies occur after long exposure to TKI. Ideally, long-term follow-up on large cohorts of CML patients under treatment is warranted. As analyses of cancer registries often do not integrate complete data on treatment, a solution could be a registry on CML trial patients after end of study.