Background
Acute promyelocytic leukemia (APL) is a unique subtype of acute myeloid leukemia (AML) that is characterized by chromosomal translocation t(15;17) with fusion of PML-RARA genes and cell-cycle arrest at the promyelocyte stage.1 APL represents a hematologic emergency because of the risk of bleeding from hyperfibrinolysis, disseminated intravascular coagulation, and thrombocytopenia. The availability of targeted therapies, such as all-trans retinoic acid (ATRA), arsenic trioxide (ATO), and other agents, has remarkably transformed the outcomes of this disorder, with current cure rates exceeding 90%.2–5 However, this result is often a projection of data from clinical trials that included carefully selected patients and may not necessarily reflect population-level survival.6,7 In addition, earlier reported improvements in the population-level outcomes of APL relied on comparisons made between the ATRA era and the pre-ATRA era, when outcomes were historically poor. For example, Chen et al7 used the SEER database to describe the outcomes of patients diagnosed with APL in 1975 through 2008. Although the investigators showed a significant improvement in outcomes over time, the study included patients starting from the pre-ATRA era.
Recently, several notable advances related to APL have been observed. These include more reliance on targeted agents, molecular monitoring with minimal residual disease (MRD) assessment, effective salvage strategies, and better supportive care. There is also enhanced awareness about early mortality and late complications of APL and strategies to mitigate these complications. However, whether these advancements have continued to improve population-level outcomes of APL over time is unclear. In this study, we used the SEER database to describe recent trends in incidence, survival, early mortality, and secondary malignancies in patients with APL diagnosed in 2000 through 2014.
Methods
In a retrospective analysis of the SEER database, ICD-0-3 code 9866 was used to identify patients with APL in the SEER 18 registry (November 2016 submission data),8 which includes the following geographic areas: Alaska Natives, Atlanta (metropolitan), Connecticut, Detroit (metropolitan), Hawaii, Iowa, New Mexico, California excluding SF/SJM/LA, rural Georgia, San Francisco–Oakland, San Jose–Monterey, Seattle (Puget Sound), Utah, Kentucky, Los Angeles, Louisiana, New Jersey, and greater Georgia. The database provided information about patient demographic characteristics, type of malignancy, and survival duration; however, it did not provide information on patient symptoms, performance status, cytogenetics, MRD, treatment drugs, stem cell transplantation, or other prognostic factors.
Selection criteria for this study included age ≥20 years at diagnosis and histologically confirmed APL as the first primary malignancy. The diagnosis period was 2000–2014. All patients were undergoing active follow-up. Patient demographic characteristics, insurance status, diagnosis period, and survival duration were obtained. Insurance status was reported in the SEER database starting from 2007. Thus, the relationship between insurance status and outcomes was evaluated in a subset of patients with known insurance information. This research did not involve interaction with human subjects or the use of any personal identifying information, therefore it was exempt from Institutional Review Board approval, and informed consent was not applicable.
Statistical Methods
Descriptive statistics were used to compare baseline demographic characteristics. Categorical variables were compared using a chi-square test. Continuous variables were compared using an ANOVA F-test and a Wilcoxon rank sum test. The diagnosis period was divided into 3 equal groups (2000–2004, 2005–2009, and 2010–2014) to assess the temporal trend in improvement of outcomes in successive 5-year cohorts.
Incidence
The age-adjusted incidence rate per 100,000 population per year and the incidence rate ratio (IRR) were calculated using SEER*Stat software (version 8.3.5), with the 2000 US population as the reference standard.
Early Mortality
Early mortality was defined as death within 30 days of diagnosis. Logistic regression was used to evaluate changes in the probability of early death after adjusting for sex, age, race/ethnicity, marital status, and insurance status. The impact of age in early mortality analysis was modeled using a natural cubic spline to avoid strong assumptions on its effect.
Survival
Overall survival (OS) was calculated as time from diagnosis to death from any cause, and patients were censored if they were alive at last follow-up or end of the study. Cause-specific survival (CSS) was calculated from time of diagnosis to death from APL. Patients were censored if they died of other causes or were alive at last follow-up or end of the study. Survival probabilities were computed using the Kaplan-Meier method and compared using the log-rank test. Because of the high prevalence of early death in APL, a conditional survival analysis was also performed on patients who did not experience an early death, starting 1 month after diagnosis. Cox proportional hazards (PH) regression was used to conduct a covariate-adjusted analysis of the effect of cohort on survival. The PH assumption and the linearity of continuous predictors (age) were assessed graphically using plots of cumulative sums of martingale residuals. An initial screening found non-PH effects only for age at diagnosis, with higher effect during the first 12 months of follow-up for CSS (24 months for OS). There were no violations of the linearity assumption. Thus, the impact of age at diagnosis was modeled using a piecewise linear effect with separate hazards during the early (≤12 months for CSS, ≤24 months for OS) and late (>12 months for CSS, >24 months for OS) follow-up periods. After these adjustments, no further deviations from the assumptions were found.
Secondary Cancers
The observed number of secondary cancers in patients with APL and the expected number in an age-, sex-, and race-adjusted population were obtained using SEER*Stat software (version 8.3.5) for each unique diagnosis year and latency group. The standardized incidence ratio (SIR) was obtained as the ratio of the observed and expected values. Poisson regression for the observed counts with log-transformed expected counts as offset was used to obtain confidence intervals and estimate SIR ratios (SIRRs) and P values.
Additional Analyses
The regression models were used to evaluate the relationship between insurance status, survival, and early death. We performed a breakpoint analysis to determine whether there was a time point where outcomes started to improve significantly. The presence of a potential breakpoint in the linear relationship between year of diagnosis and the outcomes of interest was evaluated. We obtained an estimate with standard error for each year within the appropriate regression model (logistic for early mortality, PH for CSS, and Poisson for secondary cancers). Following this, Davies’s test was used in a linear regression model of the estimates on year using the inverse squared standard error as weights.9 No significant breakpoint was found (supplemental eTable 1, available with this article at JNCCN.org), with P<.05 indicating statistical significance. Statistical analysis was performed using SEER*Stat software (version 8.3.5) and SAS version 9.2 (SAS Institute Inc).8 The “segmented” R package version 0.5-3 (R Foundation for Statistical Computing) was used for the breakpoint analysis.
Results
Baseline Characteristics
We identified 2,962 patients with APL who met the study criteria. Median age of the cohort was 48 years (range, 20–96 years), with most patients aged 40 to 59 years (40%), male (51.4%), non-Hispanic white (58.7%), and married (59.9%). Table 1 shows the summary of baseline characteristics.
Baseline Characteristics
Incidence
The overall incidence of APL was 0.33 cases per 100,000 population per year. Table 2 summarizes the variations in disease incidence by demographic characteristics. The disease incidence increased significantly with age up to 79 years (IRR range, 1.27–1.61; P<.01) and increased over the period (IRR range, 1.32–1.35; P<.01). Disease incidence was lower in women (IRR, 0.88; P<.01) compared with men, whereas it was significantly higher among Hispanic versus non-Hispanic patients (IRR, 1.26; P<.01).
Disease Incidence
Survival
Median OS and CSS were not reached. When stratified by period of diagnosis, 4-year OS was significantly higher for patients diagnosed in 2010–2014 (73.4%) compared with those diagnosed in 2005–2009 (65.6%) and 2000–2004 (57.3%) (Figure 1). Similarly, 4-year CSS was significantly higher for patients diagnosed in 2010–2014 (78.3%) compared with those diagnosed in 2005–2009 (70.8%) and 2000–2004 (60.8%) (supplemental eFigure 1). When the analysis was restricted to patients who did not die within the first 30 days of diagnosis (ie, those who did not have early mortality), the improvement in OS and CSS over time continued to remain significant (Figure 2 and supplemental eFigure 2, respectively). A similar improvement in survival over time was seen across different age groups and demographic characteristics (supplemental eFigures 3–10).
Overall survival after diagnosis.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 18, 2; 10.6004/jnccn.2019.7351
Overall survival after diagnosis.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 18, 2; 10.6004/jnccn.2019.7351
Overall survival after diagnosis.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 18, 2; 10.6004/jnccn.2019.7351
Overall survival after 30 days from diagnosis.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 18, 2; 10.6004/jnccn.2019.7351
Overall survival after 30 days from diagnosis.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 18, 2; 10.6004/jnccn.2019.7351
Overall survival after 30 days from diagnosis.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 18, 2; 10.6004/jnccn.2019.7351
Multivariate Analysis of OS and CSS
According to multivariate analysis of post-30-day OS and CSS, diagnosis in the early periods (2000–2004 and 2005–2009) was associated with significantly higher mortality compared with diagnosis in 2010–2014 (Table 3). Unmarried patients had a higher mortality risk, whereas sex and race were not significant determinants of survival after adjusting for other demographic characteristics.
Multivariate Analysis of Post-30-Day OS and CSS
Early Mortality
Early mortality improved significantly over time, as illustrated in Figure 3 (2000–2004: 25.3%; 2005–2009: 20.6%; 2010–2014: 17.1%; supplemental eTable 2), and was affected significantly by increasing age (supplemental eFigure 11). In a logistic regression model, factors such as Hispanic ethnicity and diagnosis in the earlier periods (2000–2004 and 2005–2009) were associated with higher risk of early mortality, whereas women had a lower risk of early mortality (supplemental eTable 3).
Trend in early mortality.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 18, 2; 10.6004/jnccn.2019.7351
Trend in early mortality.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 18, 2; 10.6004/jnccn.2019.7351
Trend in early mortality.
Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 18, 2; 10.6004/jnccn.2019.7351
Second Malignancies
When stratified by period of diagnosis, the risk of secondary malignancy was not significantly higher compared with the general population and did not differ significantly over time (supplemental eTable 4 and eFigure 12).
Insurance Status and Outcomes
When restricted to the period 2007 through 2014 with available insurance information, the cohort consisted of 1,834 patients, with baseline characteristics summarized in supplemental eTable 5. Uninsured versus insured status was associated with a significantly higher early mortality (supplemental eTable 6). Although OS and CSS were inferior in uninsured versus insured patients overall, a conditional survival analysis for patients who survived >30 days showed that insurance was not a significant predictor of post-30-day survival (supplemental eTable 7 and eFigures 13–15). Age and marital status remained significant predictors of post-30-day CSS according to multivariate analysis (supplemental eTable 7).
Discussion
In this population-based analysis of the most recent cohort of patients with APL, we found a significant improvement in survival and early mortality over time. We also noted variations in disease incidence and outcomes by demographic characteristics. One of the remarkable achievements in acute leukemia has been the improvement in outcomes of patients with APL through several advancements. This began with the discovery of the PML-RARA mutation and identifying the role of anthracyclines in the 1970s through the 1980s, followed by the approval of ATRA and ATO in the 2000s, the development of risk-stratified treatment recommendations, and more recently, the approach of using upfront targeted agents. Several clinical trials and population-based analyses have demonstrated improved outcomes of patients with APL treated with ATRA compared with those treated in the pre-ATRA era.2–7 However, our study has shown a continued improvement in population-level outcomes of APL over time. Although these results are encouraging, certain disparities related to outcomes continue to exist. Thus, there is a significant opportunity for more research with the goal of uniformly improving outcomes.
Our study demonstrated an APL incidence of 0.33 cases per 100,000 population per year, with an increase over time. Geographic variations in the incidence of APL have been reported previously, with a higher incidence in parts of Latin America and stable incidence in Canada.10,11 However, the reason for these variations is unclear. In the United States, prior studies by Park et al6 and Chen et al7 described the incidence of APL to be approximately 0.18 to 0.23 per 100,000 population, with increased incidence over time.
Although the reason for increasing incidence over time is unclear, it could be related to multiple factors, including heightened awareness, diagnostic accuracy, and better reporting and capturing of new cases, in addition to host factors such as obesity, which has been associated with APL.12 Hispanic patients constituted 21.5% of the study cohort and had a higher disease incidence, similar to earlier reports showing increased incidence of APL and acute lymphoblastic leukemia in this population.13,14 The etiology for this variation in incidence by sex and race/ethnicity remains unclear. Genetic differences in the types of PML-RARA fusion in Hispanic patients have been previously described,15,16 raising the possibility of underlying genetic and environmental factors that could predispose to APL.
This study showed an incremental improvement in survival of patients with APL over time, likely reflecting a change in practice with the incorporation of targeted therapies, better supportive care, more effective disease monitoring, and improved management of complications. However, population-based outcomes remained inferior, with a 4-year OS rate of 66.1% compared with rates of 86% to 95% reported in clinical trials.3–5 In addition to age, we demonstrated the role of nonbiological factors, such as marital and insurance status, in influencing APL outcomes. Marital status influenced survival, and insurance status was noted to be a significant predictor of early mortality but was not significant for post-30-day survival. Given the limitations of the SEER database, with regard to the availability of information on treatment received or initial presentation, it is difficult to pinpoint a particular cause for these differences. Factors such as delayed presentation or the inability to receive timely therapy or complete therapy in the uninsured cohort could have led to these differences. Earlier studies illustrated that nonbiological factors, such as a support system and access to care, are important predictors of outcomes in many malignancies.17–19 Our study uniquely highlighted the importance of these factors in APL, an otherwise curable disease. We did not find any significant changes in survival by race and sex, although early mortality was found to be higher in men and Hispanic patients. Studies in APL and AML have shown varying results regarding the influence of sex and race on disease outcomes.6,20–22 In our study, Hispanic patients constituted the largest proportion of patients (47.9%) in the uninsured cohort. Lack of insurance could have influenced their risk of early mortality, similar to earlier reports showing insurance-related disparities in access to care.17,23 Although APL outcomes have improved over time regardless of age group, it is important to continue research efforts to address these discrepancies and improve outcomes.
Early mortality and secondary malignancies are known to affect the prognosis of APL adversely. Early mortality results from complications, such as hemorrhage, differentiation syndrome, and infections. Early mortality rates vary from 6% to 8% in clinical trials, 19% to 26% in single-institution studies, and up to 30% in population-based studies.6,11,24–26 Early mortality was also higher in the pre-ATRA era compared with the ATRA era, although further studies have shown no significant improvement.6,26 Over time, awareness of these complications has increased among clinicians, and advances have been made in prophylactic strategies. Our study showed a significant improvement in early mortality over time, and demonstrated the role of factors such as age, sex, race/ethnicity, and insurance status in influencing early mortality. Similarly, Ho et al27 recently demonstrated an improvement in early mortality in California. In that study, 30-day mortality in 1999 through 2002 was 33.3% and decreased to 22.74% for the period of 2011 through 2014 (P<.01). Our data are limited by lack of information on elements such as treatment setting (academic vs community hospitals), interval between symptom onset and treatment initiation, and other disease-related factors that could affect outcomes. With regard to the risk of secondary malignancies, our study did not find any significant variation over time among patients. Although the earlier approach of using anthracycline-based therapy poses a risk of secondary malignancy, the availability of chemotherapy-free frontline regimens could potentially lower this risk. The follow-up time in the most recent cohort (2010–2014) was relatively short and needs to be addressed in future studies with longer follow-up.
Limitations of our study include its retrospective design and lack of information on drugs received and factors such as leukocyte count, MRD status, comorbidities, and therapies like stem cell transplantation. However, our study provides the most updated, real-world information on the incidence and outcomes of a large cohort of patients with APL from a national population-based registry. Although outcomes of APL have improved over time, outcomes of non-APL AML have remained poor in general. Given the approval of novel agents for AML in recent years, outcomes are likely to improve in the upcoming years and should be investigated in future studies.
Conclusions
The population-level outcomes of APL have continued to improve over time in the United States. This improvement suggests the clinical impact of targeted therapies, early diagnosis, appropriate management of complications, and advanced treatment approaches. Still, significant differences in disease incidence and survival, influenced by nonbiological factors, continue to exist, highlighting the need for continued research to address these discrepancies.
Acknowledgments
We would like to thank Carreen O’Connor at Medical College of Wisconsin for her assistance in manuscript preparation and language edits.
References
- 1.↑
Wang ZY, Chen Z. Acute promyelocytic leukemia: from highly fatal to highly curable. Blood 2008;111:2505–2515.
- 2.↑
Adès L, Guerci A, Raffoux E, et al.. Very long-term outcome of acute promyelocytic leukemia after treatment with all-trans retinoic acid and chemotherapy: the European APL Group experience. Blood 2010;115:1690–1696.
- 3.↑
Lo-Coco F, Avvisati G, Vignetti M, et al.. Retinoic acid and arsenic trioxide for acute promyelocytic leukemia. N Engl J Med 2013;369:111–121.
- 4.↑
Burnett AK, Russell NH, Hills RK, et al.. Arsenic trioxide and all-trans retinoic acid treatment for acute promyelocytic leukaemia in all risk groups (AML17): results of a randomised, controlled, phase 3 trial. Lancet Oncol 2015;16:1295–1305.
- 5.↑
Powell BL, Moser B, Stock W, et al.. Arsenic trioxide improves event-free and overall survival for adults with acute promyelocytic leukemia: North American Leukemia Intergroup Study C9710. Blood 2010;116:3751–3757.
- 6.↑
Park JH, Qiao B, Panageas KS, et al.. Early death rate in acute promyelocytic leukemia remains high despite all-trans retinoic acid. Blood 2011;118:1248–1254.
- 7.↑
Chen Y, Kantarjian H, Wang H, et al.. Acute promyelocytic leukemia: a population-based study on incidence and survival in the United States, 1975-2008. Cancer 2012;118:5811–5818.
- 8.↑
SEER Program. SEER*Stat Database: incidence - SEER 18 Regs Custom Data (with additional treatment fields), Nov 2016 Sub (19732014 varying) - linked to county attributes - total U.S., 1969-2015 counties, National Cancer Institute, DCCPS, Surveillance Research Program, Surveillance Systems Branch, released April 2017, based on the November 2016 submission. Available at: https://seer.cancer.gov/data-software/documentation/seerstat/nov2016/. Accessed September 4, 2019.
- 9.↑
Davies RB. Hypothesis testing when a nuisance parameter is present only under the alternative: linear model case. Biometrika 2002;89:484–489.
- 10.↑
Rego EM, Jácomo RH. Epidemiology and treatment of acute promyelocytic leukemia in Latin America. Mediterr J Hematol Infect Dis 2011;3:e2011049.
- 11.↑
Paulson K, Serebrin A, Lambert P, et al.. Acute promyelocytic leukaemia is characterized by stable incidence and improved survival that is restricted to patients managed in leukaemia referral centres: a pan-Canadian epidemiological study. Br J Haematol 2014;166:660–666.
- 12.↑
Estey E, Thall P, Kantarjian H, et al.. Association between increased body mass index and a diagnosis of acute promyelocytic leukemia in patients with acute myeloid leukemia. Leukemia 1997;11:1661–1664.
- 13.↑
Swords R, Sznol J, Elias R, et al.. Acute leukemia in adult Hispanic Americans: a large-population study. Blood Cancer J 2016;6:e484.
- 15.↑
Douer D, Santillana S, Ramezani L, et al.. Acute promyelocytic leukaemia in patients originating in Latin America is associated with an increased frequency of the BCR1 subtype of the PML/RARα fusion gene. Br J Haematol 2003;122:563–570.
- 16.↑
Ruiz-Argüelles GJ, Garcés-Eisele J, Reyes-Núñez V, et al.. More on geographic hematology: the breakpoint cluster regions of the PML/RARα fusion gene in Mexican Mestizo patients with promyelocytic leukemia are different from those in Caucasians. Leuk Lymphoma 2004;45:1365–1368.
- 17.↑
Borate UM, Mineishi S, Costa LJ. Nonbiological factors affecting survival in younger patients with acute myeloid leukemia. Cancer 2015;121:3877–3884.
- 18.↑
Zheng Z, Zhu Y, Li X, et al.. Impact of marital status during diagnosis on cancer-caused specific survival in acute myeloid leukemia patients: a case-control and population-based study. Oncotarget 2017;8:62666–62680.
- 19.↑
Guru Murthy GS, Pondaiah SK, Abedin S, et al.. Incidence and survival of T-cell acute lymphoblastic leukemia in the United States. Leuk Lymphoma 2019;60:1171–1178.
- 20.↑
Sekeres MA, Peterson B, Dodge RK, et al.. Differences in prognostic factors and outcomes in African Americans and whites with acute myeloid leukemia. Blood 2004;103:4036–4042.
- 21.↑
Patel MI, Ma Y, Mitchell BS, et al.. Age and genetics: how do prognostic factors at diagnosis explain disparities in acute myeloid leukemia? Am J Clin Oncol 2015;38:159–164.
- 22.↑
Darbinyan K, Shastri A, Budhathoki A, et al.. Hispanic ethnicity is associated with younger age at presentation but worse survival in acute myeloid leukemia. Blood Adv 2017;1:2120–2123.
- 23.↑
Master S, Munker R, Shi Z, et al.. Insurance status and other non-biological factors predict outcomes in acute myelogenous leukemia: analysis of data from the National Cancer Database. Anticancer Res 2016;36:4915–4921.
- 24.↑
Lo-Coco F, Cicconi L, Breccia M. Current standard treatment of adult acute promyelocytic leukaemia. Br J Haematol 2016;172:841–854.
- 25.↑
McClellan JS, Kohrt HE, Coutre S, et al.. Treatment advances have not improved the early death rate in acute promyelocytic leukemia. Haematologica 2012;97:133–136.
- 26.↑
Lehmann S, Ravn A, Carlsson L, et al.. Continuing high early death rate in acute promyelocytic leukemia: a population-based report from the Swedish Adult Acute Leukemia Registry. Leukemia 2011;25:1128–1134.
- 27.↑
Ho G, Li Q, Brunson A, et al.. Complications and early mortality in patients with acute promyelocytic leukemia treated in California. Am J Hematol 2018;93:E370–372.
