Long-Term Risk of Subsequent Malignant Neoplasms Among Childhood and Adolescent Lymphoma Survivors (1975-2013): A Population-Based Predictive Nomogram

Abstract

Background

Studies are needed to assess risk factors pertinent to the incidence of secondary malignancies among childhood and adolescent lymphoma survivors. We aimed to identify risk factors pertinent to the incidence of secondary malignancies and subsequently establish a clinically practical predictive nomogram.

Methods

A total of 5561 patients who were diagnosed with primary lymphoma below the age of 20 years between 1975 and 2013 and survived for at least 5 years were identified. Standardized incidence ratio (SIR) and excess risk (ER) analysis were performed by sex, age, and year when primary lymphoma was diagnosed, sites and types of primary lymphoma, and therapy strategies. Univariable and multivariable logistic regression were used to identify independent risk factors for adolescent and childhood lymphoma-related secondary malignancies. Based on 5 factors (age, time from lymphoma diagnosis, gender, lymphoma type, and therapy), a nomogram for predicting the risk of a secondary malignancy for patients with childhood and adolescent primary lymphoma was established.

Results

Among 5561 lymphoma survivors, 424 developed a secondary malignancy. Females (SIR = 5.34, 95% CI, 4.73-5.99; ER = 50.58) exhibited a higher SIR and ER than males (SIR = 3.28, 95% CI, 2.76-3.87; ER = 15.53). Blacks were at a higher risk than Caucasians or others. Nodular lymphocyte-predominant Hodgkin lymphoma survivors exhibited typically high SIR (13.13, 95% CI, 6-24.92) and ER (54.79) among all lymphoma classifications. Lymphoma survivors who underwent radiotherapy, whether they received chemotherapy or not, had typically higher SIR and ER. Among all types of secondary malignancies, “bone and joint neoplasms” (SIR = 11.07, 95% CI, 5.52-19.81) and “soft tissue neoplasms” (SIR = 12.27, 95% CI, 7.59-18.76) presented significantly high SIR whereas “breast cancer” and “endocrine cancer” associated with higher ER. The median diagnosis age of secondary malignancies was 36 years old, and the median time interval between the diagnosis of two malignancies was 23 years. A nomogram was constructed to predict the risk of secondary malignancies in patients diagnosed with primary lymphoma before 20 years of age. After internal validation, the AUC and C-index of the nomogram are 0.804 and 0.804, respectively.

Conclusion and Relevance

The established nomogram provides a convenient and reliable tool for predicting the risk of a secondary malignancy among childhood and adolescent lymphoma survivors, concluding significant concern for lymphoma survivors with high-risk estimates.

Introduction

Life expectancy is significantly shorter among individuals who survived childhood cancer.¹ Recent advancements in therapeutic modalities have optimally increased the overall 5-year survival rate to 80% among individuals suffering from pediatric cancers.² Hodgkin lymphoma is the most common cancer diagnosed in children and adolescents.³ It has been reported that lymphoma survivors may experience a 1.6 to 4.3 times higher incidence of second malignancies compared to pediatric patients who had solid tumors and were cured.⁴ A study in the Netherlands reported the incidence risk of secondary cancers in Hodgkin lymphoma patients diagnosed at the age of 15-50 years and survived for at least 5 years to be 13.9% at 30 years.⁵ Moreover, previous studies have reported the significant influence of radiotherapy or chemotherapy in the development of second malignancies among adolescents who were treated for Hodgkin lymphoma or non-Hodgkin lymphoma (NHL).^6,7 Nowadays, the incidence of secondary malignancies remains a major problem in long-term lymphoma survivors, hence risk factors related to incidence should be identified to develop a suitable screening strategy.

Furthermore, studies demonstrated that adolescent cancer survivors had a higher absolute risk of secondary malignancies compared to younger or older cancer survivors.⁵ Nomograms could be used to explore prognostic strategies by ascertaining the patient’s overall survival information.⁸ There is no reported predictive nomogram evaluating the risk of secondary malignancies for childhood and adolescent lymphoma survivors. The current study analyzed the risk of developing secondary malignancies among childhood and adolescent lymphoma survivors depending on factors such as gender, age at diagnosis, year of diagnosis, incidence site, histological type, and treatment exposure data. Such a nomogram could benefit clinicians and oncologists to guide the ongoing surveillance of high-risk survivors.

Methods

Study Population

Patient data were downloaded from a database of the Surveillance, Epidemiology, and End Results (SEER) project, the SEER 9 1975-2018 (November 2020 sub), race (White/Black/other [W/B/O]), containing 9 SEER registries and covering approximately 9% of the USA population.⁹ Ethical approval was waived for this study as the data are publicly available.

The selection criteria for this study:

1. (1) patients diagnosed with primary lymphoma less than 20 years old between January 1, 1975 and December 31, 2013

2. (2) lymphoma diagnosis was pathologically confirmed

3. (3) survival time of more than 5 years

The exclusion criteria: (1) the Ann Arbor Stage of lymphoma is unknown, (2) secondary malignancy is lymphoma, and (3) diagnosed with a secondary malignancy within 1 year after a primary lymphoma diagnosis. In this study, secondary malignancy is defined as a different type of malignancy that develops in the same individual 1 year after the initial cancer diagnosis; this strategy prevents recognizing the relapse of “primary lymphoma” as “subsequent malignancy” and helps to minimize the chances of surveillance bias, as the likelihood of incidental cancer detection could be higher immediately following the initial cancer diagnosis.¹⁰ The primary lymphoma and secondary malignancy were classified according to ICD-O-3.^11,12 Follow-up ended at the subsequent diagnosis of malignancies or death after December 2018 whichever occurred primarily.¹³

Relative and Absolute Risk Analysis

According to the data obtained from the 9 registries of the SEER database, the standardized incidence ratio (SIR) is based on the “observed count,” which refers to the number of lymphoma survivors who developed secondary malignancies for a selected cohort, and the “expected count,” which refers to the hypothetical number of secondary malignancies calculated based on the “accumulation of person-years at risk” for the selected lymphoma survivors and the “incidence of a specific malignancy in the USA general population” adjusted by gender, race, type of subsequent malignancies, age, and year of the diagnosis of a secondary malignancy. SIR describes the relative risk and was defined as the observed count divided by the expected count. It was statistically significant when the P value < .05. Confidence limits for SIR were calculated by the methods discussed by Sahai and Khurshid, assuming that the number of observed events followed the Poisson distribution.^14,15 Excess ratio describes absolute risk and the formula for ER calculation is (observed count − expected count) * 10 000)/person-years at risk.¹⁶ The expected number, SIR, and excess risk (ER) were calculated using SEER*Stat software. In SEER*Stat, the “unknown” related to treatment depicted the unavailability of evidence in medical records pertinent to chemotherapy and radiation therapy.

Logistic Regression Analysis and Nomogram

Variables were selected through univariable and multivariable logistic regression analysis to construct the nomogram. We determined a cutoff point of “age at lymphoma diagnosis” by using Receiver Operating Characteristic (ROC) analysis with the Youden index. Time after diagnosis of lymphoma was defined as “intervals between the year of diagnosis and 2018” and divided into groups at 10-year intervals. Univariable logistic regression screened out the related prognostic factors. Then, a multivariable logistic regression analysis was used to analyze the independent risk factors. The variables with a P value less than .05 were considered statistically significant.

Based on the factors with a P value less than .05 in multivariable logistical regression, a nomogram was established to predict the risk of secondary malignancies for childhood and adolescent primary lymphoma patients. Subsequently, we used a series of methods to validate this nomogram, including the area under the receiver operating characteristic curve (AUC), the calibration curve, and the consistency index (C-index). AUC and C-index were used to assess the discrimination of the model. The calibration curve (1000 bootstrap resamples) was mainly used to test its accuracy. Statistical analyses were performed using SEER (version 8.3.9.2) and R (version 4.0.1).

Results

Patient Characteristics

Our study included 5561 patients from the SEER database. Among these patients, 424 developed a secondary malignancy at least one year after the primary diagnosis. The distribution of categorical variables including gender, race, incidence site of lymphoma, histological types, therapy, age at diagnosis, and year of diagnosis was compared between all populations of lymphoma survivors and those who developed a secondary malignancy (Table 1). A higher number of males (57.7%) compared to females (42.3%) among lymphoma survivors was observed, while the number of males (33.0%) was disproportionately less than females (67.0%) among those who developed a secondary malignancy. Among the survivors who developed a second malignancy, classical Hodgkin lymphoma (78.3%) was the most common primary diagnosis. According to the results of the Chi-square test, risk factors that may play a role in the development of a secondary malignancy include sex, age, and year when primary lymphoma was diagnosed, sites, and histological types of primary lymphoma, and therapy strategies.

Relative and Absolute Risk Analysis of Secondary Malignancies

In Fig. 1, we showed the relative and absolute risks of developing a secondary malignancy for lymphoma survivors compared to the general US population. The SIR (4.42, 95% CI, 4.01-4.86) and ER (30.29 per 10 000 person-years) describe the relative and absolute risk of developing a secondary malignancy for adolescent lymphoma survivors, respectively, and these values were significantly higher than the general values. Females (SIR = 5.34, 95% CI, 4.73-5.99; ER = 50.58) exhibited a relatively higher SIR and ER than males (SIR = 3.28, 95% CI, 2.76-3.87; ER = 15.53). Blacks were at a higher risk than Caucasians or others. Adolescent Hodgkin lymphoma survivors exhibited a high relative risk and reported >3 times higher absolute risk when compared to NHL survivors. Nodular lymphocyte-predominant Hodgkin lymphoma survivors exhibited typically high SIR (13.13, 95% CI, 6-24.92) and ER (54.79) among all lymphoma types. In addition, lymphoma survivors who underwent radiotherapy, whether they received chemotherapy or not, had typically higher SIR and ER (both radiotherapy and chemotherapy: SIR 5.27, ER 34.72; radiotherapy only: SIR 5.19, ER 54.63) than those who did not receive it.

In addition, we further analyzed the risk of developing a secondary malignancy among specific-type lymphoma survivors who received different treatment strategies (Table 2). Among the patients who received radiotherapy and chemotherapy, extranodal NHL was associated with higher SIR and ER. Survivors who were treated with combined radiotherapy and chemotherapy typically exhibited a higher relative risk when compared to those who adopted other treatment modalities, except for nodal NHL patients. Patients who received radiotherapy and no/unknown chemotherapy had a relatively higher absolute risk when compared to those receiving other treatment modalities, except for extranodal NHL. Nodal HL was associated with the highest risk in cohorts who have not received radiotherapy.

In Table 3, we have analyzed the risk of secondary malignancies between different types of secondary malignancies based on two categorical variables: gender and lymphoma types. Adolescent lymphoma survivors were associated with the highest relative risk of developing secondary soft tissue neoplasm (SIR: 12.27, 95% CI, 7.59-18.76), followed by secondary bone and joint neoplasms (SIR: 11.07, 95% CI, 5.52-19.81). Soft tissue neoplasms exhibited the highest SIR among Hodgkin lymphoma survivors, while bone and joint neoplasms showed the highest SIR among NHL survivors. However, the highest absolute risk appeared for breast cancer (ER = 10.09) among Hodgkin and NHL survivors.

To elucidate the characteristics of secondary malignancies associated with the duration of lymphoma, the distribution of time intervals between the diagnosis of lymphoma and secondary malignancies and the results of the risk analysis were shown in Fig. 2. The median age of diagnosis for secondary malignancies was 36 years, and the median interval time between the diagnosis of two malignancies was 23 years. Among them, two groups (interval 1-5 years, 16-20 years) were associated with significantly high SIR. ER in various interval groups increased with the length of the interval and ranged from less than 11 to over 147 per 10 000 person-years at risk. Cumulative incidence rate vs. months from initial diagnosis was analyzed by different treatment methods Supplementary Figure 1 but no statistically significant difference was observed (Fig. 2C). The relative and absolute risks of lymphoma survivors who received various treatment modalities in different lymphoma diagnostic year segments were shown in Supplementary Table 1. Their absolute risk was decreasing as the year of diagnosis increased.

Univariable and Multivariable Analysis

The results of univariable and multivariable logistic regression analysis were reported in Table 4. Univariable analysis revealed that 6 out of 7 variables were associated with the development of secondary malignancies among primary lymphoma survivors. These variables were age, time from lymphoma diagnosis, gender, lymphoma site, types, and therapy. Considering the variables identified in the multivariable analysis, we developed a new nomogram for predicting the risk of lymphoma-related secondary malignancies among childhood and adolescent primary lymphoma patients based on factors including sex, age at diagnosis of primary lymphoma, time from lymphoma diagnosis to 2018, type, and therapy (Fig. 3).

Construction and Validation of Nomogram

Each variable was distributed on the nomogram according to its contribution to risk (Fig. 3). The level of the different factors should be identified and then the corresponding number of designated points could be added to the total. For example, being female gender adds 30 points to the total while being male adds 0. By adding all these points to reach a total depending on the real condition of a patient, we can obtain a total point number that has a corresponding decimal on the “Risk of subsequent cancer” line. The risk probability for the incidence of secondary malignancies is (the decimal × 100)%. The nomogram was internally validated using the ROC curve and calibration curve. Nomogram exhibited superior predictive power as indicated by the AUC value of 0.804. Subsequently, we employed 1000 bootstrap resamples to evaluate the accuracy of the nomogram. The concordance index (C-index) was observed as 0.804 and the bias-corrected ­c-index was 0.800. In summary, the predictive plot matched with actual probabilities. The established and validated nomogram is capable of predicting the risk of secondary malignancies among primary lymphoma survivors.

An online version of our nomogram can be accessed at https://zzuxz.shinyapps.io/Nomogram-For-Secondary-Malignancies/ to assist researchers and clinicians in obtaining accurate predictions intuitively and easily by giving clinical features as the significant inputs and subsequently reading output figures and tables generated by the web server.

Discussion

A previous study reported the cumulative incidence risk of subsequent neoplasms was 26.3% at 30 years after Hodgkin lymphoma diagnosis among 1380 childhood patients diagnosed between 1955 and 1986.¹⁷ A few reports also described the risk of secondary malignancies such as lung cancer and thyroid cancer among childhood Hodgkin lymphoma survivors.¹⁸ But, most of these studies only focused on patients of one specific lymphoma type and all age groups, and there were no studies that precisely predicted the risk of secondary malignancies in childhood lymphoma survivors. We selected lymphoma patients diagnosed below 20 years of age who had survived for at least 5 years because the proportion of such survivors has been increasing, and there is a need to unravel their health status after treatment. The current study is a large and comprehensive study that compares the relative and absolute risk of secondary malignancies for childhood and adolescent lymphoma patients in different clinical conditions and identifies risk factors to build a clinically applied predictive nomogram.

In our study, a few of the results pertinent to the incidence of secondary malignancies in primary lymphoma survivors are consistent with previous reports. Gender has been shown to play a crucial role, and female gender is one of the significant risk factors. Older age at the time of diagnosis also enhances the risk of developing secondary malignancies. Hodgkin lymphoma survivors were reported to experience the highest risk of acquiring secondary malignancies. In addition, radiotherapy has elevated the risk of subsequent malignant neoplasms up to 2.7 times (95% CI, 1.0-7.7) among lymphoma survivors of all ages; chemotherapy administration for any primary malignancy also enhances the risk of developing secondary malignancy (SIR = 6.3, 95% CI, 4.1-9.4).¹⁹

There are several established nomogram prognostic index models depending on the pretreatment indicators to predict the survival conditions of lymphoma patients, which have proven quite useful. Hongyan Li et al described the efficacy of a nomogram to provide a numerical probability of overall survival for patients suffering from extranodal natural killer/T-cell lymphoma.²⁰ Jun Cai et al constructed a nomogram to predict the overall survival for patients with diffuse large B-cell lymphoma (DLBCL), which produced a good clinical prognostic stratification for DLBCL.²¹ Analysis of the underlying factors for the risk of secondary malignancies helps clinicians to identify potential risk factors and subsequently enables medical practitioners to design a more ­patient-specific follow-up strategy. This nomogram could be a useful and novel tool to predict the long-term risk of subsequent malignancies for childhood and adolescent lymphoma survivors. The predictors included in this nomogram are considered as significant and easily available in clinical practice.

In this study, we performed a statistical analysis of subsequent malignant neoplasm incidence based on >5000 cases of lymphoma survivors observed throughout 39 years. This study is the first report that explored underlying risk factors associated with the incidence of secondary malignancies in a childhood and adolescent lymphoma patient cohort. We also developed a nomogram to quantitate the risk of secondary malignancies using a simple tool, which was associated with great predictive power. For clinicians, we illustrated the peak time of the onset of secondary malignancies and comparably common types of secondary cancers incidence among lymphoma survivors of different genders and lymphoma types. This nomogram can provide a quantified risk in real clinical practice, for survivors who exhibited a high-evaluated risk, regular, and detailed body examination should be recommended. For patients, knowing the risk of developing another neoplasm may increase their attention to their body and consequently avoid original lifestyle risk factors. However, obtaining this information may also increase the patients’ psychological burden, so we recommend clinicians use this nomogram initiatively but reach a consensus with the patients before telling them the risk of another fatal condition.

However, our study still has some limitations. First, our findings can be applicable to the US population with distinct demographic features, as we procured data from the SEER registries, but these findings need to be further validated in other racially and socially diverse populations. In addition, the nomogram has fewer variables due to the limitation of data availability. The results of laboratory tests and detailed treatment information pertinent to lymphoma should be incorporated into the analysis to further rule out confounding caused by these factors, such as radiation methods or specific types of chemotherapy drugs, older radiation methods are more likely to increase second malignancy risk than new ones. Furthermore, more studies should be conducted to stratify the patients with different risk estimates and develop a matching follow-up and screening strategy to achieve good clinical outcomes. Concomitantly, it is crucial to develop therapies with reliable curative capacity for treating primary lymphomas without carcinogenic potential.

Conclusion

Lymphoma patients diagnosed before the age of 20 years from 1975 to 2013 and who survived for at least 5 years in the SEER database were selected. We first compared the relative and absolute risk of a secondary malignancy with the general US population and then identified independent risk factors by univariable and multivariable logistic regression analyses, resulting in a quantitative nomogram for predicting the risk of secondary malignancies. This nomogram can help clinicians effectively to identify high-risk patients and raise awareness of recurrent follow-up and screening them in advance for choosing personalized medicine.

Note: Person-time at risk contributed by this subset of your cohort, expressed in years. Person-time at risk: For a particular subject, person-time at risk is the time in years between entry into an analysis and exit from an analysis. This is the time during which a subject is at risk of having an event (whether second cancer, death, etc.) as defined by the analysis. Total person-time at risk for a given subcategory is calculated by adding the person-time at risk counted in that subcategory for each subject in the analysis.

Funding

This research was supported by the Henan Natural Science Foundation of China (grant nos. 222300420534 and 222102310106) and Henan Medical Research Program (SBGJ2020002081 and LHGJ20190249).

Conflict of Interest

The authors indicated no financial relationships.

Author Contributions

All authors contributed to the conception/design: all authors. Literature search: J.L., Q.Z., N.M.B., T.L., R.S. Collection and/or assembly of data: All authors. Data analysis and interpretation: J.L., N.M.B., T.L.

Manuscript writing: J.L., Q.Z., N.M.B., T.L., R.S. Final approval of manuscript: All authors.

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.

References