The Trend of Combined Modality Treatment and its Outcomes in Elderly Patients With Primary CNS Lymphoma; A 12-Year Population-Based Analysis Using Propensity Score

The addition of radiation to chemotherapy in elderly patients with PCNSL remains controversial. Our objective was to assess the trend of combined modality treatment (CMT) and compare its survival with chemotherapy alone and radiation alone in non-HIV patients. Methods We identied 6,537 patients who received single treatment modality, combined modality treatment, or no treatment at all between 2004 and 2015 from the National Cancer Database. Factors affecting treatment selection were investigated using a logistic regression model. Annual percentage change (APC) was calculated to assess the trend of CMT use. A propensity score weighting methodology was used to compare survival outcomes.

Introduction Primary central nervous system lymphoma (PCNSL) is an aggressive form of extranodal non-Hodgkin lymphoma that involves the brain, spinal cord, meninges, or eyes without systemic involvement. Half of the patients with PCNSL are above 60 years (1). The incidence of PCNSL is four cases per million persons every year (2). However, this incidence uctuated over the years with some decline that matched the improvement in human immunode ciency virus (HIV) management. This decline did not apply to elderly patients above 65 years, where the incidence continues to rise (3,4).
The introduction of high-dose methotrexate-based (HD-MTX) regimens has improved survival signi cantly compared with radiation alone. This observation was noticed for patients younger than 70 years old. Median survival in the elderly remains 6-7 months (5)(6)(7). The best treatment approach in older patients is unclear. A recent Dutch analysis showed that the use of chemotherapy alone in patients between 61-70 years increased from 6% in the period 1989-1995 to 36% in the period 2009-2015 (8).
Among American patients, the use of chemotherapy has increased from 65.6% in 2004 to 78.8% in 2013 (p-value <0.001) (9). This increase makes chemotherapy the most common treatment modality used nowadays, compared with earlier reports where radiation was the most common therapeutic approach (10).
The addition of radiation to HD-MTX remains controversial. Radiation in PCNSL needs to involve the whole brain, given the microscopically diffuse and multifocal nature of PCNSL. Although not compared in randomized trials, there was a consensus in the late 1990s and early 2000s that adding radiation to chemotherapy would cause up to a four-fold increase in overall survival (OS), with 5-year survival reached 50% in some reports (11)(12)(13). The role of consolidation radiation after achieving complete remission with HD-MTX was studied in phase III non-inferiority trial. The study showed that the median OS was not different in patients who received consolidation radiation (32.4 months) compared to patients who did not receive radiation (37.1 months) (p-value 0.71) (14) In the elderly, delayed neurotoxicity makes the combined modality approach less attractive. Neurotoxicity usually happens several months to years after treatment. Radiation was identi ed as an independent risk factor for neurotoxicity (15). The incidence was reported as high as 33%, with risk speci cally higher in patients >60 years old (15). A long-term analysis of 80 patients found that patients who received radiation did worse on attention/executive function, motor skills, and neuropsychological composite score, which affect the quality of life (16). After the onset of neurotoxicity, median survival is 1-2 years (15,17,18).
Previous population-based analyses focused on incidence and treatment trends. However, there is a lack of studies comparing long-term survival outcomes of elderly patients diagnosed with PCNSL and treated by different modalities (8-10). This comprehensive National Cancer Database (NCDB) analysis aims to assess the trend of combined modality treatment and investigate predictors affecting treatment selection. We also sought to estimate the OS in elderly non-HIV patients diagnosed with PCNSL and treated with either chemotherapy, radiation, or combined modality as a rst-line treatment from 2004-2015, using propensity score analysis.

Study design
We conducted a retrospective cohort analysis using de-identi ed data accessed from the NCDB. The NCDB is a joint program established in 1989 by the Commission on Cancer of the American College of Surgeons and the American Cancer Society (19). This comprehensive data set integrates registry records from more than 1500 accredited hospitals, capturing approximately 70% of all incident cancers in the United States (20). According to the agreements executed with each accredited facility, data from Veteran with HIV positive or unknown status, those with unknown status regarding chemotherapy or radiation administration. We excluded patients treated outside the reporting facility because, otherwise, the NCDB does not require documentation of their treatment and outcomes. We also excluded patients who started treatment with radiation or chemotherapy >120 days after diagnosis or started radiation > 365 days from diagnosis in the combined modality group to account for immortal time bias (Figure 1: Selection process, CONSORT diagram). Treatment was categorized into four groups: chemotherapy alone, radiation alone, combined modality treatment, and no treatment.

Variables and study outcomes
Race was recoded into four categories -non-Hispanic whites, non-Hispanic blacks, Hispanics, and others. Comorbidity was captured using the Charlson/Deyo comorbidity index (21). Socioeconomic data were provided as quintiles of median household income and number of persons with less than high school education in patients' census tract of residence. The type of facility was assigned according to the Commission on Cancer accreditation category based on annual case volume and available oncology services. Geographic locations corresponded to the U.S. Census Divisions. Insurance status is captured as it appears on the admission face sheet for the patient.
The primary outcome of the study is the overall survival (OS) in the four treatment groups. We estimated median OS as well as 12-and 24-month OS. Overall survival was de ned from the time of diagnosis to the time of death. We determined predictors of receiving any treatment modality compared to no treatment and predictors of receiving CMT compared to chemotherapy alone among treated patients. We also calculated the annual percentage change (APC) for CMT to assess its trend over the study period.
Propensity score estimation In this study, we obtained the average treatment effect (ATE) de ned as an estimate of interest. Variables included in the propensity score model included age, sex, race, insurance status, median income, education, treatment facility type, type of area, comorbidity score, and distance from the treating facility.
Inverse probability of treatment weighting was used in estimating weight-adjusted OS in the four treatment groups. We used four different methods to estimate weights: multinomial regression propensity score, generalized boosted propensity score, covariate balancing propensity score, and entropy balancing method. The choice of weighting method was based on achieving small coe cients of variations, large effective sample sizes, and low covariate balance assessed using standardized bias. A standardized bias-cut off less than 0.25 was used (22). We used balance tables and love plots to assess for covariate balance before and after weighting ( Figure 2). Generalized boosted modeling was used as a nal method to estimate weights based on the above criteria. The absolute standardized mean difference among covariates was used as a balance criterion. We used the propensity score package "WeightIt" coded for R statistical program (23). We used a robust variance estimator to account for within-person homogeneity (24).

Statistical analysis
Descriptive statistics were used to compare baseline characteristics of the four treatment groups. Continuous variables were presented as mean with standard deviation or median with interquartile range (IQR). Categorical variables were presented as absolute numbers and percentages. The means of continuous variables were compared using t-test or ANOVA, and percentages were compared using Pearson chi-square or Fisher's exact test. Overall survival was compared using log-rank and GehanBreslow-Wilcoxon rank tests. A strati ed log-rank test for every three years was used to account for possible variation in available and administered treatments. Univariable logistic regression was used to determine predictors of receiving any treatment versus no treatment and of receiving CMT versus single modality treatment. Those predictors were expressed as odds ratio (OR) and 95% con dence interval (CI). Statistically signi cant variables on univariable analysis were used to build multivariable logistic regression models.
To account for missing data, we created ve multiply-imputed lists using the "mi" package. (25) All ves imputed data sets were analyzed, and the OS estimate was combined using the Rubin procedure (26).
Regression diagnostics were used to evaluate model assumptions. All statistical tests were two-sided, and P-values <0.05 were considered statistically signi cant. We used R-statistical software (version 4.0.3) for statistical analysis (27).

Cohort characteristics
We identi ed 6,537 patients diagnosed with PCNSL between 2004 and 2015. The weight-adjusted and unadjusted baseline characteristics of the four treatment groups are summarized in Table 1 and Table 2, respectively. The median age was 74 years (IQR 69-87). There were 3430 (52.5%) females. The majority of the patients were non-Hispanic whites (86.8%), had a comorbidity score of 0 (63.1%), belonged to metropolitan areas (83%), were treated at a comprehensive community cancer center or academic program (47.1%), and had Medicare insurance (84.7%). The median duration to start chemotherapy and radiation was 18 days (IQR 9-31) and 35 days (IQR 16-83), respectively. The median total radiation dose was 3600 cGy in the radiation alone and CMT groups. Median number fractions were 19 and 20 in the radiation alone and CMT groups, respectively.

Treatment Selection
Chemotherapy alone was received by 3389 (51.5%) patients, 1452 (22.2%) patients received no treatment, 856 (13.1%) patients received radiation alone, and 840 (12.8%) patients received CMT. On multivariable logistic regression analysis (Table 3), the odds of receiving any treatment decreased with increasing age and higher comorbidity score. Being at an academic/research institution and having insurance were associated with more treatment administration. Patients with private insurance and Medicare received more treatment compared with no insurance. Among treated patients (Table 4), the use of CMT was signi cantly lower in older patients, patients who received treatment at an academic/research center, those with higher income, and those who live in rural areas.

Discussion
Combined modality therapy (CMT) was received by 12.8% of patients in the study cohort, with an annual percent change (APC) of -6.0 % (95% CI -8.0 to -4.0, p-value <0.001) over the 12 years. This percentage has almost decreased by half compared with the 1990s data (10). This is directly related to a better understanding of radiation-induced neurotoxicity in the elderly, which has reduced the acceptance of radiation as a treatment option. Neurotoxicity can render patients demented, ataxic, or incontinent and thus signi cantly affect the quality of life. Also, some reports showed decreased survival with the addition of radiation, as opposed to our results. The fact that more patients will die from disease progression rather than neurotoxicity even with the use of vigorous treatment modalities argues against its use (15). All this led the most recent clinical trials to study HD-MTX in different combinations rather than add radiation (28-30). Several studies have addressed the use of reduced-dose radiation in an attempt to decrease neurotoxicity without affecting outcomes. Although using doses of 3600 cGy have been shown to delay neurotoxicity, it did not prevent it (31,32). More promising results were evident when radiation was given at lower doses (23.4 cGy) (33, 34), which is the dose recommended by the National Comprehensive Cancer Network (NCCN) guidelines for consolidation radiation after achieving complete remission. (35) Patients in our cohort who received CMT had improved median OS compared with single-modality treatment. The bene t of CMT was more prominent in the rst year and decreased after that (table 5). Eventually, the curves cross over, indicating late toxicity and likely increased death in the combined modality group. Among treated patients, increased age, receiving treatment at an academic/research center, higher income, and living in rural areas decreased the odds of receiving CMT. The complexity of administering radiation that requires planning and frequent visits in addition to its toxicity probably led to this nding in older patients and those who live in rural areas. Being in academic/research programs increased the odds of receiving any treatment and receiving chemotherapy, two essential factors that can improve outcomes. This emphasizes the importance of referring patients to a center with experience in treating this deadly disease and implementing educational programs for smaller centers to optimize management.
Around 20% of our patients did not receive any treatment. This percentage seems to be stable over time (10). Our analysis identi ed important socioeconomic factors that affected treatment selection. The odds of receiving any treatment at all was decreased with age and comorbidity score. Having insurance was a predictor of receiving treatment as well, especially with Medicare or private insurance. This is important because not receiving treatment is associated with a dismal prognosis (median OS of 2 months in our analysis). More efforts are needed to expand the availability of treatment.
Our study is limited by its retrospective nature and the associated inherent bias. Observational studies are characterized by unequal and biased allocation of patients to different treatment arms. To address this problem, we used a propensity score weighting methodology in our analysis. All baseline variables were well balanced with standardized bias scores of ≤ 0.1 (Figure 2). Randomized clinical trials remain the gold standard to obtain causal inferences because of their ability to control for observed and nonobserved confounders. However, the generalizability of randomized trials might be limited due to the strict inclusion criteria. Because population-based studies provide real-life data of a large number of patients from different parts of the U.S, they tend to be more generalizable. As a sensitivity analysis, we used various methods to estimate weights in the propensity score models. All methods resulted in similar OS estimates.
Another limitation is that NCDB lacks pertinent variables like the exact chemotherapy regimen used and dosing, toxicity information, progression, and salvage treatment details. Also, Memorial Sloan-Kettering Cancer Center and the International Extranodal Lymphoma Study Group have PCNSL prognostic scores that include variables such as performance status, serum lactate dehydrogenase, CSF total protein concentration, and involvement of deep brain structures (36, 37) that we could not account for in our survival model due to unavailability in the NCDB.
This analysis shows that adding radiation to chemotherapy is falling out of favor in the real world in the elderly population. Despite that, a considerable percentage of patients (~13%) still received CMT. Careful patient selection when choosing this treatment approach is vital to decrease neurotoxicity while preserving outcomes. Future work should investigate factors associated with increased toxicity and identify the best radiation dose and technique. Chronologic age alone should not be the only deciding criterion; a comprehensive geriatric assessment (GA), not limited to performance status, should guide treatment decisions to achieve the best outcome possible with the minimum toxicity (38, 39). The survival advantage of CMT over single modality treatment should be conformed in a randomized controlled fashion. Validation of PCNSL-speci c GAs would be the ideal tool as many of the diseaserelated symptoms are present in these GAs and should be interpreted in the right context. Our analysis also showed important social and economic factors like gender, race, insurance status, and degree of urbanization that affect treatment decisions. Careful attention by the physicians to these factors is of utmost importance in daily clinical practice.

Declarations
Contributors: All authors had full access to all the data and analysis in the study and takes responsibility for the integrity of data and the accuracy of the data analysis. Ethical approval: This study used de-identi ed data and was considered exempt from human protection oversight by the institutional review board Data Sharing: Data was provided by the national cancer database (NCDB) Transparency statement: Yazan Samhouri MD a rms that the manuscript is an honest, accurate, and transparent account of the study being reported, that no important aspects of the study have been omitted, and that any discrepancies from the study as planned have been explained.      Figure 1 selection process (CONSORT diagram)

Figure 2
We used balance tables and love plots to assess for covariate balance before and after weighting ( Figure   2).