Cognitive Trajectories in Older Adults Diagnosed With Hematologic Malignant Neoplasms.

Introduction

More people with hematologic malignant neoplasms (HMN) are surviving long-term with the development of effective, well-tolerated treatments. 1 , 2 Many patients with HMN have the potential for cure or long-term disease control. In 2024, the 5-year relative survival for the 2 most common HMNs (non-Hodgkin lymphoma and leukemias) are 74.3% and 67.0%, respectively. 3 Understanding the impact of HMN on long-term outcomes and quality of life in older survivors is increasingly important. Cognitive impairment affects quality of life and functional status in cancer survivors, and the risk of cognitive decline is an important consideration in treatment decisions for older adults with serious illnesses. 4 , 5 Cognitive decline is often reported after cancer diagnosis or treatment. Unfortunately, the literature on cancer-related cognitive impairment (CRCI) in HMN is limited, with most research focused on solid tumors. 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13

HMN biology and treatments may impact cognition differently from solid tumors. First, HMN and its treatments cause more cytopenias, which are associated with increased fatigue, decreased physical activity, and cardiovascular risk, which may impact cognition. 14 , 15 , 16 Second, since HMNs are systemic diseases affecting the immune and hematopoietic systems, their disease biology may influence cognition independent of treatment, as has been suggested in chronic lymphocytic leukemia and lymphomas. 17 , 18 , 19 , 20 Lastly, unlike solid tumors, which in early stages may be treated with surgery or radiation, most HMNs are treated with systemic therapy, which is associated with higher CRCI risk. 8 , 13 Improved understanding of what cognitive changes to expect in HMN survivors is important for patient counseling and management.

In general, the study of cognitive change in aging and cancer is fraught with methodological challenges. CRCI studies have found mixed results, likely reflecting diverse patient populations, methodologies, cognitive assessments, and choice of controls. 8 , 21 , 22 , 23 Some studies have baseline pretreatment cognitive performance, but most studies do not have precancer cognitive trajectories. Many factors can influence cognition: age, comorbidities, genetics, health behaviors, functional status, cognitive reserve, and psychosocial factors. 6 , 10 , 11 These factors are important to consider when evaluating CRCI so we understand how observed changes differ from what would be expected with normal aging.

The goal of this study was to model cognitive trajectories and compare rates of cognitive decline before, during, and after diagnosis in older adults diagnosed with HMN, compared with propensity score–matched participants without cancer. We hypothesized that the rate of cognitive decline would be faster after HMN diagnosis compared with normal aging.

Methods

Data Source

We used data from the Health and Retirement Study (HRS), which is a prospective longitudinal panel study surveying a population-representative sample of community-dwelling US adults since 1992. 24 HRS participants undergo detailed interviews every 2 years from cohort entry at age 50 or older until death or dropout, with data collected about health, cognition, and more. HRS is sponsored by the National Institute on Aging and conducted by the University of Michigan. Informed consent and data collection were approved and overseen by the University of Michigan institutional review board. Ethical approval for data analysis for this study and waiver of patient consent due to use of deidentified data were obtained from the University of California, San Francisco Committee on Human Research. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology ( STROBE ) guideline for reporting observational studies.

Study Participants

To create the HMN cohort, we identified HRS participants who agreed to release their Medicare records (approximately 80% of cohort). 24 Participants must have been diagnosed with an HMN from 1998 to 2016 and be older than 65 years when diagnosed. Participants were excluded if they had dementia at diagnosis, lacked an HRS interview within 2 years before diagnosis, or did not have any Langa-Weir cognitive scores during their follow-up. Dementia status at diagnosis was determined by the cognitive category assigned based on Langa-Weir cognitive scores or proxy cognitive scores from the HRS assessment immediately preceding diagnosis. Langa-Weir cognitive scores from 1992 to 1998 and 2016 to 2020 were included in the trajectory analysis if available.

To create the control cohort, we identified HRS participants who were propensity score–matched to an individual in the HMN cohort in the same HRS wave. Participants without cancer were excluded if they missed responding to the wave used for matching, were aged 65 years or younger at time of matching, had dementia at time of matching, had a diagnosis of any cancer (hematologic or solid) between 1998 and 2016, or did not have any Langa-Weir scores during their follow-up.

Measures

Our exposure of interest was diagnosis with an HMN. Cancer diagnoses were defined by the International Classification of Diseases, Ninth Revision and Tenth Revision diagnosis codes in Medicare inpatient and outpatient billing files between 1998 and 2016, as primary or secondary diagnoses (eTable 1 in Supplement 1 ). 25 , 26

Outcome Variables

Our primary outcome was cognitive function as assessed by the Langa-Weir global cognitive summary score, which was developed and validated for identifying dementia in HRS and correctly identifies dementia in 52.18% and normal cognition in 87.17% of cases. 27 The Langa-Weir is one of the standard cognitive assessments used extensively in many HRS studies. 28 , 29 It is composed of 4 items: immediate 10-word recall (0-10), delayed 10-word recall (0-10), serial sevens to assess working memory (0-5), and backward counting to assess attention and processing speed (0-2). Scores range from 0 to 27 points, where 0 to 6 is dementia, 7 to 11 is cognitive impairment without dementia, and 12 to 27 is normal.

Other Key Variables

Sociodemographic and health-related variables were selected a priori for incorporation into propensity scores. Values were drawn from the HRS interview within 2 years before diagnosis. When available, variables cleaned and processed by HRS or RAND were used. 30 , 31 , 32 In total, 23 variables relevant to cognition were incorporated into propensity scores: sociodemographics (age, sex, self-reported race and ethnicity, education, marital status, and wealth), comorbidities (diabetes, heart disease, stroke, and depression), health behaviors (smoking, alcohol, physical activity, and body mass index), geriatric variables (activities of daily living, instrumental activities of daily living, hearing and/or vision impairment, and falls), and psychosocial variables (fatigue, sleep, and loneliness) (eTable 2 in Supplement 1 ). Reporting race and ethnicity in this study was mandated by the US National Institutes of Health, consistent with the Inclusion of Women, Minorities, and Children policy. Three variables (stress, discrimination, and apolipoprotein E genotype) were missing in more than 5% of the HMN cohort and thus were not included in propensity scores. Modal imputation was used for missing data in variables missing in 5% or fewer cases.

Statistical Analysis

We used propensity score–based methods to evaluate the association between HMN diagnosis and cognitive outcomes. The propensity score was estimated using a logistic regression model that used multiple variables to estimate HMN diagnosis. To match participants with HMN to those without cancer, we first calculated propensity scores for eligible HMN participants using data from the HRS assessment immediately preceding diagnosis (baseline assessment). Then we calculated propensity scores for each potential eligible control in each HRS wave. We matched HMN participants to those without cancer in the same wave using greedy nearest-neighbor matching using a 1:3 ratio and calipers equal to 0.2 of the SD of the logit of the propensity score. We evaluated the quality of matching by comparing standardized mean differences between groups for each covariate. 33

Time 0 for HMN participants was the date of diagnosis, which could be 0 to 730 days from their baseline assessment. A pseudo-time 0 was assigned for the control group based on their HMN match, such that pseudo-time 0 was the same number of days from the control’s matched baseline assessment. If the control died before pseudo-time 0, a second round of matching was performed to maximize the number of matches.

Model Selection

After trialing different ways of modeling cognitive trajectories, we compared 2 models with most face validity: a cubic spline with 4 knots (−1, −0.25, 0.25, and 1 year) and a piecewise linear spline with 2 knots (−1 and 1 year). The cubic and linear models had similar cognitive slopes, Akaike information criterion, and bayesian information criterion, with the linear model showing slightly better fit. The linear model also enabled direct comparison of slopes and CIs between cohorts. Thus, we chose the piecewise linear spline for our final analyses.

Sensitivity and Supplemental Analyses

There is potential bias due the competing risk of death if patients with cancer were more likely than those without to die or drop out, such that the fittest individuals who survived longer would be overrepresented in the postdiagnosis cognitive trajectory. Thus, we performed 2 sensitivity analyses. First, we applied inverse probability weighting (IPW) for mortality, which adjusts for the individual’s survival probability, such that cognitive data of those with a lower survival probability (eg, more similar to participants who died) were weighted more. 34 Second, we repeated the analyses while limiting the analytic cohort to those who completed at least 2 HRS assessments before and 2 assessments after time 0. This approach ensured we were comparing participants who survived for at least 2 to 4 years after time 0 and had nonmissing HRS assessments before and after diagnosis to establish a cognitive trajectory.

As supplemental analysis, we repeated the analyses with survey weighting so the study sample more accurately represented the population of community-dwelling US adults. We did not use the weighted analysis as the primary approach since the analytic dataset represented only a small portion of the full HRS population, but it is still important to compare weighted and unweighted results. 35 Statistical analyses were conducted using SAS version 9.4 (SAS Institute) and Stata version 18.0 (Stata Corp). Data were analyzed from April 2022 to April 2024.

Results

Participant Characteristics and Propensity Score-Matching

At baseline, there were 668 participants in the HMN cohort (mean [SD] age, 76.8 [7.6] years; 343 [51.3%] male; 72 [10.8%] Black, 33 [4.9%] Hispanic, and 585 [87.6%] White) and 1994 participants in the control cohort (mean [SD] age, 76.5 [7.3] years; 1020 [51.2%] male; 226 [11.3%] Black, 91 [4.6%] Hispanic, and 1726 [86.6%] White). The mean (SD) follow-up from first to last HRS assessment evaluated was 13.0 (6.8) years in the HMN cohort and 16.2 (6.7) years in the control cohort. To define the HMN cohort, we identified 1085 HRS respondents with linked Medicare data who had an HMN diagnosis recorded between 1998 to 2016. Of these, we excluded 83 for being 65 years or younger at diagnosis, 89 for having dementia at diagnosis, and 244 for lacking adequate assessments for analysis, resulting in 669 eligible individuals. After propensity score–matching, 658 had 3 matches and 10 had 2 matches, resulting in 668 individuals in the HMN cohort and 1994 participants in the control cohort ( Figure 1 ). The 244 individuals with an HMN diagnosis who were ineligible due to lacking adequate assessments were not significantly different from the HMN cohort in age, sex, race, or ethnicity but were more likely to have lower levels of education (χ 2 3 = 15.0; P = .002).

At baseline, 168 patients in the HMN cohort (25.1%) and 487 patients in the control cohort (24.4%) had cognitive impairment without dementia. Within the HMN cohort, the most common diagnoses were myeloproliferative neoplasm (127 patients [19.0%]), chronic leukemia (106 patients [15.9%]), and multiple myeloma (97 patients [14.5%]). Overall, 477 patients (71.4%) had diagnoses usually treated with chronic or intermittent therapy (eg, myelodysplastic syndrome, indolent lymphomas, or multiple myeloma), and 87 patients (13.0%) had aggressive diagnoses usually treated with curative intent therapy (eg, acute leukemias or aggressive lymphomas). Only 96 patients (14.4%) in the HMN group received chemotherapy. Baseline characteristics are described in Table 1 . Matched cohorts were well-balanced on propensity score distribution and baseline characteristics. Standardized mean differences for all covariates were less than 0.1 ( Figure 2 ).

Cognitive Trajectories of HMN and Control Cohorts

Cognitive trajectories in the propensity-matched cohorts are shown in Figure 3 A, and the corresponding Langa-Weir cognitive slopes are presented in Table 2 . At baseline before time 0, similar rates of cognitive decline as measured by Langa-Weir score were seen in the HMN (−0.17; 95% CI, −0.19 to −0.15) and control (−0.17; 95% CI, −0.18 to −0.16) ( P > .99) cohorts. In the 2 years peridiagnosis, the HMN group seemed to have a faster rate of cognitive decline (−0.51; 95% CI, −0.65 to –0.36) compared with the control group (−0.39; 95% CI, −0.46 to –0.32); however, this difference was not significant ( P = .18). At 1 year postdiagnosis and beyond, the rate of cognitive decline was slower in the HMN cohort (−0.18; 95% CI, −0.23 to –0.14) than the control cohort (−0.24; 95% CI, −0.26 to –0.23; P = .02). For reference, a slope of −0.24 means that cognitive performance decreases by about 1 point in 4 years, which is the equivalent of missing 1 word on a 10-word recall or missing 1 subtraction on serial sevens.

Sensitivity and Supplemental Analyses

In both sensitivity analyses accounting for the competing risk of death, there were no significant differences in rates of cognitive decline between the 2 cohorts during any time period ( Figure 3 B and 3 C; Table 2 ). Using IPW for mortality, rates of cognitive decline were similar in the HMN (−0.27; 95% CI, −0.34 to –0.19) and control cohorts (−0.30; 95% CI, −0.33 to –0.27). Limiting the analytic cohort to those with longer pretime and posttime 0 follow-up, rates of cognitive decline were also similar in the HMN (−0.22; 95% CI, −0.27 to –0.16) and control cohorts (−0.23; 95% CI, −0.26 to –0.20). Notably, with the latter approach, the HMN group had absolute Langa-Weir scores that were consistently higher than the control group ( Figure 3 C). Supplemental analysis using survey weights produced globally similar results (eTable 3 in Supplement 1 ).

Discussion

Contrary to our hypothesis, the rate of cognitive decline in older adults after HMN diagnosis was not faster compared with normal aging. At baseline before and in the 2 years surrounding time 0, the HMN and control cohorts had similar rates of cognitive decline. Posttime 0, the rate of cognitive decline appeared slower in the HMN cohort than the control cohort, but this difference was no longer statistically significant after accounting for the competing risk of death.

Our results are concordant with other longitudinal cognitive studies in patients with cancer, which found slower or similar rates of cognitive decline in cancer survivors compared with participants without cancer. To our knowledge, no prior longitudinal studies with prediagnosis cognitive trajectories have focused on HMN. Ospina-Romero et al 36 found that HRS participants aged 50 years or older with cancer had better memory function and slower memory decline after diagnosis compared with participants without cancer. To address survival bias, their sensitivity analysis in individuals with similar follow-up time yielded similar results. Wang et al 37 found that a history of cancer did not exacerbate cognitive functioning in adults aged 65 years or older in HRS, and adults aged 75 to 84 years had slower cognitive decline compared with participants without cancer. A meta-analysis 23 of longitudinal studies investigating cognitive performance in adults treated with chemotherapy found that patients’ performance improved from baseline to follow-up more than the improvement seen in the control group. In a population-based study, Anstey et al 38 , 39 found that those aged 60 years and older with cancer who did not receive chemotherapy had similar cognitive performance as participants without cancer.

The counterintuitive finding of slower cognitive decline in cancer survivors compared with participants without cancer may support the proposed inverse relationship between cancer and Alzheimer disease, where factors related to carcinogenesis may be neuroprotective. 40 , 41 , 42 , 43 , 44 , 45 Another possible explanation is posttraumatic growth among cancer survivors, which posits that cancer survivors may be incentivized to adopt lifestyle changes beneficial to cognitive functioning. 46 , 47 Alternatively, this paradoxical inverse relationship may be due to methodological biases, such as competing risks, survival bias, or handling of potential confounders. 48 , 49 , 50

In our study accounting for the competing risk of death, after 1 year or more postdiagnosis, the HMN group demonstrated similar rates of cognitive decline as matched participants without cancer. The postdiagnosis rate of cognitive decline in the HMN group was steeper with IPW (−0.27) than in the primary analysis (−0.18), consistent with the concern that those with a higher probability of mortality (therefore upweighted with IPW) exhibit faster cognitive decline that would not be reflected in the primary analysis. These results suggest that the competing risk of death does bias results toward better cognitive performance in survivors.

It is important to acknowledge that our HMN cohort is composed predominantly of indolent diseases, and only 14.4% received chemotherapy. Thus, these cognitive trajectories represent primarily indolent diagnoses which may not require therapy for years. Cognitive trajectories may look different if the cohort included mostly aggressive diagnoses or those treated with chemotherapy. One future direction of this work is to describe patterns of cognitive decline in different disease and treatment subgroups while accounting for potential confounders for chemotherapy receipt.

Strengths and Limitations

Strengths of this study include the focus on HMN, a large population-representative sample, longitudinal cognitive data that precede cancer diagnosis, use of Medicare claims data rather than self-report of cancer to minimize misclassification, and accounting for the competing risk of death. This study has several limitations. First, the trajectories included only Langa-Weir scores based on assessments completed by the participant. Participants who only had proxy report for cognition were not included in the trajectory analysis given the small proportion (<5% at baseline) and lack of a validated way to map proxy scores to Langa-Weir scores, although proxy scores were used to determine baseline dementia status. Second, we acknowledge limitations of the Langa-Weir, which only correctly identifies dementia in 52.18% of cases and does not cover all cognitive domains. However, it is useful for evaluating change over time and covers most domains implicated in CRCI (eg, learning and memory, attention, working memory, processing speed). Third, different HMN diagnoses exhibit heterogenous clinical courses and are managed differently, and we do not account for this heterogeneity due to the small resulting sample sizes. The CRCI literature suggests that both disease and treatment may impact cognition, and thus different disease or treatment subgroups may exhibit different cognitive trajectories, which is an important area for future research. Fourth, with propensity score-based matching, there is the possibility of unmeasured confounding. However, we matched on a very broad set of variables and our prediagnosis cognitive trajectories are essentially identical between the 2 groups, suggesting that they are well-matched on risk factors for cognition. Fifth, our cohorts included individuals who completed adequate assessments for analysis and thus represent a highly selected group, which may impact generalizability to the standard population. Additionally, our selection method of participants without cancer may miss individuals cured of cancer before 1998, which would not be captured by diagnosis code review during 1998 to 2016.

Conclusions

In conclusion, after accounting for the competing risk of death, the cognitive trajectories of older adults with HMN paralleled that of normal aging. The competing risk of death should be accounted for in longitudinal studies of cognition in older adults with cancer to avoid potential bias. Despite frequent concerns about CRCI, we did not find evidence that HMN diagnosis accelerates cognitive decline more than normal aging, which can be helpful for counseling patients on what to expect cognitively after diagnosis. Future directions include identifying predictors of clinically meaningful cognitive decline and exploring whether different diagnosis and treatment types will result in heterogeneous cognitive trajectories.

eTable 1. Definition of Cancer-Related Exposure and Covariates

eTable 2. Sociodemographic and Health-Related Variables Included in Propensity Score Model

eTable 3. Change in Langa-Weir Cognitive Summary Score in Hematologic Malignant Neoplasm and Propensity Score–Matched Noncancer Control Cohorts in Models With Survey Weighting

Data Sharing Statement

References

1. Pulte D, Jansen L, Brenner H. Changes in long term survival after diagnosis with common hematologic malignancies in the early 21st century. Blood Cancer J. 2020;10(5):56. doi: 10.1038/s41408-020-0323-4

2. Pulte D, Jansen L, Castro FA, Brenner H. Changes in the survival of older patients with hematologic malignancies in the early 21st century. Cancer. 2016;122(13):2031-2040. doi: 10.1002/cncr.30003

3. Surveillance Research Program, National Cancer Institute . SEER*Explorer: an interactive website for SEER cancer statistics. Accessed April 30, 2024. https://seer.cancer.gov/statistics-network/explorer/

4. Fried TR, Van Ness PH, Byers AL, Towle VR, O’Leary JR, Dubin JA. Changes in preferences for life-sustaining treatment among older persons with advanced illness. J Gen Intern Med. 2007;22(4):495-501. doi: 10.1007/s11606-007-0104-9

5. Fried TR, Bradley EH, Towle VR, Allore H. Understanding the treatment preferences of seriously ill patients. N Engl J Med. 2002;346(14):1061-1066. doi: 10.1056/NEJMsa012528

6. Kerstens C, Wildiers HPMW, Schroyen G, et al. A systematic review on the potential acceleration of neurocognitive aging in older cancer survivors. Cancers (Basel). 2023;15(4):1215. doi: 10.3390/cancers15041215

7. Williams AM, Zent CS, Janelsins MC. What is known and unknown about chemotherapy-related cognitive impairment in patients with haematological malignancies and areas of needed research. Br J Haematol. 2016;174(6):835-846. doi: 10.1111/bjh.14211

8. Lange M, Joly F, Vardy J, et al. Cancer-related cognitive impairment: an update on state of the art, detection, and management strategies in cancer survivors. Ann Oncol. 2019;30(12):1925-1940. doi: 10.1093/annonc/mdz410

9. Loh KP, Janelsins MC, Mohile SG, et al. Chemotherapy-related cognitive impairment in older patients with cancer. J Geriatr Oncol. 2016;7(4):270-280. doi: 10.1016/j.jgo.2016.04.008

10. Ahles TA, Root JC. Cognitive effects of cancer and cancer treatments. Annu Rev Clin Psychol. 2018;14:425-451. doi: 10.1146/annurev-clinpsy-050817-084903

11. Mandelblatt JS, Hurria A, McDonald BC, et al. ; Thinking and Living With Cancer Study . Cognitive effects of cancer and its treatments at the intersection of aging: what do we know; what do we need to know? Semin Oncol. 2013;40(6):709-725. doi: 10.1053/j.seminoncol.2013.09.006

12. Pergolotti M, Battisti NML, Padgett L, et al. Embracing the complexity: older adults with cancer-related cognitive decline-a young international society of geriatric oncology position paper. J Geriatr Oncol. 2020;11(2):237-243. doi: 10.1016/j.jgo.2019.09.002

13. Janelsins MC, Kesler SR, Ahles TA, Morrow GR. Prevalence, mechanisms, and management of cancer-related cognitive impairment. Int Rev Psychiatry. 2014;26(1):102-113. doi: 10.3109/09540261.2013.864260

14. Littlewood T, Mandelli F. The effects of anemia in hematologic malignancies: more than a symptom. Semin Oncol. 2002;29(3)(suppl 8):40-44. doi: 10.1016/S0093-7754(02)70173-5

15. Straus DJ, Testa MA, Sarokhan BJ, et al. Quality-of-life and health benefits of early treatment of mild anemia: a randomized trial of epoetin alfa in patients receiving chemotherapy for hematologic malignancies. Cancer. 2006;107(8):1909-1917. doi: 10.1002/cncr.22221

16. Mozos I, Borzak G, Caraba A, Mihaescu R. Arterial stiffness in hematologic malignancies. Onco Targets Ther. 2017;10:1381-1388. doi: 10.2147/OTT.S126852

17. Williams AM, van Wijngaarden E, Seplaki CL, et al. Cognitive function in patients with chronic lymphocytic leukemia: a cross-sectional study examining effects of disease and treatment. Leuk Lymphoma. 2020;61(7):1627-1635. doi: 10.1080/10428194.2020.1728748

18. Juríčková V, Fayette D, Jonáš J, Fajnerová I, Kozák T, Horáček J. Pretreatment cancer-related cognitive impairment in hodgkin lymphoma patients. Curr Oncol. 2023;30(10):9028-9038. doi: 10.3390/curroncol30100652

19. Fayette D, Juríčková V, Kozák T, et al. Cognitive impairment associated with Hodgkin’s lymphoma and chemotherapy. Neurosci Lett. 2023;797:137082. doi: 10.1016/j.neulet.2023.137082

20. Janelsins MC, Mohamed M, Peppone LJ, et al. Longitudinal changes in cognitive function in a nationwide cohort study of patients with lymphoma treated with chemotherapy. J Natl Cancer Inst. 2022;114(1):47-59. doi: 10.1093/jnci/djab133

21. Ahles TA, Root JC, Ryan EL. Cancer- and cancer treatment-associated cognitive change: an update on the state of the science. J Clin Oncol. 2012;30(30):3675-3686. doi: 10.1200/JCO.2012.43.0116

22. Bernstein LJ, McCreath GA, Komeylian Z, Rich JB. Cognitive impairment in breast cancer survivors treated with chemotherapy depends on control group type and cognitive domains assessed: a multilevel meta-analysis. Neurosci Biobehav Rev. 2017;83:417-428. doi: 10.1016/j.neubiorev.2017.10.028

23. Lindner OC, Phillips B, McCabe MG, et al. A meta-analysis of cognitive impairment following adult cancer chemotherapy. Neuropsychology. 2014;28(5):726-740. doi: 10.1037/neu0000064

24. Sonnega A, Faul JD, Ofstedal MB, Langa KM, Phillips JW, Weir DR. Cohort profile: the health and retirement study (HRS). Int J Epidemiol. 2014;43(2):576-585. doi: 10.1093/ije/dyu067

25. Medicode . International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). Medicode; 1997.

26. World Health Organization . International Statistical Classification of Diseases, Tenth Revision (ICD-10). World Health Organization; 1992.

27. Crimmins EM, Kim JK, Langa KM, Weir DR. Assessment of cognition using surveys and neuropsychological assessment: the Health and Retirement Study and the Aging, Demographics, and Memory Study. J Gerontol B Psychol Sci Soc Sci. 2011;66 Suppl 1(Suppl 1):i162-171. doi: 10.1093/geronb/gbr048

28. Langa KM, Kabeto M, Weir D. Report on race and cognitive impairment using HRS in 2010 Alzheimer’s disease facts and figures. Alzheimer’s Dement. 2010;6(1):158-194. doi: 10.1016/j.jalz.2010.01.009

29. Langa KM, Plassman BL, Wallace RB, et al. The aging, demographics, and memory study: study design and methods. Neuroepidemiology. 2005;25(4):181-191. doi: 10.1159/000087448

30. RAND . RAND HRS Longitudinal File. 2020. Accessed July 25, 2024. https://hrsdata.isr.umich.edu/data-products/rand-hrs-longitudinal-file-2020

31. Health and Retirement Study . Harmonized Health and Retirement Study documentation version C public use dataset. 2022. Accessed July 30, 2024. https://hrs.isr.umich.edu/data-products/access-to-public-data/conditions-of-use

32. Health and Retirement Study . RAND Health and Retirement Study longitudinal file V1 public use dataset. 2020. Accessed July 30, 2024. https://hrs.isr.umich.edu/data-products/access-to-public-data/conditions-of-use

33. Austin PC. A comparison of 12 algorithms for matching on the propensity score. Stat Med. 2014;33(6):1057-1069. doi: 10.1002/sim.6004

34. Seaman SR, White IR. Review of inverse probability weighting for dealing with missing data. Stat Methods Med Res. 2013;22(3):278-295. doi: 10.1177/0962280210395740

35. Si Y, Lee S, Heeringa SG. Population weighting in statistical analysis. JAMA Intern Med. 2024;184(1):98-99. doi: 10.1001/jamainternmed.2023.6300

36. Ospina-Romero M, Abdiwahab E, Kobayashi L, et al. Rate of memory change before and after cancer diagnosis. JAMA Netw Open. 2019;2(6):e196160. doi: 10.1001/jamanetworkopen.2019.6160

37. Wang Y, Haaksma ML, Ramakers IHGB, et al. Cognitive and functional progression of dementia in two longitudinal studies. Int J Geriatr Psychiatry. 2019;34(11):1623-1632. doi: 10.1002/gps.5175

38. Anstey KJ, Sargent-Cox K, Cherbuin N, Sachdev PS. Self-reported history of chemotherapy and cognitive decline in adults aged 60 and older: the PATH through life project. J Gerontol A Biol Sci Med Sci. 2015;70(6):729-735. doi: 10.1093/gerona/glt195

39. Kim S, Cherbuin N, Anstey KJ. Cancer and cognitive function: the PATH through life project. J Gerontol A Biol Sci Med Sci. 2017;72(9):1226-1232. doi: 10.1093/gerona/glw254

40. Zhang DD, Ou YN, Yang L, et al. Investigating the association between cancer and dementia risk: a longitudinal cohort study. Alzheimers Res Ther. 2022;14(1):146. doi: 10.1186/s13195-022-01090-9

41. Ma LL, Yu JT, Wang HF, et al. Association between cancer and Alzheimer’s disease: systematic review and meta-analysis. J Alzheimers Dis. 2014;42(2):565-573. doi: 10.3233/JAD-140168

42. Driver JA, Beiser A, Au R, et al. Inverse association between cancer and Alzheimer’s disease: results from the Framingham Heart Study. BMJ. 2012;344:e1442. doi: 10.1136/bmj.e1442

43. Driver JA. Inverse association between cancer and neurodegenerative disease: review of the epidemiologic and biological evidence. Biogerontology. 2014;15(6):547-557. doi: 10.1007/s10522-014-9523-2

44. Shafi O. Inverse relationship between Alzheimer’s disease and cancer, and other factors contributing to Alzheimer’s disease: a systematic review. BMC Neurol. 2016;16(1):236. doi: 10.1186/s12883-016-0765-2

45. Bowles EJA, Walker RL, Anderson ML, Dublin S, Crane PK, Larson EB. Risk of Alzheimer’s disease or dementia following a cancer diagnosis. PLoS One. 2017;12(6):e0179857. doi: 10.1371/journal.pone.0179857

46. Tedeschi RG, Calhoun LG. Target article: “posttraumatic growth: conceptual foundations and empirical evidence”. Psychol Inq. 2004;15(1):1-18. doi: 10.1207/s15327965pli1501_01

47. Marioni RE, Proust-Lima C, Amieva H, et al. Cognitive lifestyle jointly predicts longitudinal cognitive decline and mortality risk. Eur J Epidemiol. 2014;29(3):211-219. doi: 10.1007/s10654-014-9881-8

48. Ospina-Romero M, Glymour MM, Hayes-Larson E, et al. Association between alzheimer disease and cancer with evaluation of study biases: a systematic review and meta-analysis. JAMA Netw Open. 2020;3(11):e2025515. doi: 10.1001/jamanetworkopen.2020.25515

49. Hayes-Larson E, Ackley SF, Zimmerman SC, et al. The competing risk of death and selective survival cannot fully explain the inverse cancer-dementia association. Alzheimers Dement. 2020;16(12):1696-1703. doi: 10.1002/alz.12168

50. Hayes-Larson E, Shaw C, Ackley SF, et al. The Role of dementia diagnostic delay in the inverse cancer-dementia association. J Gerontol A Biol Sci Med Sci. 2022;77(6):1254-1260. doi: 10.1093/gerona/glab341

Associated Data

Supplementary Materials

eTable 1. Definition of Cancer-Related Exposure and Covariates

eTable 2. Sociodemographic and Health-Related Variables Included in Propensity Score Model

eTable 3. Change in Langa-Weir Cognitive Summary Score in Hematologic Malignant Neoplasm and Propensity Score–Matched Noncancer Control Cohorts in Models With Survey Weighting

Data Sharing Statement