Abstract
INTRODUCTION
Cancer survivors are less likely than comparably aged individuals without a cancer history to develop Alzheimer's disease and related dementias (ADRD).
METHODS
In the UK Biobank, we investigated associations between cancer history and five structural magnetic resonance imaging (MRI) markers for ADRD risk, using linear mixed‐effects models to assess differences in mean values and quantile regression to examine whether associations varied across the distribution of MRI markers.
RESULTS
Cancer history was associated with smaller mean hippocampal volume (b = ‐19 mm3, 95% CI = ‐36, ‐1) and lower mean cortical thickness in the Alzheimer's disease signature region (b = ‐0.004 mm, 95% CI = ‐0.007, ‐0.000). Quantile regressions indicated individuals most vulnerable to ADRD were more affected by cancer history.
DISCUSSION
Some brain MRI markers associated with ADRD risk were elevated in adults with a history of cancer. The magnitude of the adverse associations varied across quantiles of neuroimaging markers, and the pattern suggests possible harmful associations for individuals already at high ADRD risk.
Highlights
We found no evidence of an inverse association between cancer history and ADRD‐related neurodegeneration.
Cancer history was associated with smaller mean hippocampal volume and lower mean cortical thickness in the Alzheimer's disease signature region.
Quantile regressions indicated individuals most vulnerable to ADRD were more affected by cancer history.
Keywords: Alzheimer's disease and related dementias, cancer, neuroimaging, quantile regression
1. BACKGROUND
Numerous epidemiologic studies have identified an inverse association between both prevalent and incident diagnosis of cancer of any type and subsequent risk of incident Alzheimer's disease and related dementias (ADRD). 1 , 2 , 3 , 4 Some studies have also found that individuals with a history of cancer had a lower neuropathologic burden of Alzheimer's disease (AD) compared to age‐matched decedents without a history of cancer. 5 , 6 , 7 Possible explanations for the observed inverse association include selective survival of healthier cancer survivors who are at lower risk of ADRD and shared biological or genetic mechanisms that elevate cancer risk but reduce ADRD risk. 8 , 9 , 10 , 11 , 12 The inverse association between cancer and ADRD has sparked substantial interest because biological explanations if validated, could offer important insights into the underlying biology of ADRD. While existing research has focused on incident ADRD or cognitive function as the outcome of interest, there is a lack of evidence on the relationship between cancer history and biomarkers of ADRD risk—although there is research on treatment‐induced structural and functional brain changes among breast cancer survivors, 13 , 14 , 15 only two prior studies have evaluated the association between cancer diagnosis and MRI markers of structural brain aging. 16 , 17
Understanding this relationship could help to disentangle potential biases (e.g., diagnostic bias) from causal mechanisms in the cancer‐ADRD relationship. Measuring ADRD onset is difficult due to the slow etiologic development of the disease, phenotypic heterogeneity, and frequent missed or delayed diagnoses. 18 , 19 , 20 Brain changes detectable from neuroimaging likely precede diagnosed ADRD by decades. 21 Evaluating whether cancer history is associated with neuroimaging markers of ADRD risk avoids the potential for diagnostic bias of ADRD, which occurs when clinicians overlook ADRD symptoms due to cancer diagnoses and treatments or when patients with a cancer history have more frequent contact with clinicians, increasing the chance of ADRD diagnosis.
Previous investigations have revealed site‐specific associations between cancer and ADRD, 8 prompting us to examine various cancer types. Nevertheless, the sample sizes in two prior studies were too small (n = 2,043 and 1609) to provide conclusive estimates for individual cancer types. 16 , 17 Furthermore, prior studies use linear regression models, which estimate the effects on the mean values of the neuroimaging measures, but do not evaluate the effects on the overall distribution of the biomarkers. Focusing on the mean, without consideration of high or low quantiles, precludes more comprehensive insights into heterogeneous exposure associations across the distribution of outcomes. 22 For example, any beneficial or harmful impact of cancer on brain volumes may be largest for people with otherwise low (or high) volumetric values. The variability of the effects of cancer history across the distribution of each MRI marker, which serves as a proxy for ADRD risk, remains unclear.
In this manuscript, we used the UK Biobank neuroimaging sample to test the hypothesis that people with a history of cancer have brain MRI characteristics associated with a lower risk of incident ADRD compared to otherwise similar individuals without a history of cancer. We examined both all cancer sites and 10 common individual cancer types. We also examined the heterogeneous effect of cancer across quantiles of each MRI marker to test the hypothesis that individuals most susceptible to ADRD are affected the most by their cancer history.
2. METHODS
2.1. Study population
The UK Biobank is a prospective volunteer cohort of 502,490 adults aged 40‐69 years who attended 1 of 22 assessment centers across the United Kingdom from 2006 to 2010. At the baseline visit, participants completed physical, physiological, and medical assessments. In 2014, the UK Biobank invited participants for brain MRIs at four clinics using identical protocols. At the time of writing, MRI data were available for 43,102 participants, among whom 5514 had data from a repeated imaging visit. Ethics approval was obtained from the National Health Service National Research Ethics Service (16/NW/0274), and all UK Biobank participants signed informed consent before information collection.
RESEARCH IN CONTEXT
Systematic review: The authors reviewed the relevant literature in PubMed. Evidence from relatively small observational studies did not support a linkage between cancer and neuroimaging biomarkers that would explain an inverse association of cancer history with subsequent dementia.
Interpretation: This study expands prior research by examining the associations between cancer history and five structural magnetic resonance imaging (MRI) markers for Alzheimer's disease and related dementias (ADRD) risk in a large cohort. Overall, our findings did not support an inverse link between cancer diagnosis and ADRD‐related neurodegeneration. Rather, we find a suggestion of adverse effects of cancer history on some neuroimaging markers, including harmful associations for individuals already at high ADRD risk.
Future directions: Further research should assess longitudinal changes in cognition and neuroimaging markers before and after cancer treatments.
2.2. Brain imaging data
Neuroimaging variables were selected a priori based on previous studies showing their associations with cognitive decline or ADRD pathologies. They included total gray matter volume, 23 , 24 total brain volume, 25 hippocampal volume, 26 , 27 white matter hyperintensity volume, 28 , 29 and mean cortical thickness in the AD signature region, comprising six regions of interest: entorhinal, inferior temporal, middle temporal, inferior parietal, fusiform, and precuneus. 30 All MRIs were carried out using similar scanners (Siemens Skyra 3T scanner with a standard 32‐channel head coil). Full details on image acquisition, processing, and quality control are available from the UK Biobank Brain Imaging Documentation and protocol publications. 31 In brief, the T1‐weighted anatomic images were acquired using three‐dimensional magnetization prepared for rapid gradient‐echo (3D MPRAGE) at a resolution of 1 × 1 × 1 mm. Total white matter hyperintensity volumes were derived based on T1 and T2 fluid‐attenuated inversion recovery (FLAIR) using the Brain Intensity Abnormality Classification Algorithm (BIANCA). 32 Regional estimates of cortical thickness and surface area were processed using FreeSurfer v.5.3 based on the Desikan–Killiany atlas parcellation. 33 Hippocampal volume was estimated by summing left and right hippocampal volumes. All volumetric measures were corrected for skull size using a residual method 34 that calculated adjusted brain volumes from residuals of a linear regression between raw volumes and intracranial volume. The mean cortical thickness in the AD signature region was calculated by the surface area‐weighted average of cortical thicknesses across six regions of interest.
2.3. Ascertainment of cancer cases
Cancer diagnoses were identified via linkage to hospital admission, cancer registries, and self‐reported medical conditions using International Classification of Diseases ICD‐10 and ICD‐9 codes (Table S1). We included all cancer types except non‐melanoma skin cancer (NMSC) in our primary analysis. For all participants, an indicator for cancer diagnosis denoted at least one diagnosis any time prior to the first imaging visit. A time‐updated indicator of cancer diagnosis was available for 5514 participants who had a repeat imaging visit. In secondary analyses, we repeated the variable construction for breast cancer 2 , 35 and prostate cancer, 2 , 36 which are the two most common cancers in the UK Biobank and have been linked to ADRD risk in previous studies. In addition, evidence of gray matter volume loss and white matter microstructural disruption has been reported among breast cancer survivors treated with chemotherapy. 13 , 14 To account for the potential impact of chemotherapy‐induced brain changes, we also evaluated NMSC, for which chemotherapy is not a common treatment. 37 We additionally evaluated the remaining seven cancer sites from the top 10 most prevalent cancer sites in our sample, including malignant melanoma, uterine cancer, lymphoma, cervical cancer, testicular cancer, ovarian cancer, and bladder cancer.
2.4. Ascertainment of ADRD cases
ADRD cases were identified from linkage to hospital admission, primary care, and death records using a comprehensive list of ICD‐10 and ICD‐9 codes described elsewhere. 9 In brief, we included AD, vascular dementia, frontotemporal dementia, Lewy body dementia, alcohol‐related dementia, and Creutzfeldt‐Jakob disease. We defined ADRD onset as the first date of ADRD diagnosis.
2.5. Assessment of covariates
We controlled for covariates that plausibly influence both cancer and ADRD risk. We conceptualized two sets of covariates based on their temporality relative to cancer diagnosis (Figure S1). Our “base model” included covariates that could not be affected by cancer diagnosis: age (linear and quadratic terms 38 , 39 , 40 ), sex (female, male), race (White, Black, Asian, Other), and binary apolipoprotein E (APOE)‐ε4 carrier status. In our fully adjusted models, we additionally controlled for covariates that could both affect and be affected by cancer diagnosis: educational attainment (professional/university degree, secondary, vocational, other qualifications), Townsend deprivation index, body mass index (BMI), ever smoking, ever alcohol use, high physical activity (≥75 min/week of vigorous activity or ≥150 min/week of moderate activity 41 ), and assessment center (Cheadle, Reading, Newcastle, Bristol). Most covariates were assessed at study enrollment from 2006 to 2010. Information on BMI, smoking, alcohol use, and physical activity was collected during the imaging visits. Townsend deprivation index is a composite score measuring area‐level socioeconomic status based on employment, home ownership, car ownership, and household overcrowdedness. 42 During each visit, trained staff measured height and weight, and BMI was derived by dividing weight (kg) by the square of height (m2). Participants reported details on ever smoking, ever alcohol use, and physical activity via touchscreen questionnaires.
2.6. Statistical analyses
We summarized characteristics of the study sample stratified by cancer status. To confirm the relevance of the selected MRI measures for ADRD, we used Cox proportional hazards regression models to estimate adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for associations between each MRI marker and incident ADRD. For these analyses, we treated the imaging visit as the baseline and only included participants with no ADRD diagnosis at that time. Participants were followed up to date of ADRD diagnosis, death, or censoring on September 30, 2021 (the latest date of ADRD diagnosis in the imaging cohort), whichever came first. We adjusted for the same covariate sets as in the primary analyses. We scaled brain volume and cortical thickness measures by dividing them by the sample standard deviation and estimated separate models for each individual measure.
To evaluate whether mean values of the neuroimaging outcomes differed by cancer status, we used linear mixed‐effects models with individual‐level random intercepts to account for repeated within‐person MRI measures. To examine whether the effects of a prior cancer diagnosis on neuroimaging outcomes varied across the distribution of ADRD risk, we specified quantile regression models at the 10th, 25th, 50th, 75th, and 90th percentiles of each neuroimaging outcome. We used cluster bootstrapping to account for repeated within‐person measures. Quantile regression allows estimation of the relationship between an exposure and an outcome across cut‐points of the outcome distribution, such as quartiles or deciles. 43 The quantile regression coefficients quantify how much the location each specified quantile of the neuroimaging outcome distribution differs between those with and without a history of cancer. If the coefficients are similar in each of the five quantile regression models (10th, 25th, 50th, 75th, and 90th percentiles), we would expect the associations of cancer history with each neuroimaging outcome are not differential across the entire distribution of the outcome and vice versa. We used the Wald test to check for heterogeneous associations of cancer with neuroimaging outcomes across quantiles. In secondary analyses, we analyzed the associations of breast cancer, prostate cancer, NMSC, and seven common cancer sites with mean neuroimaging outcomes. For each individual cancer type, we excluded participants with any other cancer diagnosis. To offer context about the magnitude of effects, we fitted linear mixed‐effects models with random intercepts and linear age. This allowed us to report the predicted differences in biomarkers associated with a 1‐year increment in age within this sample.
We conducted three sensitivity analyses. First, to minimize potential short‐term effects of chemotherapy, we excluded participants with a cancer diagnosis within 5 years before the MRI scan. Second, to model potential time‐varying associations of cancer with neuroimaging outcomes, we split the exposure variable into four categories based on the time between the most recent cancer diagnosis and the MRI scan: 0–1 year, >1–5 years, >5–10 years, and >10 years. Third, we applied inverse probability weighting (IPW) 44 to address potential selection bias 45 due to selection into the imaging cohort from the parent UK Biobank study. We used a logistic regression model to calculate the inverse probability of selection into the imaging cohort and then assigned stabilized weights to each participant and performed inverse probability weighted analyses to address informative selection. Model specifications and weights assessment are presented in Supplemental Data 1.
All statistical analyses were performed using R version 4.0.5.
3. RESULTS
3.1. Participant characteristics
The analytical sample included 43,102 participants (Table 1). The mean age at the first imaging visit was 64.5 years (SD = 7.7 years). Before the imaging visit, 5514 (12.8%) participants had a recorded cancer diagnosis. Participants with a history of cancer were older, more likely to be female and White, and averaged less physical activity.
TABLE 1.
Demographic, clinical, and imaging characteristics of participants in the analytical sample at the first imaging visit, stratified by cancer history
| Characteristic |
No cancer diagnosis (n = 37,588) |
Cancer diagnosis (n = 5514) |
|---|---|---|
| Age (year), median (interquartile) | 64.5 (58.1‐70.1) | 67.7 (61.4‐72.4) |
| Sex | ||
| Female | 19,337 (51.4) | 3371 (61.1) |
| Male | 18,251 (48.6) | 2143 (38.9) |
| Race | ||
| White | 36,302 (96.6) | 5397 (97.9) |
| Black | 321 (0.9) | 30 (0.5) |
| Asian | 602 (1.6) | 44 (0.8) |
| Other | 353 (0.9) | 43 (0.8) |
| APOE‐ε4 carrier | ||
| Yes | 10,187 (72.2) | 1462 (72.8) |
| No | 26,416 (27.8) | 3905 (27.2) |
| Education | ||
| Higher | 19,496 (51.9) | 2769 (50.2) |
| Secondary | 13,622 (36.2) | 2029 (36.8) |
| Vocational | 2035 (5.6) | 285 (5.2) |
| Other | 2435 (6.5) | 431 (7.8) |
| Townsend deprivation index b | ||
| 1 (least deprived) | 8959 (23.9) | 1330 (24.2) |
| 2‐4 | 23,390 (62.3) | 3414 (62.0) |
| 5 (most deprived) | 5205 (13.9) | 762 (13.8) |
| BMI (kg/m2), mean (SD) | 26.5 (4.4) | 26.4 (4.4) |
| Ever smoked | ||
| Yes | 13,697 (36.8) | 2232 (41.0) |
| No | 23,549 (63.2) | 3207 (59.0) |
| Ever used alcohol | ||
| Yes | 36,123 (96.8) | 5292 (96.8) |
| No | 1213 (3.2) | 173 (3.2) |
| High physical activity | ||
| Yes | 31,743 (85.4) | 4570 (84.2) |
| No | 5406 (14.6) | 857 (15.8) |
| GM (mm3), mean (SD) | 615,546 (56,068) | 605,119 (53,262) |
| TBV (mm3), mean (SD) | 1,161,750 (111,880) | 1,143,852 (106,584) |
| HV (mm3), mean (SD) | 8079 (808) | 7919 (783) |
| WMH (mm3), mean (SD) | 4968 (6,498) | 6033 (7,761) |
| CTAD (mm), mean (SD) | 2.82 (0.12) | 2.81 (0.12) |
Note: Values are mean (SD), count (%), or median (interquartile).
Abbreviations: APOE, apolipoprotein E; BMI, body mass index; CTAD, cortical thickness in the AD signature region; GM, gray matter; HV, hippocampal volume; SD, standard deviation; TBV, total brain volume; WMH, white matter hyperintensity.
Percentages may not sum to 100 because of rounding.
Index quintiles, combining social class, employment, car availability, and housing.
3.2. Associations of brain MRI variables with incident dementia
All MRI variables were significantly associated with ADRD incidence (Table 2). In the fully adjusted models, each standard deviation increase in gray matter volume (SD = 31,251 mm3), total brain volume (SD = 47,163 mm3), hippocampal volume (SD = 678 mm3), and mean cortical thickness in the AD signature region (SD = 0.12 mm) was associated with 64% (HR = 0.36, 95% CI = 0.29–0.44), 50% (HR = 0.50, 95% CI = 0.42–0.60), 56% (HR = 0.44, 95% CI = 0.36–0.54), and 48% (HR = 0.52, 95% CI = 0.42–0.65) decrease in the hazard of ADRD, respectively. For each SD increase in white matter hyperintensity volume (SD = 6616 mm3), there was a 27% increase in the hazard of ADRD (HR = 1.27, 95% CI = 1.11–1.45).
TABLE 2.
Association of neuroimaging outcomes with incident ADRD diagnosis: HRs from Cox proportional hazards regression models
| Demographics‐adjusted base model c | Fully adjusted model d | ||||
|---|---|---|---|---|---|
| Exposure a | N b | HR (95% CI) | p‐Value | HR (95% CI) | p‐Value |
| GM | 41,178 | 0.34 (0.28‐0.41) | <0.001 | 0.36 (0.29‐0.44) | <0.001 |
| TBV | 41,178 | 0.49 (0.41‐0.58) | <0.001 | 0.50 (0.42‐0.60) | <0.001 |
| HV | 41,939 | 0.41 (0.35‐0.49) | <0.001 | 0.44 (0.36‐0.54) | <0.001 |
| WMH | 39,870 | 1.26 (1.11‐1.42) | <0.001 | 1.27 (1.11‐1.45) | <0.001 |
| CTAD | 41,939 | 0.51 (0.42‐0.63) | <0.001 | 0.52 (0.42‐0.65) | <0.001 |
Abbreviations: ADRD, Alzheimer's disease and related dementias; CI, confidence interval; CTAD, cortical thickness in the AD signature region; GM, gray matter; HR, hazard ratio; HV, hippocampal volume; TBV, total brain volume; WMH, white matter hyperintensity.
Each exposure was scaled by dividing values by the sample standard deviation. All volumetric measures are in mm3 and cortical thickness is in mm.
Number of participants included in the analyses per outcome.
Adjusted for age, sex, race, and APOE‐ε4 carrier status.
Further adjusted for education and Townsend deprivation index, body mass index, ever smoking, ever using alcohol, physical activity, and assessment center.
3.3. Associations of cancer history with neuroimaging outcomes
Compared with participants without a history of cancer, participants with a history of cancer showed smaller hippocampal volume (β = −19 mm3, 95% CI = −36 to −1) and lower cortical thickness in the AD signature region (β = −0.004 mm, 95% CI = −0.007 to −0.000) in fully adjusted models (Table 3). There were no differences in gray matter volume, total brain volume, or white matter hyperintensity volumes between participants with and without a cancer history.
TABLE 3.
Covariate‐adjusted differences in mean values of neuroimaging outcomes comparing participants with versus without cancer history from linear mixed effects regression models
| Demographics‐adjusted base model c | Fully adjusted model d | ||||||
|---|---|---|---|---|---|---|---|
| Outcome a | N b | Estimate | 95% CI | p‐Value | Estimate | 95% CI | p‐Value |
| GM | 41,378 | −452 | −1146 to 241 | 0.201 | −350 | −1056 to 355 | 0.330 |
| TBV | 41,378 | −293 | −1401 to 815 | 0.604 | −376 | −1510 to 758 | 0.515 |
| HV | 42,036 | −23 | −40 to ‐6 | 0.008 | −19 | −36 to ‐1 | 0.037 |
| WMH | 40,119 | 216 | 39 to 392 | 0.017 | 178 | −1 to 357 | 0.051 |
| CTAD | 42,036 | −0.004 | −0.007 to −0.001 | 0.007 | −0.004 | −0.007 to −0.000 | 0.026 |
Abbreviations: CI, confidence interval; CTAD, cortical thickness in the AD signature region; GM, gray matter; HV, hippocampal volume; TBV, total brain volume; WMH, white matter hyperintensity.
All volumetric measures are in mm3 and cortical thickness is in mm.
Number of participants included in each analysis. Mixed effects models with random intercepts for individuals were used to account for a small number of repeated MRIs.
Adjusted for age, sex, race, and APOE‐ε4 carrier status.
Further adjusted for education and Townsend deprivation index, body mass index, ever smoking, ever using alcohol, physical activity, and assessment center.
The distributions of each neuroimaging outcome, stratified by cancer history, are shown in Figure 1. The quantile regression results estimating associations of cancer history with the 10th, 25th, 50th, 75th, and 90th percentiles of each neuroimaging outcome are shown in Figure 2 and Table S2. No significant differences in any quantiles of total gray matter volume and total brain volume were observed. The association between cancer history and hippocampal volume did not show clear patterns across quantiles, but cancer history was associated with a significantly smaller 75th percentile of the distribution of hippocampal volume (β 75 = −26 mm3, 95% CI = −54 to −1). This point estimate was similar to the estimated association with the 10th percentile (β 10 = −22 mm3, 95% CI = −52 to 13). The magnitude of the associations between cancer history and white matter hyperintensity volume were largest for the higher values of white matter hyperintensity; for example, cancer history was associated with a 250 mm3 larger 75th percentile (95% CI = 71 to 446) and a 552 mm3 larger 90th percentile (95% CI = 250 to 1002). Comparing participants with versus without cancer history, the adjusted differences in 10th, 25th, 50th, 75th, and 90th percentiles of the distributions of white matter hyperintensity volumes were significantly different (p‐value < 0.001). For cortical thickness in the AD signature region, associations were mainly observed in the lowest and median percentiles (β 10 = −0.006 mm, 95% CI = −0.011 to −0.000; β 50 = −0.005 mm, 95% CI = −0.009 to −0.001), and the associations across percentiles were significantly different (p‐value = 0.003).
FIGURE 1.
The distributions of each neuroimaging outcome, stratified by cancer history
FIGURE 2.
Covariate‐adjusted differences in the 10th, 25th, 50th, 75th, and 90th percentiles of the distributions of each neuroimaging outcome, comparing participants with versus without cancer history in quantile regression models. Demographics‐adjusted base model was adjusted for age, sex, race, and apolipoprotein E (APOE)‐ε4 carrier status. Fully adjusted model was further adjusted for education and Townsend deprivation index, body mass index (BMI), ever smoking, ever using alcohol, physical activity, and assessment center
When we investigated individual common cancer types, we found no significant mean differences in any neuroimaging outcome between participants with an NMSC history and participants without any cancer history (Table S3). Participants with a history of breast cancer compared to participants without a cancer history showed a significantly lower total brain volume (β = −3450 mm3, 95% CI = −5908 to −993), higher white matter hyperintensity volume (β = 512 mm3, 95% CI = 132–891), and lower cortical thickness in the AD signature region (β = −0.008 mm, 95% CI = −0.014 to −0.001). Prostate cancer was significantly associated with a higher total brain volume (β = 2832 mm3, 95% CI = 114 to 5551) but no other MRI markers. Significant associations were also observed between lymphoma and a lower total brain volume (β = −7033 mm3, 95% CI = −13,358 to −709) and a lower hippocampal volume (β = −149 mm3, 95% CI = −247 to −52).
3.4. Sensitivity analyses
After we excluded participants with a cancer history within 5 years before the MRI scan, no significant differences were observed across neuroimaging outcomes (Table S4). We did not observe clear patterns when we split the exposure based on the time between the most recent cancer diagnosis and the MRI scan, although most significant differences were detected among participants with a cancer diagnosis within 1–5 years prior to the MRI scan (Table S5: β = −1516 mm3, 95% CI = −2797 to −236 for gray matter volume; β = 391 mm3, 95% CI = 51–731 for white matter hyperintensity volumes; and β = −0.012 mm, 95% CI = −0.018 to −0.006 for cortical thickness in the AD signature region). In addition, a cancer diagnosis within 5–10 years before the MRI scan was associated with a lower hippocampal volume (β = −35 mm3, 95% CI = −68 to −1). The inverse probability weighted analyses to estimate results if the neuroimaging sample had the same characteristics as the full UK Biobank study population did not substantially alter results (Table S6).
4. DISCUSSION
In a large sample of UK adults, we found that adults with a history of cancer had some brain MRI markers associated with higher ADRD risk. Individuals with a history of cancer averaged slightly lower hippocampal volume and lower cortical thickness in the AD signature region than those without a cancer history, although the sensitivity analysis suggested these findings were driven by those with a cancer diagnosis within 5 years of the MRI. Cancer history was associated with higher variability across quantiles of white matter hyperintensity volume, that is, there were larger adverse associations with high (high‐risk) quantiles of white matter hyperintensity volume and null associations with low quantiles of white matter hyperintensity volume. Similar patterns were seen for cortical thickness, such that the adverse association with cancer history was largest at low (high‐risk) quantiles rather than high quantiles of cortical thickness. However, the association with cancer history at 50th quantile of cortical thickness was comparable with that at low quantiles. Survivors of breast cancer, an invasive cancer type, had significantly lower total brain volume and higher white matter hyperintensity volume compared to participants without a cancer history. Prostate cancer was associated with a slightly larger total brain volume. Lymphoma was associated with lower total brain volume and hippocampal volume. This association was not evident in other individual cancer sites.
Our results were consistent with the two previous studies describing the link between cancers at any site and structural brain aging. 16 , 17 In the Framingham Heart Study (n = 2043), Gupta et al. found that, compared with those without a cancer history, cancer survivors did not have a significant difference in total cerebral brain volume, temporal brain volume, temporal horn volume, or white matter hyperintensity volumes. 16 Nudelman et al. conducted a voxel‐based morphometric analysis of cerebral gray matter density (GMD) in the Alzheimer's Disease Neuroimaging Initiative cohort (n = 1609) and did not find any region with increased GMD among cancer survivors. 17 Instead, they found that a cancer history was associated with lower GMD in the right superior frontal gyrus. Although this area has not been linked to AD pathogenesis or diagnosis, 46 it has been associated with cancer treatments. 47 , 48 Our results expand on these findings in a sample over 10 times larger than the combined previous samples, with a wider range of neuroimaging measures, and evaluated whether effects differed across the distribution of neuroimaging outcomes. Our results indicate that, if anything, cancer history is associated with a higher risk of ADRD. Nevertheless, the associations that achieved statistical significance were generally of small magnitude, equivalent to 0.54–4.26 years of aging in this sample (Table S7). This suggests that, despite subtle links between cancer and several neuroimaging markers indicative of higher ADRD risk, the impact may not be clinically meaningful. The quantile regression results bolster this interpretation by showing differential effects across the distribution of the outcomes. Although the pattern did not persist in all neuroimaging markers, differentially harmful effects of cancer history at higher‐risk quantiles of white matter hyperintensity volume and cortical thickness suggest that individuals most at‐risk for dementia may also be affected the most by cancer history.
Our results also indicated that cancer treatment may play a role in ADRD‐related neurodegeneration. In the sensitivity analysis with a 5‐year washout period before the MRI scan, all adverse associations between cancer history and neuroimaging markers were attenuated and close to null. The attenuated differences among people who had survived 5+ years after diagnosis suggest that recent cancer treatment may have driven the loss in hippocampal volume and thinning in cortical thickness in the AD signature region. This is also supported by the association between breast cancer and adverse neuroimaging outcomes due to known chemotherapy‐induced brain changes among breast cancer patients. Although our study did not evaluate this directly, these associations may reflect short‐term effects of chemotherapy. 13 , 14 The sensitivity analysis may also suggest potential selective survival that people with a cancer history further in the past who survived 5+ years may potentially have healthier brains relative to the full population of people diagnosed with cancer.
Although the inverse relationship between cancer and ADRD has been studied and reported in multiple epidemiological studies, 1 , 2 , 3 , 4 , 8 , 9 , 10 , 11 , 12 our results did not support an inverse link between cancer diagnosis and ADRD‐related neurodegeneration, except a small positive association between prostate cancer and total brain volume. Potential biases in observational studies may have yielded previously observed associations. Most recent studies evaluating the cancer‐ADRD link have been based on clinical diagnosis of ADRD, which may face methodological challenges, including missed, delayed, or misclassified diagnoses. For example, several simulation‐based studies have suggested the importance of accounting for competing risk of death and diagnostic bias. 10 , 12 Future work should address the biological mechanisms linking underlying cancer and subsequent ADRD and account for various study biases.
The lack of information on cancer treatment is a significant limitation of this study, especially given the potential effect of chemotherapy and hormone therapy on brain structures. 13 , 14 , 49 While UK Biobank is linked to the UK national cancer registries, which are regarded as the gold standard for identifying cancer outcomes in the United Kingdom, the completeness of data might experience a time lag due to the complicated curation process from multiple sources. 50 UK Biobank has future plans to expand its linkage, aiming to incorporate information on cancer treatment, which can serve as a step toward addressing our current limitation. We additionally lacked some important biomarkers, such as amyloid burden. Further, the UK Biobank is a highly selected volunteer sample with unusually high socioeconomic status and healthy individuals compared to the UK population. 51 This selection process may bias observed associations. 52 Finally, selective survival bias may bias the estimated association between cancer and neuroimaging biomarkers, specifically making this association look more beneficial than it would in the absence of elevated mortality after cancer diagnoses. Individuals with a history of cancer may be less likely to survive long enough to participate in the MRI visit, which could potentially influence the observed association between cancer and neuroimaging markers. A major strength of our study is its measurement of cancer and structural MRI in a large cohort. The UK Biobank has a large sample size for participants with structural MRI, measured based on a uniformly high‐quality image acquisition protocol. The availability of both cancer registry and self‐report data on cancer history increased the validity of exposure measurement.
In this large sample, we did not find evidence to support the hypothesis that cancer history is associated with lower ADRD risk as measured by brain MRI markers. We find instead a suggestion of adverse effects of cancer history on some neuroimaging markers, including possible harmful associations for individuals already at high ADRD risk. Future studies should evaluate longitudinal changes in cognition and neuroimaging markers before and after cancer treatments.
CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest. Author disclosures are available in the supporting information.
CONSENT STATEMENT
All human subjects provided informed consent.
Supporting information
Supporting Information
Supporting Information
ACKNOWLEDGMENTS
The authors have nothing to report. This study was supported by National Institute on Aging (NIA) RF1AG059872 (Glymour, Graff, Wang), NIA T32AG049663 (Sims), NIA K99AG073454 (Ackley), and NIA K99AG075317 (Hayes‐Larson).
Wang J, Sims KD, Ackley SF, et al. Association of cancer history with structural brain aging markers of Alzheimer's disease and related dementias risk. Alzheimer's Dement. 2024;20:880–889. 10.1002/alz.13497
REFERENCES
- 1. Musicco M, Adorni F, Santo SD, et al. Inverse occurrence of cancer and Alzheimer disease: a population‐based incidence study. Neurology. 2013;81:322‐328. doi: 10.1212/WNL.0b013e31829c5ec1 [DOI] [PubMed] [Google Scholar]
- 2. Freedman DM, Wu J, Chen H, et al. Associations between cancer and Alzheimer's disease in a U.S. Medicare population. Cancer Med. 2016;5:2965‐2976. doi: 10.1002/cam4.850 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Zhang Q, Guo S, Zhang X, et al. Inverse relationship between cancer and Alzheimer's disease: a systemic review meta‐analysis. Neurol Sci. 2015;36:1987‐1994. doi: 10.1007/s10072-015-2282-2 [DOI] [PubMed] [Google Scholar]
- 4. Ma L‐L, Yu J‐T, Wang H‐F, et al. Association between cancer and Alzheimer's disease: systematic review and meta‐analysis. J Alzheimer's Dis. 2014;42:565‐573. doi: 10.3233/JAD-140168 [DOI] [PubMed] [Google Scholar]
- 5. Yarchoan M, James BD, Shah RC, et al. Association of cancer history with Alzheimer's disease dementia and neuropathology. J Alzheimers Dis. 2017;56:699‐706. doi: 10.3233/JAD-160977 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Karanth SD, Katsumata Y, Nelson PT, et al. Cancer diagnosis is associated with a lower burden of dementia and less Alzheimer's‐type neuropathology. Brain. 2022;145:2518‐2527. doi: 10.1093/brain/awac035 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Lachner C, Day GS, Camsari GB, et al. Cancer and vascular comorbidity effects on dementia risk and neuropathology in the oldest‐old. J Alzheimer's Dis. 2022;90:405‐417. doi: 10.3233/JAD-220440 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. 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:e2025515. doi: 10.1001/jamanetworkopen.2020.25515 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Wang J, Buto P, Ackley SF, et al. Association between cancer and dementia risk in the UK Biobank: evidence of diagnostic bias. European Journal of Epidemiology. Published online August 27, 2023. doi: 10.1007/s10654-023-01036-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Hayes‐Larson E, Shaw C, Ackley SF, et al. The role of dementia diagnostic delay in the Inverse Cancer–Dementia Association. J Gerontol: Series A. 2022;77:1254‐1260. doi: 10.1093/gerona/glab341 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. van der Willik KD, SB Schagen, Ikram MA. Cancer and dementia: two sides of the same coin? Eur J Clin Invest. 2018;48:e13019. doi: 10.1111/eci.13019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. 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;16:1696‐1703. doi: 10.1002/alz.12168 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Chen BT, Jin T, Patel SK, et al. Gray matter density reduction associated with adjuvant chemotherapy in older women with breast cancer. Breast Cancer Res Treat. 2018;172:363‐370. doi: 10.1007/s10549-018-4911-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Feng Y, Zhang XD, Zheng G, Zhang LJ. Chemotherapy‐induced brain changes in breast cancer survivors: evaluation with multimodality magnetic resonance imaging. Brain Imaging Behav. 2019;13:1799‐1814. doi: 10.1007/s11682-019-00074-y [DOI] [PubMed] [Google Scholar]
- 15. McDonald BC, Van Dyk K, Deardorff RL, et al. Multimodal MRI examination of structural and functional brain changes in older women with breast cancer in the first year of antiestrogen hormonal therapy. Breast Cancer Res Treat. 2022;194:113‐126. doi: 10.1007/s10549-022-06597-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Gupta A, Preis SR, Beiser AS, et al. Relationship of cancer to brain aging markers of Alzheimer's disease: the Framingham Heart Study. Adv Geriatr Med Res. 2019;1:e190006. doi: 10.20900/agmr20190006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Nudelman KNH, Risacher SL, West JD, et al. Association of cancer history with Alzheimer's disease onset and structural brain changes. Front Physiol. 2014;5:423. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Jack CR, Bennett DA, Blennow K, et al. NIA‐AA research framework: toward a biological definition of Alzheimer's disease. Alzheimers Dement. 2018;14:535‐562. doi: 10.1016/j.jalz.2018.02.018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Dubois B, Hampel H, Feldman HH, et al. Preclinical Alzheimer's disease: definition, natural history, and diagnostic criteria. Alzheimers Dement. 2016;12:292‐323. doi: 10.1016/j.jalz.2016.02.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Reitz C, Mayeux R. Alzheimer disease: epidemiology, diagnostic criteria, risk factors and biomarkers. Biochem Pharmacol. 2014;88:640‐651. doi: 10.1016/j.bcp.2013.12.024 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Jack CR, Knopman DS, Jagust WJ, et al. Tracking pathophysiological processes in Alzheimer's disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol. 2013;12:207‐216. doi: 10.1016/S1474-4422(12)70291-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Khadka A, Hebert J, Glymour MM, et al. Quantile regressions as a tool to evaluate how an exposure shifts and reshapes the outcome distribution: a primer for epidemiologists. medRxiv. 2023. doi: 10.1101/2023.05.02.23289415 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Royle NA, Booth T, Valdés Hernández MC, et al. Estimated maximal and current brain volume predict cognitive ability in old age. Neurobiol Aging. 2013;34:2726‐2733. doi: 10.1016/j.neurobiolaging.2013.05.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Chang Y‐T, Huang C‐W, Chang Y‐H, et al. Amyloid burden in the hippocampus and default mode network. Medicine (Baltimore). 2015;94:e763. doi: 10.1097/MD.0000000000000763 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Sluimer JD, van der Flier WM, Karas GB, et al. Whole‐brain atrophy rate and cognitive decline: longitudinal MR study of memory clinic patients. Radiology. 2008;248:590‐598. doi: 10.1148/radiol.2482070938 [DOI] [PubMed] [Google Scholar]
- 26. Aschenbrenner AJ, Gordon BA, Benzinger TLS, Morris JC, Hassenstab JJ. Influence of tau PET, amyloid PET, and hippocampal volume on cognition in Alzheimer disease. Neurology. 2018;91:e859‐866. doi: 10.1212/WNL.0000000000006075 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Jack CR, Petersen RC, Xu YC, et al. Prediction of AD with MRI‐based hippocampal volume in mild cognitive impairment. Neurology. 1999;52:1397‐1403. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.d’Arbeloff T, Elliott ML, Knodt AR, et al. White matter hyperintensities are common in midlife and already associated with cognitive decline. Brain Commun. 2019;1:fcz041. doi: 10.1093/braincomms/fcz041 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Carmichael O, Schwarz C, Drucker D, et al. Longitudinal changes in white matter disease and cognition in the first year of the Alzheimer Disease Neuroimaging Initiative. Arch Neurol. 2010;67:1370‐1378. doi: 10.1001/archneurol.2010.284 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Schwarz CG, Gunter JL, Wiste HJ, et al. A large‐scale comparison of cortical thickness and volume methods for measuring Alzheimer's disease severity. NeuroImage Clin. 2016;11:802‐812. doi: 10.1016/j.nicl.2016.05.017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Alfaro‐Almagro F, Jenkinson M, Bangerter NK, et al. Image processing and quality control for the first 10,000 brain imaging datasets from UK Biobank. Neuroimage. 2018;166:400‐424. doi: 10.1016/j.neuroimage.2017.10.034 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Griffanti L, Zamboni G, Khan A, et al. BIANCA (Brain Intensity AbNormality Classification Algorithm): a new tool for automated segmentation of white matter hyperintensities. Neuroimage. 2016;141:191‐205. doi: 10.1016/j.neuroimage.2016.07.018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Desikan RS, Ségonne F, Fischl B, et al. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage. 2006;31:968‐980. doi: 10.1016/j.neuroimage.2006.01.021 [DOI] [PubMed] [Google Scholar]
- 34. Voevodskaya O, Simmons A, Nordenskjöld R, et al. The effects of intracranial volume adjustment approaches on multiple regional MRI volumes in healthy aging and Alzheimer's disease. Front Aging Neurosci. 2014;6:264. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Ording AG, Horváth‐Puhó E, Veres K, et al. Cancer and risk of Alzheimer's disease: small association in a nationwide cohort study. Alzheimers Dement. 2020;16:953‐964. doi: 10.1002/alz.12090 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Sun M, Wang Y, Sundquist J, Sundquist K, Ji J. The association between cancer and dementia: a National Cohort Study in Sweden. Front Oncol. 2020;10:73. doi: 10.3389/fonc.2020.00073 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Fahradyan A, Howell AC, Wolfswinkel EM, Tsuha M, Sheth P, Wong AK. Updates on the management of non‐melanoma skin cancer (NMSC). Healthcare (Basel). 2017;5:E82. doi: 10.3390/healthcare5040082 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. DeCarli C, Massaro J, Harvey D, et al. Measures of brain morphology and infarction in the Framingham Heart Study: establishing what is normal. Neurobiol Aging. 2005;26:491‐510. doi: 10.1016/j.neurobiolaging.2004.05.004 [DOI] [PubMed] [Google Scholar]
- 39. Driscoll I, Davatzikos C, An Y, et al. Longitudinal pattern of regional brain volume change differentiates normal aging from MCI. Neurology. 2009;72:1906‐1913. doi: 10.1212/WNL.0b013e3181a82634 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Thambisetty M, Wan J, Carass A, An Y, Prince JL, Resnick SM. Longitudinal changes in cortical thickness associated with normal aging. Neuroimage. 2010;52:1215‐1223. doi: 10.1016/j.neuroimage.2010.04.258 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Lloyd‐Jones DM, Hong Y, Labarthe D, et al. Defining and setting national goals for cardiovascular health promotion and disease reduction. Circulation. 2010;121:586‐613. doi: 10.1161/CIRCULATIONAHA.109.192703 [DOI] [PubMed] [Google Scholar]
- 42. Townsend P, Phillimore P, Beattie A. Health and Deprivation: Inequality and the North. Routledge; 1988. [Google Scholar]
- 43. Beyerlein A. Quantile regression—opportunities and challenges from a user's perspective. Am J Epidemiol. 2014;180:330‐331. doi: 10.1093/aje/kwu178 [DOI] [PubMed] [Google Scholar]
- 44. Hernán MA, Hernández‐Díaz S, Robins JM. A structural approach to selection bias. Epidemiol. 2004;15:615‐625. [DOI] [PubMed] [Google Scholar]
- 45. Bradley V, Nichols TE. Addressing selection bias in the UK Biobank neurological imaging cohort. medRxiv. 2022. doi: 10.1101/2022.01.13.22269266 [DOI] [Google Scholar]
- 46. Busatto GF, Diniz BS, Zanetti MV. Voxel‐based morphometry in Alzheimer's disease. Expert Rev Neurother. 2008;8:1691‐1702. doi: 10.1586/14737175.8.11.1691 [DOI] [PubMed] [Google Scholar]
- 47. Correa DD, Root JC, Kryza‐Lacombe M, et al. Brain structure and function in patients with ovarian cancer treated with first‐line chemotherapy: a pilot study. Brain Imaging Behav. 2017;11:1652‐1663. doi: 10.1007/s11682-016-9608-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. McDonald BC, Conroy SK, Ahles TA, West JD, Saykin AJ. Gray matter reduction associated with systemic chemotherapy for breast cancer: a prospective MRI study. Breast Cancer Res Treat. 2010;123:819‐828. doi: 10.1007/s10549-010-1088-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49. Kantarci K, Tosakulwong N, Lesnick TG, et al. Effects of hormone therapy on brain structure: a randomized controlled trial. Neurology. 2016;87:887‐896. doi: 10.1212/WNL.0000000000002970 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Conroy MC, Lacey B, Bešević J, et al. UK Biobank: a globally important resource for cancer research. Br J Cancer. 2023;128:519‐527. doi: 10.1038/s41416-022-02053-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51. Fry A, Littlejohns TJ, Sudlow C, et al. Comparison of sociodemographic and health‐related characteristics of UK Biobank participants with those of the general population. Am J Epidemiol. 2017;186:1026‐1034. doi: 10.1093/aje/kwx246 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52. Batty GD, Gale CR, Kivimäki M, Deary IJ, Bell S. Comparison of risk factor associations in UK Biobank against representative, general population based studies with conventional response rates: prospective cohort study and individual participant meta‐analysis. BMJ. 2020;368:m131. doi: 10.1136/bmj.m131 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supporting Information
Supporting Information