Abstract
Background:
Cross-sectional studies suggest self-reported cancer history is associated with decreased risk of Alzheimer’s disease (AD). However, little is known about how self-reported cancer affects longitudinal AD progression, the primary outcome in clinical trials and observational studies.
Objective:
To determine self-reported cancer history’s effect on longitudinal AD progression in an observational study.
Methods:
We utilized data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) to evaluate progression to AD by self-reported all-cancer, breast, prostate, colorectal, or non-melanoma skin cancer history. Linear mixed effects models were used to examine baseline differences and rates of progression on the Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) by self-reported cancer history. Age at AD onset was examined using consensus clinical diagnoses with Cox proportional hazards regression.
Results:
Among 1271 participants, models revealed no significant differences in progression over time but did reveal significantly lower baseline ADAS-Cog score, indicating better cognition at a given age in those with self-reported cancer history. Cox models indicated those with self-reported cancer history had significantly later age of AD onset (HR: 0.67, 95% CI: 0.53–0.85) after adjustment for covariates.
Conclusion:
Participants with self-reported cancer history entered ADNI with better cognition and later age of AD onset, but progressed similarly to participants without such history, indicating differences in AD between those with and without self-reported cancer history emerge early in the disease course. Such differences in longitudinal progression by self-reported cancer history could affect AD trials and observational studies, given the current focus on early disease course. Further investigation is warranted with detailed longitudinal assessment of cancer and AD.
Keywords: Alzheimer’s disease, mild cognitive impairment, disease progression, cancer, cognitive impairment
INTRODUCTION
Alzheimer’s disease (AD) is associated with older age and comorbidities including diabetes, hypertension, and cardiovascular disease [1]. Cancer seems to have an inverse relationship with AD, primarily based on self-reported history of cancer among older adults [2][3][4][5]. Most studies have examined the cross-sectional relationship between self-reported cancer history and AD, which would not allow for determination of when in the course of AD this inverse relationship begins to occur. To date, no study has examined effects of self-reported cancer history on longitudinal AD progression. Differences in progression would be of particular importance for observational studies and clinical trials in AD, which focus on longitudinal outcomes.
To better understand the implications of cross-sectional findings for longitudinal studies, we examined progression of AD in patients with self-reported cancer history in an observational cohort study. We estimated the rate of decline on the Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) based on self-reported cancer history overall and specifically in breast, prostate, colorectal, and non-melanoma skin cancers. We also estimated time to AD onset based on self-reported cancer history in this cohort. In addition, we estimated the age at which ADAS-cog scores began to diverge between those with and without self-reported cancer history, due to potential pre-existing impairment in these groups. This initial analysis is intended to spur further research on the effect of cancer history on the progression of AD, particularly the implications for clinical trials and observational studies.
METHODS
Study Population
Participants were from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) ADNI3 database downloaded August 27, 2018 [6]. ADNI is a longitudinal multicenter study to develop standards for clinical, imaging, genetic, and biochemical biomarkers for early detection and tracking of AD progression. [6] Inclusion criteria for the ADNI parent study are ages 55 through 90 years old at various stages of cognitive functioning (normal, MCI, and dementia due to AD), good overall health with no diseases preventing enrollment, and stable doses of permitted medications (including memory-related therapies) for four weeks prior to enrollment [6][7]. Diagnoses of cognitive impairment were made following ADNI guidelines including NINDS-ADRDA criteria for probable AD [8] and Petersen criteria for MCI [9]. The ADNI study was reviewed and approved by Institutional Review Boards (IRBs) at each study site and written informed consent was obtained from each participant. This report was exempted from IRB approval by the UAB IRB.
A total of 1271 participants with MCI or AD were used in the analysis. Participants were included in this analysis if they had two or more ADAS-Cog scores and were diagnosed as MCI or AD at baseline. Participants were not included from this analysis if they were diagnosed as cognitively normal at baseline, or if they were missing information on dementia diagnosis, exam dates for baseline or follow-up ADAS-Cog scores, demographics, or baseline Mini Mental State Examination scores. For the sub-analyses, participants were required to have APOE genotype information. For the survival analysis, participants were not included if they lacked data on change in diagnosis (from MCI to AD).
Data Collection
Demographic and Clinical Information
Demographic and clinical information was collected at baseline. Variables examined included sex, race, age, education, and APOE genotype. Education level was total years of school [10]. Age in ADNI was not collected specifically, rather, birth year and birth month. To calculate age, we used the 15th of the month for all participants. Race, ethnicity (Hispanic vs. non-Hispanic), sex, and marital status (married, widowed, divorced, never married, or unknown) were self-reported. Race was categorized into White vs. non-White for this report.
Self-Reported Cancer History
Self-reported comorbidities were obtained from the ADNI medical history file. These variables were collected using standardized interview procedures across study sites by a nurse or other medical professional. Participants or caregivers were queried regarding current or previous diseases, which were recorded as system affected, description of the problem, dates of the problem, and whether it was current or resolved [10][11][12]. Our self-reported cancer history group consisted of all participants with self-reported previous malignancy (if the organ system code for “Malignancy” was positive). Additionally, Perl regular expressions [13] were used to search other organ systems to determine self-reports of cancers which may have been coded by the organ system affected (Supplement 1). Perl regular expressions were also used to classify self-reported cancers into specific cancer types: breast, prostate, colorectal, and skin (not including melanoma). Although non-melanoma skin cancer is typically not part of cancer statistics, we have included it in our study based on results of previous analyses in the ADNI dataset [14]. An all-cancer group was also created to determine whether effects on progression of AD is a general effect of self-reported cancer or is specific to the type of cancer reported. Our reference group consisted of all ADNI participants without a self-reported history of cancer.
Alzheimer’s Disease Progression
Alzheimer’s Disease Assessment Scale-Cognitive Subscale
The ADAS-Cog [15] is a neuropsychological assessment evaluating memory, orientation, language, praxis, and word-finding difficulty. The scale is scored from 0 to 70 errors with higher scores indicating more cognitive impairment. Scores were collected at baseline and at follow-up visits every 6 months. A two-point difference is considered to be a significant change in most AD clinical trials [16].
Conversion to AD
For progression from MCI to AD, diagnostic summaries were collected via review of all visit documents across study sites by site principal investigators (PIs) who were medical professionals, usually a physician, trained in diagnostics of normal cognition, MCI, and various stages of AD. Diagnoses were adjudicated by the clinical monitor PI and the clinical conversion committee [10][11][12]. Participant status at each visit was categorized as stable, reverted, or converted, with the former two considered no progression and latter considered progression. Time to progression was calculated as the difference between the participant’s visit date where conversion was documented and their birthdate. Additionally, participants who entered the study with AD were queried as to the self-reported onset of their AD symptoms. Only year was collected, so date of conversion was set as the middle of the year in which onset occurred (June 15). To calculate time to progression, the date of conversion was subtracted from participants’ birthdate.
Statistical Analysis
Bivariate analyses using t-tests and chi-square tests for continuous and categorical variables, respectively, were used to assess potential covariates. Fisher’s exact test was used where expected cell counts were low for categorical variables. Linear mixed models (random coefficient models) [17] were used to examine progression on the ADAS-Cog by overall and specific self-reported cancer history status, adjusting for covariates of race, sex, education, with further adjustment for APOE ε4 allele carrier status (positive or negative) in subanalyses. Age was used as the time variable for mixed modeling, centered on the mean age based on cognitive diagnosis (both AD and MCI, MCI only, or AD only). Separate age-centered time variables were created for each disease group. As non-melanoma skin cancer is often not included in cancer classifications, a sensitivity analysis was performed with an all-cancer group that did not include non-melanoma skin cancer. Instead, participants reporting only non-melanoma skin cancer were categorized as having no self-reported cancer for sensitivity analysis, while participants reporting non-melanoma skin cancer plus an additional cancer were retained in the self-reported all-cancer group due to the presence of another cancer type.
Kaplan-Meier analyses with log-rank tests were used to summarize the time to onset of AD based on self-reported cancer history. Cox proportional hazards models were used to assess time to onset of AD by overall and specific self-reported cancer history controlling for relevant covariates, with censoring at the last recorded visit if progression had not occurred. Participant age was used as the time variable in this analysis. Education, baseline ADAS-Cog score, and APOE ε4 allele carrier status (positive or negative) were included in the final model regardless of statistical significance in bivariate analyses due to a priori knowledge of their association with cognitive decline [18][19][20][21]. Race was not included in Cox regression analysis. Over 90% of the ADNI cohort is White, non-Hispanic so adjusting for race in the model would not produce appropriate sample size in each group causing model convergence issues. Analyses included examination of proportional hazards assumptions. All statistical analyses were assessed at α=0.05 significance level using SAS Version 9.4 (SAS Institute, Inc., Cary, NC). SAS code is included in Supplement 1.
RESULTS
Bivariate analyses indicated those in the self-reported all-cancer history group were older, more likely to be White, have a diagnosis of MCI at baseline, and more likely to be married. Those with history of non-melanoma skin cancer were also more highly educated (Table 1). Table 2 includes the frequency of cancers among participants, including those with more than one cancer by group. A total of 52 participants had more than one cancer and a total of 9 had more than two cancers.
Table 1.
Baseline Characteristics by Self-Reported Cancer History Status*
| Variable | All (n=367)§ | Breast (n=36) | Prostate (n=103) | Colorectal (n=29) | Non-Melanoma Skin (n=165) | No Cancer (n=904) |
|---|---|---|---|---|---|---|
| Age (years) | 76.14 ± 7.00 | 75.50 ± 6.70 | 77.40 ± 6.44 | 79.38 ± 6.13 | 75.65 ± 7.34 | 72.70 ± 7.76 |
| Education (years) | 15.97 ± 2.86 | 14.97 ± 3.07 | 16.02 ± 2.53 | 15.31 ± 3.33 | 16.23 ± 2.94 | 15.63 ± 2.89 |
| Sex | ||||||
| Male | 253 (68.94) | 0 | 103 (100) | 17 (58.62) | 120 (72.73) | 487 (53.87) |
| Female | 114 (31.06) | 36 (100) | 0 | 12 (41.38) | 45 (27.27) | 417 (46.13) |
| Race | ||||||
| White | 357 (97.28) | 35 (97.22) | 99 (96.12) | 28 (96.55) | 160 (96.97) | 828 (91.59) |
| Non-White | 10 (2.72) | 1 (2.78) | 4 (3.88) | 1 (3.45) | 5 (3.03) | 76 (8.41) |
| Ethnicity | ||||||
| Hispanic | 1 (1.91) | 1 (2.78) | 3 (2.91) | 0 | 1 (0.61) | 34 (3.76) |
| Non-Hispanic | 359 (97.82) | 35 (97.22) | 99 (96.12) | 29 (100) | 164 (99.39) | 864 (95.58) |
| Unknown | 1 (0.27) | 0 | 1 (0.97) | 0 | 0 | 6 (0.66) |
| Marital Status | ||||||
| Married | 303 (82.56) | 23 (63.89) | 96 (93.20) | 20 (68.97) | 137 (83.03) | 694 (76.77) |
| Divorced | 33 (8.99) | 6 (16.67) | 4 (3.88) | 7 (24.14) | 13 (7.88) | 103 (11.39) |
| Widowed | 20 (5.45) | 4 (11.11) | 2 (1.94) | 2 (6.90) | 10 (6.06) | 78 (8.63) |
| Never Married | 7 (1.91) | 3 (8.33) | 0 | 0 | 3 (1.82) | 26 (2.88) |
| Unknown | 4 (1.09) | 0 | 1 (0.97) | 0 | 2 (1.21) | 3 (0.33) |
| Baseline Diagnosis | ||||||
| Mild Cognitive Impairment (MCI) | 282 (76.84) | 26 (72.22) | 78 (75.73) | 24 (82.76) | 133 (80.61) | 637 (70.46) |
| Alzheimer’s Disease (AD) | 85 (23.16) | 10 (27.78) | 25 (24.27) | 5 (17.24) | 32 (19.39) | 267 (29.54) |
| APOE e4 Status† | ||||||
| Positive | 183 (51.84) | 19 (52.78) | 42 (42.42) | 15 (51.72) | 84 (52.50) | 477 (56.18) |
| Negative | 170 (48.16) | 17 (47.22) | 57 (57.58) | 14 (48.28) | 76 (47.50) | 372 (43.82) |
| Baseline ADAS-Cog Score‡ | 12.53 ± 6.42 | 10.41 ± 5.08 | 13.29 ± 6.59 | 11.16 ± 4.92 | 12.22 ± 6.07 | 13.20 ± 6.78 |
Assessed at α=0.05 significance level using t-tests for continuous variables and chi-square tests for categorical variables (Fisher’s exact tests where expected cell counts were low). Bold font indicates statistically significant differences compared to the no cancer column. Some variables may not sum to column total due to missingness in the variable: for the all cancer group ethnicity n=6 are missing.
APOE ε4=apolipoprotein ε4 allele; not all participants had APOE ε4 genotyping (n=1202)
ADAS-Cog: Alzheimer’s Disease Assessment Scale-Cognitive Subscale, scores range from 0 to 70
The All column includes 34 cancers which were not individually classified (367–333=34).
Table 2.
Frequency of Synchronous Cancers.
| Breast | Prostate | Colorectal | Non-Melanoma Skin | Other | None | |
|---|---|---|---|---|---|---|
| Breast | 26 | 0 | 1 | 3 | 4 | --- |
| Prostate | 0 | 80 | 2 | 11 | 4 | --- |
| Colorectal | 1 | 2 | 13 | 3 | 5 | --- |
| Non-Melanoma Skin | 3 | 11 | 3 | 123 | 19 | --- |
| Other | 4 | 4 | 5 | 19 | 64 | --- |
| None | --- | --- | --- | --- | --- | 904 |
Frequency does not add up to total sample size of 1271 due to a frequency of n=9 with greater than 2 cancer types: Breast-colorectal-skin (n=1), Prostate-colorectal-other (n=3), Prostate-skin-other (n=3), Breast-skin-other (n=1), and Colorectal-skin-other (n=1)
In mixed model analysis, no significant differences were seen in rate of progression over time on the ADAS-Cog for the self-reported all-cancer history group (βslope: 0.06, 95% CI: −0.15, 0.27), regardless of baseline diagnosis (MCI or AD) (Table 3). This finding remained in sensitivity analyses that did not include non-melanoma skin cancer in the all-cancer group (βslope: 0.02, 95% CI: −0.23, 0.27). However, self-reported history of cancer was associated with a statistically significant −4.42-point difference in intercept for ADAS-Cog score compared to no self-reported cancer history, after adjustment for race, sex, and education (95% CI: −5.90, −2.94). Again, this finding remained in sensitivity analyses that did not include non-melanoma skin cancer history in the all-cancer group (βintercept: −3.38, 95% CI: −5.11, −1.65). Similar differences in intercepts were seen across self-reported cancer types [Breast: −6.72 (95%CI: −10.92, −2.53); Prostate: −4.27 (95% CI: −6.88, −1.65); Colorectal: −10.76 (95% CI: −15.69, −5.84); Non-Melanoma Skin: −4.82 (95% CI: −6.81, −2.84)]. Further adjustment for APOE ε4 status did not significantly change results. Stratified analyses by baseline diagnosis showed that MCI patients had results similar to that observed in overall analyses within each self-reported cancer history category. However, in MCI patients with colorectal cancer history, there was evidence of a faster progression on ADAS-Cog over time [0.72, (95% CI: 0.04–1.40)] (Table 4). The AD group did not exhibit any significant differences in intercepts for ADAS-cog score for overall or specific self-reported cancer history, nor any differences in rate of progression by self-reported cancer history type. These findings were unchanged in sensitivity analyses that did not include non-melanoma skin cancer in the all-cancer group. Figure 1 shows a graphical depiction of predicted ADAS-Cog scores by age based on unadjusted mixed model regression results.
Table 3.
Beta Estimates and 95% CI for Random Coefficients Models Based on Self-Reported History of Specific Cancer Types Regardless of Baseline Cognitive Diagnosis (MCI or AD)*
| Model 1 | ||
|---|---|---|
| Cancer Type | β (95% CI)-Intercept | β (95% CI)-Slope |
| All Cancer vs. No Cancer (n=1271) | −4.42 (−5.90, −2.94) | 0.06 (−0.15, 0.27) |
| Breast Cancer vs. No Cancers (n=940) | −6.72 (−10.92, −2.53) | 0.33 (−0.25, 0.91) |
| Prostate Cancer vs. No Cancers (n=1007) | −4.27 (−6.88, −1.65) | −0.03 (−0.37, 0.31) |
| Colorectal Cancer vs. No Cancers (n=933) | −10.76 (−15.69, −5.84) | 0.54 (−0.08, 1.16) |
| Non-Melanoma Skin Cancers vs. No Cancers (n=1069) | −4.82 (−6.81, −2.84) | 0.04 (−0.22, 0.31) |
| Model 2 | ||
| All Cancer vs. No Cancer (n=1202) | −4.17 (−5.67, −2.67) | 0.01 (−0.21, 0.23) |
| Breast Cancer vs. No Cancers (n=885) | −6.27 (−10.41, −2.13) | 0.27 (−0.33, 0.86) |
| Prostate Cancer vs. No Cancers (n=948) | −3.41 (−6.08, −0.75) | −0.12 (−0.46, 0.23) |
| Colorectal Cancer vs. No Cancers (n=878) | −10.81 (−15.68, −5.94) | 0.55 (−0.09, 1.18) |
| Non-Melanoma Skin Cancer vs. No Cancers (n=1009) | −4.66 (−6.64, −2.68) | 0.02 (−0.25, 0.30) |
Estimated using linear random coefficients models of change in Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) with a random coefficient for time at α=0.05 significance level.
Model 1: Adjusted for race (White vs. Non-White), sex, and education
Model 2: Adjusted for variables in Model 1 plus apolipoprotein ε4 status (yes vs. no)
For sensitivity analysis with participants with only non-melanoma skin cancer history, estimates remained similar (intercept: β: −3.38, 95% CI: −5.11, −1.65; slope: β: 0.02, 95% CI: −0.23, 0.27).
95% CI: 95% confidence interval
MCI: Mild Cognitive Impairment
AD: Alzheimer’s disease
Table 4.
Beta estimates and 95% CI for Random Coefficients Models Based on Self-Reported History of Specific Cancer Types by MCI or AD Diagnosis*
| MCI—Without ApoE4 | ||
|---|---|---|
| Cancer Type | β (95% CI)-Intercept | β (95% CI)-Slope |
| All Cancer vs. No Cancer (n=919) | −3.80 (−5.32, −2.28) | 0.14 (−0.09, 0.37) |
| Breast Cancer vs. No Cancers (n=663) | −5.23 (−9.66, −0.80) | 0.45 (−0.20, 1.11) |
| Prostate Cancer vs. No Cancers (n=715) | −4.53 (−7.20, −1.86) | 0.06 (−0.30, 0.42) |
| Colorectal Cancer vs. No Cancer (n=661) | −9.22 (−14.08, −4.36) | 0.73 (0.07, 1.39) |
| Non-Melanoma Skin Cancer vs. No Cancer (n=770) | −3.80 (−5.80, −1.80) | 0.08 (−0.20, 0.37) |
| AD—Without ApoE4 | ||
| All Cancer vs. No Cancer (n=352) | 0.04 (−2.13, 2.20) | −0.20 (−0.46, 0.06) |
| Breast Cancer vs. No Cancers (n=277) | −2.85 (−8.55, 2.85) | −0.31 (−1.02, 0.40) |
| Prostate Cancer vs. No Cancers (n=292) | 1.45 (−2.42, 5.33) | −0.26 (−0.67, 0.14) |
| Colorectal Cancer vs. No Cancer (n=272) | −6.12 (−19.04, 6.80) | 0.03 (−1.23, 1.29) |
| Non-Melanoma Skin Cancer vs. No Cancer (n=299) | −0.45 (−3.81, 2.90) | −0.04 (−0.44, 0.35) |
| MCI—With ApoE4 | ||
| All Cancer vs. No Cancer (n=863) | −3.72 (−5.29, −2.15) | 0.09 (−0.15, 0.33) |
| Breast Cancer vs. No Cancers (n=619) | −4.68 (−9.13, −0.23) | 0.39 (−0.28, 1.06) |
| Prostate Cancer vs. No Cancers (n=668) | −3.96 (−6.73, −1.19) | −0.03 (−0.41, 0.34) |
| Colorectal Cancer vs. No Cancer (n=617) | −9.49 (−14.37, −4.61) | 0.72 (0.04, 1.40) |
| Non-Melanoma Skin Cancer vs. No Cancer (n=721) | −3.84 (−5.88, −1.81) | 0.04 (−0.26, 0.34) |
| AD—With ApoE4 | ||
| All Cancer vs. No Cancer (n=339) | 0 (−2.21, 2.22) | −0.19 (−0.46, 0.07) |
| Breast Cancer vs. No Cancers (n=266) | −2.98 (−8.78, 2.83) | −0.31 (−1.03, 0.42) |
| Prostate Cancer vs. No Cancers (n=280) | 1.46 (−2.50, 5.42) | −0.27 (−0.68, 0.14) |
| Colorectal Cancer vs. No Cancer (n=261) | −6.37 (−19.55, 6.80) | 0.08 (−1.22, 1.38) |
| Non-Melanoma Skin Cancer vs. No Cancer (n=288) | −0.35 (−3.76, 3.06) | −0.02 (−0.42, 0.38) |
Estimated using linear random coefficients models of change in Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) with a random coefficient for time at α=0.05 significance level.
Models adjusted for race (White vs. Non-White), sex, education, and apolipoprotein ε4 status (yes vs. no) if indicated
95% CI: 95% confidence interval; MCI: mild cognitive impairment; AD: Alzheimer’s disease
Figure 1. Predicted ADAS-Cog Scores Over Time by Self-Reported Cancer History in MCI.
The group with no self-reported cancer history begins with a higher Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) score, indicating worse cognition at baseline, but progresses similar to the group with self-reported cancer history. ADAS-Cog scores were predicted over time using unadjusted linear mixed effects models.
Kaplan-Meier analysis of unadjusted time to event probabilities between the self-reported all-cancer history group compared to the no self-reported cancer history group showed differences between the curves that were statistically significant (χ2=60.57, p=<0.0001). Those with self-reported cancer history showed evidence of later age of onset of AD than those without self-reported cancer history. (Figure 2).
Figure 2. Kaplan-Meier Analysis by Self-Reported Cancer History.
Participants with MCI and self-reported cancer history (Cancer +) have a later age of onset of AD compared to those without self-reported cancer history (Cancer −) in subset of participants with measured APOE ε4 status (n=1202).
Consistent with Kaplan-Meier analyses, Cox regression showed a significantly later age of onset of AD in participants with any self-reported cancer history compared to those without any self-reported cancer history (HR: 0.68, 95% CI: 0.54–0.86) (Table 5). This association remained after adjustment for age, education, and baseline ADAS-Cog score (HR: 0.71, 95% CI 0.56–0.90), and in subanalyses adjusting for APOE ε4 status (HR: 0.67, 95% CI: 0.53–0.85) (Table 5). Self-reported prostate cancer and non-melanoma skin cancer history exhibited similar associations to the overall cancer history group even after full adjustment, but the later age of onset was not significant in those with self-reported breast or colorectal cancer history (Table 5).
Table 5.
Hazard Ratios (HR) and 95% CI for Time-to-Onset of AD Based on Self-Reported Cancer History (n=1202)*
| Comparison | HR | 95% CI | p-value |
|---|---|---|---|
| All-Cancer vs. No Cancer | |||
| Without ApoE4 | 0.71 | 0.56–0.90 | 0.0040 |
| With ApoE4 | 0.67 | 0.53–0.85 | 0.0009 |
| Breast Cancer vs. No Cancer | |||
| Without ApoE4 | 0.67 | 0.30–1.52 | 0.3419 |
| With ApoE4 | 0.65 | 0.29–1.47 | 0.3013 |
| Prostate Cancer vs. No Cancer | |||
| Without ApoE4 | 0.60 | 0.41–0.89 | 0.0113 |
| With ApoE4 | 0.59 | 0.40–0.87 | 0.0085 |
| Colorectal Cancer vs. No Cancer | |||
| Without ApoE4 | 0.96 | 0.56–1.65 | 0.8795 |
| With ApoE4 | 0.88 | 0.51–1.52 | 0.6465 |
| Non-Melanoma Skin Cancer vs. No Cancer | |||
| Without ApoE4 | 0.74 | 0.55–1.01 | 0.0579 |
| With ApoE4 | 0.71 | 0.52–0.96 | 0.0273 |
Assessed using Cox proportional hazards models at the α=0.05 significance level
Models adjusted for education in years and baseline Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) score, and apolipoprotein (ApoE4) ε4 status (yes or no) if indicated (n=858 with ApoE4)
95% CI: 95% confidence interval; AD: Alzheimer’s disease
DISCUSSION
In this secondary analysis of the ADNI observational cohort, participants with self-reported cancer history and MCI at baseline had an approximately 4-point lower difference in mixed model intercepts after controlling for covariates. This suggests that ADAS-Cog score for those with self-reported cancer history was lower for a given age compared to those with no self-reported cancer history in the MCI sample, corresponding to less cognitive impairment in the former. Further analysis of the mixed model suggests no difference in slopes of ADAS-Cog scores (rate of cognitive decline) over time between the self-reported cancer history and no self-reported cancer history groups. However, due to heterogeneity among cancer types, the self-reported all-cancer results must be interpreted with caution. Separate analysis by specific self-reported cancer type revealed similar results to the self-reported all-cancer history analysis after adjustment for relevant covariates, except for self-reported colorectal cancer. In those with AD at baseline, no significant differences in ADAS-Cog intercepts (score at a given age) and slope (rate of cognitive decline) were seen for overall and specific self-reported cancer history groups. Additionally, consistent with previous literature [14], self-reported cancer history in ADNI was associated with later age of onset of AD after adjusting for education, baseline ADAS-Cog score, and APOE ε4 status for self-reported overall, prostate, and non-melanoma skin cancer, though not for self-reported breast and colorectal cancer. Sensitivity analyses that did not include self-reported non-melanoma skin cancer in the all-cancer group yielded similar results regardless of baseline cognitive diagnosis. Collectively, these results indicate that participants with self-reported cancer history do not have greater progression of cognitive impairment in AD than participants with no self-reported cancer history. During the course of MCI, those with self-reported cancer history also progressed at a similar rate to those without self-reported cancer history. Thus, differences in cognitive scores between those with and without self-reported cancer history appear to develop early in the course of their cognitive trajectories. These results are consistent with the inverse relationship between AD and self-reported cancer history in cross-sectional analyses [2][3][4][5], but our analysis expands on these results by examining the timing of this effect.
The findings from this analysis emphasize the need for more detailed longitudinal follow-up studies of cancer and AD. Studies of cancer among AD patients have focused mostly on self-reported cancer history, which is limited by potential recall bias inherent in self-report. Self-reported cancer history is further limited by lack of detailed information about the cancer treatments and staging which would be useful in assessing any relationship between treatment and cognition at AD onset. Studies of AD among cancer patients have focused on cross-sectional follow-up AD assessment in individuals with documented cancer history. This approach has its own limitations in that it could be subject to survivor bias, where patients with cancer may die prior to assessment diagnosis of AD, and does not provide information on longitudinal progression of AD.
The findings from this analysis must also be viewed in the context of what is known about short-term cognitive effects of cancer. Cancer is associated with cognitive deficits prior to, during, and after treatment. Evidence is mixed for how long these may last [22][23] and prognosis for cognitive deficits may differ based on treatment modality [24]. The cognitive deficits observed in cancer patients include executive function, verbal skills, and memory [22][23][24][25]. Similar patterns of deficits are observed in AD progression and worsen over time [26]. Since cancer patients may already possess symptoms in similar domains, progression of AD would intuitively be more rapid than the general population. However, our results indicate that this may not be the case. This is in contrast to pre-existing cognitive deficits due to other disorders, which are usually associated with higher risk of AD [27]. While it is possible that some of these discrepancies arise due the limitations of self-report, studies have indicated relatively high sensitivity and specificity in self-reported cancer history compared to cancer registries, especially for the cancers under study in this project [28][29][30]. This further emphasizes the need for detailed longitudinal studies of cancer and AD to determine if the results of our analysis are due to cancer itself or to the self-reporting of cancer history.
These results clarify the implications of self-reported cancer history for observational studies and treatment trials for individuals with AD. Participants in AD studies with self-reported cancer history have a similar rate of progression of AD as those without self-reported cancer history. Thus, longitudinal studies of the course of AD, and treatment trials examining the effects of interventions on AD progression, would seem to not be affected by self-reported cancer history of the participants. However, given the small sample sizes for specific self-reported cancer types in our analysis, this conclusion should be interpreted as preliminary, and differences between individuals with and without self-reported cancer history may be apparent in larger studies. Participants in AD studies with self-reported cancer history also have fewer cognitive deficits at baseline and later age of onset for AD. Thus, as clinical trials for AD increasingly shift to treatment in the earliest stages of the disease, self-reported cancer history may assume a much greater importance as a confounder of treatment effects. The 4-point baseline difference in cognition between participants with and without self-reported history of cancer is greater than the typical effect size for cholinesterase inhibitors, and similar to the effect size for other potential confounders in AD clinical trials [31][32][33]. If self-reporting does not seem to be a driver of our results, this could have implications for identifying factors which may predispose those with cancer history to future development of AD, and identify the need for more focused treatment of AD cognitive deficits in trial participants with prior cancer.
Strengths and Limitations
This study has several notable strengths. First, this study is the first to examine longitudinal progression based on self-reported cancer history and specific self-reported cancers rather than previous studies examining only cross-sectional associations. Longitudinal analysis allows for examination of rates of progression using slopes (rate of cognitive decline) and intercepts (level of impairment at a particular age) in mixed models, while cross-sectional analyses cannot distinguish between the two. Secondly, participants in self-reported cancer and no cancer history groups come from the same primary study base, increasing internal validity. Thirdly, data collection procedures are highly standardized across study sites limiting potential for information bias.
The primary limitation to consider is that cancer history was determined via self-report, which can be inaccurate compared to objectively confirmed cancers via biopsy or imaging as noted above; verification of cancer diagnoses through medical records would strengthen findings. Other limitations include potential selection biases, in that most cancer survivors included in ADNI would be less advanced or in remission longer than the general population. This would be due to study exclusion criteria and likelihood that participants with more advanced cancers would be deceased prior to developing MCI or AD. This selection bias would correspond to less invasive, less damaging treatments and more time to recover from potential cancer-related cognitive impairment, resulting in less severe cognitive deficits for those with cancer history enrolled in ADNI compared to those with cancer history in the general population. We do not know the extent to which this bias would influence self-reported cancer history. Inability to account for specific cancer treatments could explain some of the associations seen given known toxic effects of chemotherapy on the brain. This also could have affected ability to self-report cancer status and recall details about their cancers [34][35]. Furthermore, age of onset in the AD group is self-reported and not measured, which could bias the results if inaccurate. We were also unable to adjust for comorbidities in our analysis due to lack of detailed data in many ADNI participants for calculating a comorbidity score. This is similar to previous analyses in ADNI assessing cognition and brain abnormalities in cancer [14]. Additionally, while the results from specific self-reported cancer history analyses show no evidence of a difference in cognitive decline over time, sample size was limited for each self-reported cancer type and even further limited when also stratifying by baseline diagnosis. Some of the association in the self-reported all-cancer history analyses could be driven by self-reported cancers not specifically examined in this analysis. However, the self-reported cancers included in this analysis are very common and would be expected to reflect the most common self-reported cancer types seen in a general clinic setting. Most importantly, use of an AD observational study like ADNI does not allow us to draw epidemiological inferences such as whether or not cancer is a risk factor for AD.
In conclusion, these analyses suggest that, in observational studies of AD, self-reported all-cancer history is associated with better cognitive scores at a given age in AD patients, but no difference in disease progression over the course of AD. These results echo previous cross-sectional studies indicating an older age of onset for those with self-reported cancer history and with self-reported history of specific cancer subtypes, but this is also the first study focusing specifically on the relationship between self-reported cancer and longitudinal AD progression. Our results demonstrate the need for detailed longitudinal studies of cancer history and and how this affects AD and its progression to clarify complex relationship between the two diseases.
Supplementary Material
ACKNOWLEDGEMENTS
Data used in preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at http://adni.loni.usc.edu/wp-content/themes/freshnews-devv2/documents/policy/ADNI_Acknowledgement_List%205-29-18.pdf
Funding acknowledgments: Funding for this report was provided by NIH R01 AG057684 (MEF, GRC, LSS, REK) and by NIH R01 AG037561 (GRC, LSS, REK). The funding agency has no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
Footnotes
| Elements of Financial/ Personal Conflicts | MEF | KLT | GRC | LSS | REK |
|---|---|---|---|---|---|
| Yes/No | Yes/No | Yes/No | Yes/No | Yes/No | |
| Employment or Affiliation | No | No | No | No | No |
| Grants/Funds | No | Yes | Yes | Yes | No |
| Honoraria | No | No | Yes | Yes | No |
| Speaker Forum | No | No | No | No | No |
| Consultant | No | No | Yes | Yes | No |
| Stocks | No | No | No | No | No |
| Royalties | No | No | No | No | No |
| Expert Testimony | No | No | No | Yes | No |
| Board Member | No | No | No | No | No |
| Patents | No | No | No | No | No |
| Personal Relationship | No | No | No | No | No |
REFERENCES:
- [1].Bauer K, Schwarzkopf L, Graessel E, Holle R (2014) A claims data-based comparison of comorbidity in individuals with and without dementia. BMC Geriatr 14, 10. doi: 10.1186/1471-2318-14-10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [2].Roe CM, Behrens MI, Xiong C, Miller JP, Morris JC (2005) Alzheimer disease and cancer. Neurology 64, 895–898. doi: 10.1212/01.WNL.0000152889.94785.51. [DOI] [PubMed] [Google Scholar]
- [3].Roe CM, Fitzpatrick AL, Xiong C, Sieh W, Kuller L, Miller JP, Williams MM, Kopan R, Behrens MI, Morris JC (2010) Cancer linked to Alzheimer disease but not vascular dementia. Neurology 74, 106–112. doi: 10.1212/WNL.0b013e3181c91873. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Musicco M, Adorni F, Di Santo S, Prinelli F, Pettenati C, Caltagirone C, Palmer K, Russo A (2013) Inverse occurrence of cancer and Alzheimer disease: a population-based incidence study Neurology 81, 322–328. doi: 10.1212/WNL.0b013e31829c5ec1. [DOI] [PubMed] [Google Scholar]
- [5].Driver JA, Beiser A, Au R, Kreger BE, Splansky GL, Kurth T, Kiel DP, Lu KP, Seshadri S, Wolf PA (2012) Inverse association between cancer and Alzheimer’s disease: results from the Framingham Heart Study. BMJ 344, e1442. doi: 10.1136/bmj.e1442. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].About the ADNI Study. Alzheimer’s Disease Neuroimaging Initiative (online). Available at: http://adni.loni.usc.edu/about/. Accessed October 19, 2018.
- [7].Petersen RC, Aisen PS, Beckett LA, Donohue MC, Gamst AC, Harvey DJ, Jack CR, Jagust WJ, Shaw LM, Toga AW, Trojanowski JQ, Weiner MW (2010) Alzheimer’s Disease Neuroimaging Initiative (ADNI): clinical characterization. Neurology 74, 201–209. doi: 10.1212/WNL.0b013e3181cb3e25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].McKhann GDD, Folstein M, Katzman R, Price D, Stadlan EM (1984) Clinical diagnosis of Alzheimer’s disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 34, 939–944. [DOI] [PubMed] [Google Scholar]
- [9].Petersen RC (2004) Mild cognitive impairment as a diagnostic entity. J Intern Med 256, 183–194. doi: 10.1111/j.1365-2796.2004.01388.x. [DOI] [PubMed] [Google Scholar]
- [10].Procedures manual. Alzheimer’s Disease Neuroimaging Initiative (online). Available at: http://adni.loni.usc.edu/wp-content/uploads/2010/09/ADNI_GeneralProceduresManual.pdf. Accessed October 19, 2018.
- [11].Study documents from the ADNI Study. Alzheimer’s Disease Neuroimaging Initiative (online). Available at: http://adni.loni.usc.edu/methods/documents/. Accessed October 19, 2018.
- [12].Clinical protocol. Alzheimer’s Disease Neuroimaging Initiative (online). Available at: http://adni.loni.usc.edu/wp-content/themes/freshnews-dev-v2/documents/clinical/ADNI-1_Protocol.pdf. Accessed October 19, 2018.
- [13].Friedl JEF (2006) Mastering Regular Expressions. 3rd ed. O’Reilly Media, Inc., Sebastopol, CA. [Google Scholar]
- [14].Nudelman KNH, Risacher SL, West JD, McDonald BC, Gao S, Saykin AJ (2014) Association of cancer history with Alzheimer’s disease onset and structural brain changes. Front Physiol 5, 423. doi: 10.3389/fphys.2014.00423. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Rosen WG, Mohs RC, Davis KL (1984) A new rating scale for Alzheimer’s Disease. Am J Psychiatry. 141, 1356–1364. doi: 10.1176/ajp.141.11.1356. [DOI] [PubMed] [Google Scholar]
- [16].Schneider LS, Sano M (2009) Current Alzheimer’s disease clinical trials: methods and placebo outcomes. Alzheimers Dement 5, 388–397. doi: 10.1016j.jalz.2009.07.038. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Brown H, Prescott R (2015) Applied Mixed Models in Medicine. Statistics in Practice. 3rd ed. Hoboken, NJ: John Wiley & Sons, Ltd, Hoboken, NJ. [Google Scholar]
- [18].Kennedy RE, Cutter GR, Wang G, Schneider LS (2016) Post hoc analyses of ApoE genotype-defined subgroups in clinical trials. J Alzheimers Dis 50, 1205–1215. doi: 10.3233/JAD-150847 [DOI] [PubMed] [Google Scholar]
- [19].Kennedy RE, Cutter GR, Schneider LS (2014) Effect of APOE genotype status on targeted clinical trials outcomes and efficiency in dementia and mild cognitive impairment resulting from Alzheimer’s disease. Alzheimers Dement 10, 349–359. doi: 10.1016/j.jalz.2013.03.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Schneider LS, Kennedy RE, Wang G, Cutter GR (2015) Differences in Alzheimer disease clinical trial outcomes based on age of the participants. Neurology 84, 1121–1127. doi: 10.1212/WNL.0000000000001376. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Kennedy RE, Cutter GR, Wang G, Schneider LS (2015) Using baseline cognitive severity for enriching Alzheimer’s disease clinical trials: How does Mini-Mental State Examination predict rate of change? Alzheimers Dement (NY) 1, 46–52. doi: 10.1016/j.trci.2015.03.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].Jean-Pierre P, McDonald BC (2016) Neuroepidemiology of cancer and treatment-related neurocognitive dysfunction in adult-onset cancer patients and survivors. Handb Clin Neurol 138, 297–309. doi: 10.1016/B978-0-12-802973-2.00017-3. [DOI] [PubMed] [Google Scholar]
- [23].Vannorsdall TD (2017) Cognitive changes related to cancer therapy. Med Clin North Am 101, 1115–1134. doi: 10.1016/j.mcna.2017.06.006. [DOI] [PubMed] [Google Scholar]
- [24].Janelsins MC, Kesler SR, Ahles TA, Morrow GR (2014) Prevalence, mechanisms, and management of cancer-related cognitive impairment. Int Rev Psychiatry 26, 102–113. doi: 10.3109/09540261.2013.864260. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [25].Hodgson KD, Hutchinson AD, Wilson CJ, Nettelbeck T (2013) A meta-analysis of the effects of chemotherapy on cognition in patients with cancer. Cancer Treat Rev 39, 297–304. doi: 10.1016/j.ctrv.2012.11.001. [DOI] [PubMed] [Google Scholar]
- [26].Fleming KC, Adams AC, Petersen RC (1995) Dementia: diagnosis and evaluation. Mayo Clin Proc 70, 1093–1107. doi: 10.1016/S0025-6196(11)64447-9. [DOI] [PubMed] [Google Scholar]
- [27].Huang AR, Strombotne KL, Mokyr HE (2018) Adolescent cognitive aptitudes and later-in-life Alzheimer disease and related disorders. JAMA Netw Open 1, e181726. doi: 10.1001/jamanetworkopen.2018.1726. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [28].Earle C (2017) Sensitivity and specificity of self-reported cancer history compared to cancer registry. J Clin Oncol 35, 234. doi: 10.1200/JCO.2017.35.8_suppl.234. [DOI] [Google Scholar]
- [29].Bergmann MM, Calle EE, Mervis CA, Miracle-McMahill HL, Thun MJ, Heath CW (1998) Validity of self-reported cancers in a prospective cohort study in comparison with data from state cancer registries. Am J Epidemiol 147, 556–562. Doi: 10.1093/oxfordjournals.aje.a009487. [DOI] [PubMed] [Google Scholar]
- [30].Cho S, Shin A, Song D, Park JK, Kim Y, Choi JY, Kang D, Lee JK (2017) Validity of self-reported cancer history in the health examinees (HEXA) study: A comparison of self-report and cancer registry records. Cancer Epidemiol 50, 16–21. doi: 10.1016/j.canep.2017.07.010. [DOI] [PubMed] [Google Scholar]
- [31].Kennedy RE, Cutter GR, Fowler ME, Schneider LS (2018) Association of concomitant use of cholinesterase inhibitors or memantine with cognitive decline in Alzheimer clinical trials: a meta-analysis. JAMA Netw Open 1, e184080. doi: 10.1001/jamanetworkopen.2018.4080. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [32].Schneider LS, Kennedy RE, Wang G, Cutter GR (2015) Differences in Alzheimer disease clinical trial outcomes based on age of the participants. Neurology 84, 1121–1127. doi: 10.1212/WNL.0000000000001376. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [33].Kennedy RE, Schneider LS, Cutter GR (2012) Biomarker positive and negative subjects in the ADNI cohort: clinical characterization. Curr Alzheimer Res 9, 1135–1141. doi: 10.2174/156720512804142976. [DOI] [PubMed] [Google Scholar]
- [34].Pourhassan Shamchi S, Zirakchian Zadeh M, Østergaard B, Kim J, Raynor WY, Khosravi M, Taghvaei R, Nielsen AL, Gerke O, Werner TJ, Holdgaard P, Abildgaard N, Revheim ME, Høilund-Carlsen PF, Alavi A (2020) Brain glucose metabolism in patients with newly diagnosed multiple myeloma significantly decreases after high-dose chemotherapy followed by autologous stem cell transplantation. Nucl Med Commun 41, 288–293. doi: 10.1097/MNM.0000000000001144. [DOI] [PubMed] [Google Scholar]
- [35].Inagaki M, Yoshikawa E, Matsuoka Y, Sugawara Y, Nakano T, Akechi T, Wada N, Imoto S, Murakami K, Uchitomi Y (2007) Smaller regional volumes of brain gray and white matter demonstrated in breast cancer survivors exposed to adjuvant chemotherapy. Cancer 109, 146–156. doi: 10.1002/cncr.22368. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.