Exploring the association between cancer and cognitive impairment in the Australian Imaging Biomarkers and Lifestyle (AIBL) study

An inverse association between cancer and Alzheimer’s disease (AD) has been demonstrated; however, the association between cancer and mild cognitive impairment (MCI), and the association between cancer and cognitive decline are yet to be clarified. The AIBL dataset was used to address these knowledge gaps. The crude and adjusted odds ratios for MCI/AD and cognitive decline were compared between participants with/without cancer (referred to as C+ and C− participants). A 37% reduction in odds for AD was observed in C+ participants compared to C− participants after adjusting for all confounders. The overall risk for MCI and AD in C+ participants was reduced by 27% and 31%, respectively. The odds of cognitive decline from MCI to AD was reduced by 59% in C+ participants after adjusting for all confounders. The risk of cognitive decline from MCI to AD was halved in C+ participants. The estimated mean change in Clinical Dementia Rating-Sum of boxes (CDR-SOB) score per year was 0.23 units/year higher in C− participants than in C+ participants. Overall, an inverse association between cancer and MCI/AD was observed in AIBL, which is in line with previous reports. Importantly, an inverse association between cancer and cognitive decline has also been identified.

strategies to intervene cognitive decline in AD.Last but not least, although the association between cancer and AD has been explored in European 6,13,[16][17][18] and American studies 2,3,[10][11][12]14,15,19 , and the influence of geographic and demographic factors on this association was noted, an Australian study is lacking.
To address these knowledge gaps, we leveraged longitudinal data from the Australian Imaging Biomarkers and Lifestyle (AIBL) study 21,22 and examined the risk for MCI/AD in participants with cancer and the risk for cognitive decline (i.e.cognitive unimpaired [CU] to MCI, CU to AD, and MCI to AD) due to cancer.Confounding factors were adjusted for or stratified.In addition, change in Clinical Dementia Rating Scale Sum of Boxes scores over time (ΔCDR-SOB, a surrogate for cognitive decline) was calculated and compared between participants with and without cancer.

Participants
The AIBL study was approved by the institutional ethics committees at St. Vincent's Health and the University of Melbourne, and all participants have provided written informed consent prior to study enrolment.All methods were performed in accordance with the Australian code for the responsible conduct of research.To date, 2854 participants have been recruited to AIBL.Longitudinal data was collected over an average 7.2-year period, with an overall participation rate of 84.9%.The selection process for participants in the current study is summarized in Fig. 1.Of the total 2854 AIBL participants, 718 participants were excluded from the study due to unknown cancer and/or cognitive status at enrolment.Of the included participants (n = 2136), 55.6% were female (n = 1187) and 44.4% were male (n = 949).The participants were categorized as CU, MCI, AD, and PRO (those who experienced progression in cognition category during the follow-up period).Cognitive assessment is detailed in the Clinical Diagnosis section below.

History of cancer (exposure)
History of cancer was self-reported at enrolment and the occurrence of cancer was self-reported during the follow-up period.Cancer status of the participants was categorized as (1) self-report of cancer at enrollment or during follow-up (C+); (2) no self-report of cancer at enrolment or during follow-up (C−).For participants with change in cognition category, only those who were diagnosed with cancer before cognition category change were included in the analysis.

Clinical diagnosis of MCI and AD (outcome)
The criteria used for clinical diagnosis of MCI and AD has been described in previous studies 21,22 .Cognitive status of the participants was determined by neurologists and neuropsychologists using amyloid PET imaging and neuropsychology tests.Participants who were identified as CU or MCI at enrolment but experienced cognition category change over the follow-up period (i.e.CU to MCI, CU to AD, and MCI to AD) were assigned to the PRO group.Participants who were identified as CU, MCI, or AD at enrolment and throughout the study were categorized into the CU, MCI, and AD groups.

Potential confounders
Sex, APOEε4 (carrier or noncarrier), smoking (never, former, current), alcohol (non/light/moderate/heavy drinker), and education were determined from self-report and medical records.In addition, a blood sample was taken for APOE genetic testing for participants who gave consent.Light, moderate, and heavy alcohol consumption was defined as < 3 days per week, 3 to 6 days per week, and daily consumption, respectively.Descriptive statistics for the potential confounders are provided in Supplementary Data Table S1.www.nature.com/scientificreports/546 C + participants, 367 (67.2%), 47 (8.6%), 73 (13.4%) and 59 (10.8%) belonged to the CU, MCI, AD, and PRO groups, respectively (Fig. 2B).The averaged age and distribution of sex for participants under each cognition group is also provided in Table 1.

The association (OR) between cancer and AD is significant but not for cancer and MCI
To measure the association between cancer and MCI/AD, the crude and adjusted ORs were calculated (  2, + All, column 2).In conclusion, an inverse association between cancer and AD was observed and remained significant even after adjusting for all confounders in this study.

The association (RR) between cancer and MCI/AD is moderated by sex and APOE ε4 allele type
The crude risk of MCI (RR 0.73 [0.54-0.99],p = 0.042) and AD (RR 0.69 [0.54-0.87],p = 0.001) was significantly reduced in C+ participants when compared to C− participants (Table 3).Stratum-specific RR for confounders   S2.In conclusion, a significant inverse association was observed between cancer and MCI/AD.However, the stratum-specific RR showed that the association between cancer and MCI/AD contained differences for each stratum of the confounder.The stratum-specific RR was not calculated for smoking (three subgroups), alcohol (four subgroups), education (six subgroups), or their combinations, given the small sample size resulting from stratification.

The association (RR) between cancer and cognitive decline is moderated by different PRO categories
The association between cancer and cognitive decline was examined in the PRO group (Table 4, column 1), and no significant association was observed from either crude or adjusted models.Different categories of PRO (i.e.CU to MCI, MCI to AD, and CU to AD) were therefore analyzed meticulously.Progression from CU to MCI occurred in 88 participants, of whom 29 (24.79%) were C+.Unadjusted and adjusted logistic regression models revealed a nonsignificant association between cancer and progression from CU to MCI (Table 4, column 2).A total of 41 participants experienced cognitive decline from CU to AD, of whom 15 (36.59%) were C+; however, no statistically significant association was observed (Table 4, column 3).The same analyses were performed for participants who progressed from MCI to AD (n = 95), of whom 15 (15.79%) were C+.The crude logistic regression model showed a 0.48-fold relative decrease (OR 0.48 [0.27-0.84],p = 0.010) in the progression from MCI to AD in C+ participants compared with C− participants.In adjusted logistic regression models, cancer was inversely associated with the progression from MCI to AD, after separate adjustment for each confounder (ORs 0.45-0.51).A similar association was also found when all confounders were accounted for (0.41 [0.21-0.82],p = 0.011), and all these odds ratios reached statistical significance (Table 4, column 4).

The association (RR) between cancer and progression from MCI to AD is significant, while that between cancer and progression from CU to MCI or CU to AD is nonsignificant
The risk reduction of cognitive decline in C+ compared to C− participants was not statistically significant (0.80 [0.61-1.05],p = 0.105) (Table 5, PRO).RRs in each cognitive decline category were therefore assessed.Only the association between cancer and progression from MCI to AD was significant but not for any other cognitive decline categories.A 0.5-fold reduction (0.50 [0.29-0.85],p = 0.008) in the risk of progression from MCI to AD www.nature.com/scientificreports/ was observed in C+ participants compared with C− participants (Table 5, MCI to AD).The stratum-specific RR was not calculated for sex (two subgroups), APOE ε4 allele type (two subgroups), smoking (three subgroups), alcohol (four subgroups), education (six subgroups), or their combinations due to small sample size resulting from subgrouping that were impractical for analysis.

The increase in CDR-SOB score over time is slower in participants with history of cancer than those without
The CDR-SOB is a global score regularly used in clinical and research settings to stage dementia severity 23 .Generally, more severe cognitive impairment was associated with a higher score.To understand the association between cancer and cognitive decline, the estimated mean change in CDR-SOB scores normalized by time (ΔCDR-SOB) was compared between C+ and C− participants (Fig. 3).The ΔCDR-SOB score was 0.34 and 0.57 units/year for C+ and C− participants, respectively (Table 6).

Discussion
The present study examined the association between cancer and MCI/AD in AIBL participants.We noted that C+ participants were less likely to develop AD than C− participants, which is consistent with previous studies 6,11,[16][17][18]24,25 . Our sudies have also revealed a nonsignificant association between cancer and MCI.It must be emphasized that studying the association between cancer and MCI is challenging as MCI can be caused by different conditions, including AD, Lewy body dementia, Parkinson's disease and even chemotherapy [26][27][28][29] .
In future studies, MCI should be carefully characterized to eliminate disease heterogeneity from analysis; for example, the brain levels of amyloid-beta (Aβ) can be assessed, so that the association between cancer and Aβ-positive MCI can be precisely investigated.
The strength of our study is that we employed a large longitudinal dataset from AIBL to study the association between cancer and cognitive decline, which has not yet been comprehensively studied.The cognition category for the participants was clearly defined in our longitudinal study, so that we can ensure the temporal sequence between exposure (i.e.cancer) and the outcome (i.e.change in cognition category) to avoid reverse causality.All participants were included in the study for data analysis, except for those who had no record for cancer history or cognitive status, which reduced the risk of survival bias.The average age and gender distribution of the current study was comparable to a community-based cohort study conducted by University of Kentucky Alzheimer's Disease Research Center (UK-ADRC) 30 .We showed that C+ participants were less likely to progress from MCI to AD compared with C− participants, while no significant association between cancer and cognitive decline from CU to MCI was observed.This observation has demonstrated the potential role of cancer on the progression of MCI to AD, but not on the development of MCI.The association between cancer and cognitive decline from CU to AD (n = 41) was found to be nonsignificant, which is not consistent with the association between cancer and change in ΔCDR-SOB score over time.Comparing the estimated mean change in ΔCDR-SOB score per year, it is noted that cognitive decline is slower in C+ participants than C− participants (n = 546 for C+, n = 1590 for C−).This discrepancy could be due to a much smaller sample size available for the former analysis approach.
In our study, there are several limitations that need to be considered, which may restrict the inference of a true association between exposure and outcome.History of cancer was self-reported by participants, and therefore recall bias is possible, especially in old participants.In addition, details of chemotherapy and types of cancer were not recorded.Therefore, an investigation of the association between cancer therapies or specific cancer types and subsequent development of MCI, AD, or PRO was not possible.A systematic review has revealed that various types of cancer can have distinct effects on the risk of subsequent cognitive decline.It suggests that the association between cancer and disease progression is not uniform across all cancer types 24 .Therefore, recording more details on cancer subtypes should be considered for future AIBL studies.In addition, the composition of our cohort was mainly Caucasians residing in Victoria (60%), and Western Australia (40%).This will limit the generalizability of our results to Australians residing in regions outside of our study area and Australians with diverse ethnic backgrounds.In addition, the AIBL study excludes participants who had cancer in the past 5 years from recruitment (excluding skin or in situ prostate cancer), which can potentially contribute to selection bias.
The inverse association between cancer and AD can possibly be explained by the underlying pathobiological processes.At molecular level, the TP53 gene encodes the p53 protein, which is a well-recognized tumor suppressor 31,32 .In most cancer cases, the p53 gene is mutated and loses its tumor suppressive effect 33 .The accumulation of Aβ and hyperphosphorylation of tau are the hallmarks of AD 34 .Aβ promotes p53-dependent neuronal apoptosis by activating the p53 promoter, which is believed to be responsible for neurological disorders, such as AD 35,36 .The hyperphosphorylation of tau is also p53-dependent 37 .Other oncogenic molecules may also be involved in the inverse relationship between cancer and AD.Transforming growth factor beta (TGFβ) overexpression upregulates the expression of amyloid precursor protein (APP) and contribute to Aβ accumulation, while producing anti-proliferative effect that may suppress cancer 38,39 .Tumor overexpress PD-L1 helps tumor survival by suppressing the immune system 40,41 .Interestingly, PD-L1 suppresses neuroinflammation and the associated AD pathology 42 .The mitochondria regulate cell survival or apoptosis via regulating the production of reactive oxygen species (ROS).Moderate ROS production promotes the cancer cells growth and proliferation, while high levels of ROS in AD promote neurodegeneration 43 .P-glycoprotein at the blood-brain barrier clears Aβ from the brain 44 , while P-glycoprotein plays an immunosurveillance role in cancer 45 .A hypothetical pathobiological interaction between cancer and AD is illustrated in Fig. 4.
In conclusion, we observed a negative association between cancer and MCI/AD in AIBL, which is consistent with previous studies.We also highlighted a lower incidence and prevalence of MCI/AD progression in participants with cancer.In addition, cancer is associated with cognitive decline (change of CDR-SOB score) at a slower rate.Further studies are required to investigate the pathobiology underlying the two diseases to explain their inverse association observed in epidemiological studies.

Figure 3 .
Figure 3. ΔCDR-SOB for AIBL participants with/without cancer.The ΔCDR-SOB was plotted for C− (without cancer, left boxplot) and C+ (with cancer, right boxplot) participants.The lowest ΔCDR-SOB for C− participants was −1 and the highest was 12.The lowest ΔCDR-SOB for C+ participants were −0.5 and the highest was 6.The mean ΔCDR-SOB for C− and C+ was 0.57 and 0.34 units/year, respectively.

Table 2 .
Odds of MCI & AD in participants with cancer.Numbers in the brackets represent 95% CI.Odds of

Table 3 .
Risk of MCI & AD in participants with cancer.Numbers in the brackets represent 95% CI.

Table 4 .
The association (OR) between cancer and cognitive decline.Numbers in the brackets represent 95% CI.

Table 5 .
The association (RR) between cancer and cognitive decline.

Table 6 .
The association between the CDR-SOB score over time and cancer history status.Mean change for ΔCDR-SOB between C+ and C− participants.