Epidemiological studies on the relationship between Alzheimer disease (AD) and cancer indicates that individuals with a diagnosis of AD have a lower incidence of cancer, and cancer survivors have a lower incidence of AD.1–5 Studies by Roe et al3 and Driver et al4 are 2 of several well-known prospective population-based analyses that have shown this inverse relationship between AD and cancer. In Roe et al’s3 study, patients with diagnosis of cancer had a 43% lower risk of developing AD, and patients with a diagnosis AD had a 69% lower risk of developing subsequent cancer. However, such findings could be limited to selective mortality; the incidence of AD in cancer patients may appear reduced due to the fact that cancer patients are more likely to die before AD onset. Driver et al4 sought to resolve this issue of selective mortality by investigating the relationship between cancer and AD using the data from the Framingham Heart Study and restricting their analysis to patients who survived to at least 80 years of age. They found that cancer survivors had a 33% decreased risk of subsequent AD diagnosis, with the lowest risk of AD in survivors of smoking-related cancers (74%), whereas patients with AD had a 61% decrease risk of developing cancer.4 In Ou et al’s5 nationwide population study, it was found that there was a decreased in the overall incidence of cancer in AD patients, especially lung cancers.
A growing body of evidence suggests that cancer and AD may have common genetic and molecular abnormalities to explain this inverse epidemiological relationship. Various studies have explored the possibility of common biological mechanisms to explain the inverse association.2,6,7 The link between cancer and AD appears to be related to cell cycle dysregulation.8 Proteins P53, Wnt, and Pin1 have been implicated in the cell cycle regulation of both disease processes.4,6,7,9–12 Cancer is characterized by uncontrolled proliferation of cells, whereas AD has a propensity toward apoptosis in neuronal cells. The Pin1 gene has a role in both cancer and AD through its regulation of the cell cycle and signaling process, response to DNA damage, and regulation of tau and β-amyloid (Aβ) precursor proteins.4,6,13,14 Pin1 is found to be underexpressed in brain tissues of AD patients.6,13,15,16 In contrast, Pin1 is found to be overexpressed in several cancer types.10,13,16,17 In Bao et al’s10 pathologic tissue analysis of several cancer types, colon, hepatocellular, prostate, cervical, lung, breast, and ovarian cancers overexpressed Pin1, whereas esophageal, pancreatic, stomach, and bladder cancers did not. Subsequent investigations have corroborated Bao and colleagues’ findings.18–23 Animal studies have showed that Pin1 knockout prevents the formation of oncogenes, but increases tau and Aβ-related pathologies.14,16,24–27
Our objective was to observe the relationship between AD and 10 cancer types using discharge data from a nationally representative database. Cancer types were further stratified based on Pin1 expression to investigate Pin1’s role in the relationship between cancer and AD.
MATERIALS AND METHODS
Data were obtained from the Nationwide Inpatient Sample (NIS) from 1999 to 2008 to examine the relationship between AD and cancer. The NIS represents a 20% stratified sample of US community hospitals from more than 1000 short-term and nonfederal hospitals across 37 states. It contains all-payer data on 8 million hospital inpatient stays in each year of the NIS. It captures discharge-level information on 1 primary, 14 secondary diagnosis, and 15 procedures using International Classification of Disease, 9th edition, clinical modification (ICD-9-CM) and the Clinical Classification Software (CCS) codes. Each data entry includes a unique identifier, demographic variables, discharge disposition, primary insurance payers, total hospital charges, and length of stay. The sampling strategy selects hospitals from State Inpatient Database according to a defined strata based on ownership, bed size, teaching status, urban or rural location, and region. All discharges from sampled hospitals for the calendar year are then selected for inclusion into NIS. To allow extrapolation for national estimates, both hospital and discharge weights are provided. Detailed information on the design of the NIS is available at http://.hcup-us.ahrq.gov.
The population study consisted of records indicating age at admission of 60 years and above with a primary diagnosis of cancer. To further analyze cancer hospitalizations, 10 cancer types that have previously demonstrated a relationship with AD were selected subsequent to literature review.5,28 NIS uses the ICD-9-CM and CCS codes, which were utilized to identify the following cancers: esophagus (CCS category 12), stomach (CCS category 13), colon (CCS category 14), pancreas (CCS category 17), lung (CCS category 19), breast (CCS category 24), cervix (CCS category 26), ovarian (CCS category 27), prostate (CCS category 29), and bladder (CCS category 32).29 After logistic regression analysis was performed to examine the relationship of cancer type and secondary diagnosis of AD, the correlation of Pin1 expression in these cancer types and AD was explored. Cancer types were separated in 2 groups according to their concentration level of Pin1. Bao et al10 demonstrated that colon, prostate, cervical, lung, breast, and ovarian cancers overexpressed Pin1 and thus, these cancers comprised the overexpressed Pin1 group. Esophagus, pancreas, stomach, and bladder cancers did not demonstrate elevated levels of Pin1 on pathologic tissue analysis by Bao et al10 and thus, these cancers comprised the underexpressed Pin1 group.
Main Outcome Measure
Secondary diagnosis of AD was the outcome measure in cancer patients and cancer patients without secondary diagnosis of AD diagnosis served as the reference group. The diagnosis was identified by ICD-9-CM code of 331.0.
Propensity Score Matching
A 1:1 case-control matching analysis based on propensity scores was used to match cases to controls based on the predicted probability of our main exposure, cancer. Propensity score matching is commonly used to reduce selection bias in observational studies. The Greedy matching techniques, developed by Parsons is the matching macro used in this study.30 First, the macro makes best matches, then it makes next-best matches; it follows this hierarchical sequence until no more matches can be made. The results are matched pair samples with closely matched individual pairs and balanced case and control groups. Matched control cases, which did not have diagnosis of cancer or AD and were greater than age 60, were selected for our reference group. The following covariates were used to calculate the propensity score, age, sex, race, primary payer, and hospital characteristics (location and region). For this study, the following races were of interest: whites, African Americans, Hispanics, and Asians or Pacific Islander. Approximately 25% of NIS discharges are missing race because a handful of states do not report that data. Records with missing values of race were excluded from this analysis. Primary payer included Medicare, Medicaid, private insurance, and other (self-pay, no charge, etc.).
A descriptive analysis on cancer admissions with and without a secondary diagnosis of AD was performed. We examined the prevalence of AD within each cancer type. We evaluated the 10 cancer types in 1 model creating a single dummy variable with 11 levels, which included the reference group. A logistic regression analysis was performed to evaluate the association between prevalence of AD in patients with a primary diagnosis of cancer and patients without a cancer diagnosis as reference. Crude and multivariate odds ratio (OR) and 95% confidence intervals of cancer patients with a secondary diagnosis of AD (CancerAD group) versus cancer patients without a secondary diagnosis of AD (CancerNAD group) were computed. The multivariate analysis included the adjusted potential confounders: age, sex, race, primary payer, and hospital characteristics (region/location). With multiple comparisons being made, we corrected for a Bonferroni adjustment in crude and multivariate analysis.
In addition, cancer types were grouped into Pin1 overexpressors and Pin1 underexpressors. Similarly, a logistic regression analysis was performed to evaluate the association between Pin1 expression in cancer and secondary diagnosis of AD. Crude and multivariate OR and 95% confidence intervals of Pin1 underexpression in cancer and secondary diagnosis of AD were computed. The Pin1 overexpression group served as the reference. The multivariate analysis was also adjusted for potential confounders age, sex, race, primary payer, and hospital characteristics (region and location).
Appropriate NIS weighted sampling was applied during both analyses. All data analyses were conducted using SAS (v 9.3; SAS Institute Inc, Cary, NC) and SPSS (v 20.0; IBM Corp, Armonk, NY).
As shown in Table 1, 32,019 patients with cancer and a secondary diagnosis of AD (CancerAD group) and 3,132,597 patients with cancer without a secondary diagnosis of AD (CancerNAD group) were hospitalized from 1999 to 2008. There were statistically significant differences between these 2 groups for all demographic measures. The mean ages at admission in CancerAD group and the CancerNAD group were 81.8±0.09 and 73.1±0.03, respectively. CancerAD group averaged 7.4±0.01 days in length of stay, whereas CancerNAD group averaged 6.8±0.01. There were more female (57.2%), white (81.9%), and Medicare insurance (91.0%) found in CancerAD group than CancerNAD group. Mortality was higher in the CancerNAD group in comparison to CancerAD group (4.0% vs. 3.3%, respectively). From 1999 to 2008, an estimated 3,188,854 patients were hospitalized without a primary diagnosis of cancer or AD, based on the 1 to 1 propensity score matching. The noncancer group had a mean age of 72.2±0.05 and a shorter length of stay 5.4±0.02. Among the noncancer control group, similar to cancer admissions, the most common sociodemographic categories were female, white, and Medicare insurance. Mortality for noncancer patients was the same as CancerNAD group at 4.0%. Although the most common geographic locations for hospitalizations were urban areas and the southern part of the United States, the proportion of cancer hospitalizations were higher in these areas compared with the noncancer controls.
Prevalence of Secondary Diagnosis of AD Within Cancer Discharges
From 1999 to 2008, overall, 1.0% of the discharged patients with a primary diagnosis of cancer had a secondary diagnosis of AD (Table 2). Discharged patients with bladder cancer had the highest prevalence of AD (1.5%), whereas discharged patients with prostate cancer had the lowest prevalence of AD (0.5%) (Table 2).
Logistic Regression Analysis
The crude analysis revealed a significantly reduced likelihood of secondary AD diagnosis among all cancer types at discharge, where patients with bladder cancer had the higher association with a secondary diagnosis of AD [OR: 0.74 (0.68 to 0.80)] and prostate cancer had the least association with secondary diagnosis of AD [OR: 0.26 (0.24 to 0.29)] (Table 3). Upon adjusting for age, sex, race, and hospital characteristics (region/location), the OR of secondary diagnosis of AD in cancer patients ranged from OR: 0.35 to 0.51. Discharged patients with prostate [crude OR: 0.26 (0.24 to 0.29); multivariate OR: 0.39 (0.35 to 0.43)], ovarian [crude OR: 0.38 (0.32 to 0.44); multivariate OR: 0.35 (0.30 to 0.41)], and lung [crude OR: 0.39 (0.36 to 0.41); multivariate OR: 0.41 (0.39 to 0.44)] cancers revealed the strongest inverse relationship with AD (Table 3). The likelihood of secondary diagnosis of AD in discharged patients with cancer was higher among cancers with underexpression of Pin1 when compared with cancers with an overexpressed Pin1 [crude OR: 1.37 (1.29 to 1.45), multivariate OR: 1.08 (1.02 to 1.14)] (Table 4). The odds of secondary AD diagnosis was significantly lower for all cancer types and with cancers having an overexpressed Pin1 in crude and multivariate analysis with P<0.001. After testing for the Bonferroni correction, there was no change in significance (P<0.001).
Using a nationally representative database, we were able to further establish the inverse relationship between cancer and AD. Our study demonstrated an inverse relationship between hospitalized patients with a primary diagnosis of 10 cancer types and secondary diagnosis of AD based on discharge summaries. Patients with primary diagnosis of each cancer type had a lower prevalence of a secondary diagnosis of AD when compared with matched controls (Table 2). Our cross-sectional case-control study yielded similar results to prior studies, which further strengthens this inverse relationship.2–5 Patients with a diagnosis of prostate, ovarian, and lung cancers were found to have the lowest incidence of secondary diagnosis of AD at discharge in our study, suggesting a strong inverse association (Table 3). These findings are not limited to selective mortality as we controlled for this bias by including discharge data of patient at least 60 years of age and who survived to 80 years old or greater. Most epidemiological studies on this topic have shown that males with cancer have a lower risk of AD and vice-versa, when compared with females.2–5 Interestingly, our study revealed that females with a primary diagnosis of cancer had a higher likelihood of a secondary diagnosis of AD when compared with males (Table 1). This discrepancy warrants further investigation.
Our study also showed that discharged patients with cancer types associated with Pin1 underexpression had a higher likelihood of a secondary AD diagnosis when compared with patients with cancer types that overexpressed Pin1, suggesting that Pin1 overexpression is associated with reduced prevalence of AD (Table 4). Animal studies have shown that Pin1 knockout prevents the formation of oncogenes and suppresses tumor and cell growth.14,16,20,24,25 Concurrently, an underexpressed Pin1 could accelerate neurodegeneration in an Alzheimer brain.26 Animal studies have demonstrated that higher levels of Pin1 in postnatal neurons reversed neurodegeneration.6,9,31 The loss of Pin1 function in animal studies causes tau and Aβ-related pathologies in an age-dependent manner, similar to changes seen in AD.6,9,31 Biochemically, Pin1 facilitates the isomerization and dephosphorylation of the pathogenic cis-tau confirmation to the physiological trans-tau confirmation.7,13,16,32 Thus, underexpression of Pin1 can lead to an increase in hyperphosphorylated tau and neurofibrillary tangles, which are pathogenic hallmarks of AD. In addition, Pin1 is involved in the isomerization and intracellular processing of Aβ proteins. Investigators have demonstrated that underexpression of Pin1 leads to reduced intracellular localization and processing, resulting in extracellular precipitation of Aβ, another pathogenic hallmark of AD.14,33–35 There seems to be a delicate balance in cellular Pin1 expression: overexpression will promote the development of cancer, whereas underexpression will lead to AD development. Prostate, lung, and ovarian cancers types overexpress Pin1.10 Our study demonstrated that these cancer types at discharge were associated with the lowest prevalence of secondary diagnosis of AD, which suggests that Pin1 overexpression in these cancer types may play a role in the strong inverse association with AD. Given Pin1’s direct relationship in both cancer and AD, it has received wide spread attention as potential target for therapy and should be investigated further.24,36–40 To the best of our knowledge, this is the first study to stratify cancer type per Pin1 expression and assessed its relationship with AD.
This type of research could be fraught with limitations. A disadvantage of this cross-sectional study is that it does not establish a cause and effect relationship between AD and cancer. Our study has characteristics of a case-control study where we compared patients with a primary diagnosis of cancer and secondary diagnosis of AD (CancerAD group) with patients with a primary diagnosis of cancer and without a secondary diagnosis of AD (CancerNAD group), relying on the accuracy of the database for selected ICD-9 codes. These codes are prone to error and cannot differentiate between late or early onset of AD. Furthermore, since cases are admission records and not unique individuals, there could be multiple admission records of the same individual, thus increasing or decreasing cases with secondary diagnosis of AD. In addition, in the minority population, the diagnosis of dementia has proven to be more challenging as language and cultural nuances add layers of complexity to the process of diagnosis. Even though the NIS is a nationwide representation, it may not be a true representation of incidence or prevalence of dementia in the general population.
Survival bias, or a decrease in incidence or under diagnoses of cancer, could be a cause of low incidence rate and, in turn, be the cause of a possible spurious inverse relationship. Few studies have looked at the survival bias of cancer patients dying before developing AD. To account for survival bias, Driver et al4 conducted a separate analysis with participants who survived to age 80 and still found an inverse relationship between cancer and probable AD, which is what we have done during our selection process. Despite these limitations, the power of NIS database was an important strength of our study and to the best of our knowledge; a sample of this size has not previously been used to evaluate the relationship between 10 cancer types and AD.
In addition, we grouped the 10 cancer types according to their Pin1 expression that was outlined in Bao et al10 pathologic study. This inherently assumes that the cancer types in our study have Pin1 overexpression or underexpression without tissue pathologic confirmation. Despite this limitation, several literature reports have confirmed Pin1 expression or underexpression in these cancer types, which are in agreement with Bao and colleagues initial discovery.18–23
We were able to demonstrate an inverse relationship between 10 cancer types and AD based on discharge data from the NIS database, which is consistent with several other epidemiologic studies.2–5 Patients with prostate, ovarian, and lung cancers types had a stronger inverse relationship with secondary diagnosis of AD than other cancer types. In addition, patients with cancer types that are Pin1 underexpressors were found to have a significantly higher likelihood of a secondary diagnosis of AD at discharge. Our findings suggest that Pin1 plays a role in the inverse relationship between AD and cancer, which could point to future interventions. Further analysis is necessary to accurately determine the significance of Pin1 and its role in the inverse relationship between AD and cancer. As the population ages rapidly and the number of those suffering from cancer and AD grows at an alarming rate, better understanding of the relationship between such common diseases and their molecular underpinnings as possible pathways to a cure can be pivotal.
1. Roe CM, Behrens MI, Xiong C, et al. Alzheimer disease
. Neurology. 2005;64:895–898.
2. Lanni C, Racchi M, Memo M, et al. p53 at the crossroads between cancer
and neurodegeneration. Free Radic Biol Med. 2012;52:1727–1733.
3. Roe CM, Fitzpatrick AL, Xiong C, et al. Cancer
linked to Alzheimer disease
but not vascular dementia
. Neurology. 2010;74:106–112.
4. Driver JA, Beiser A, Au R, et al. Inverse association between cancer
and Alzheimer’s disease: results from the Framingham Heart Study. BMJ. 2012;344:e1442.
5. Ou SM, Lee YJ, Hu YW, et al. Does Alzheimer’s disease protect against cancers? A nationwide population-based study. Neuroepidemiology. 2013;40:42–49.
6. Behrens MI, Lendon C, Roe CM. A common biological mechanism in cancer
and Alzheimer’s disease? Curr Alzheimer Res. 2009;6:196–204.
7. Lu KP. Pinning down cell signaling, cancer
and Alzheimer’s disease. Trends Biochem Sci. 2004;29:200–209.
8. Nagy Z. The dysregulation of the cell cycle and the diagnosis of Alzheimer’s disease. Biochim Biophys Acta. 2007;1772:402–408.
9. Driver JA, Lu KP. Pin1
: a new genetic link between Alzheimer’s disease, cancer
and aging. Curr Aging Sci. 2010;3:158–165.
10. Bao L, Kimzey A, Sauter G, et al. Prevalent overexpression of prolyl isomerase Pin1
in human cancers. Am J Pathol. 2004;164:1727–1737.
11. Nixon DW. The inverse relationship between cancer
and Alzheimer’s disease: a possible mechanism. Curr Alzheimer Res. 2017;14:883–893.
12. De Ferrari GV, Inestrosa NC. Wnt signaling function in Alzheimer’s disease. Brain Res Brain Res Rev. 2000;33:1–12.
13. Lu PJ, Wulf G, Zhou XZ, et al. The prolyl isomerase Pin1
restores the function of Alzheimer-associated phosphorylated tau protein. Nature. 1999;399:784–788.
14. Pastorino L, Sun A, Lu PJ, et al. The prolyl isomerase Pin1
regulates amyloid precursor protein processing and amyloid-beta production. Nature. 2006;440:528–534.
15. Mattson MP. Apoptosis in neurodegenerative disorders. Nat Rev Mol Cell Biol. 2000;1:120–129.
16. Liou YC, Sun A, Ryo A, et al. Role of the prolyl isomerase Pin1
in protecting against age-dependent neurodegeneration. Nature. 2003;424:556–561.
17. Ryan KM, Vousden KH. Cancer
: pinning a change on p53. Nature. 2002;419:795; 797.
18. Ryo A, Uemura H, Ishiguro H, et al. Stable suppression of tumorigenicity by Pin1
-targeted RNA interference in prostate cancer
. Clin Cancer
19. Wang XJ, Xu B, Mullins AB, et al. Conformationally locked isostere of phosphoSer-cis-Pro inhibits Pin1
23-fold better than phosphoSer-trans-Pro isostere. J Am Chem Soc. 2004;126:15533–15542.
20. Wulf G, Ryo A, Liou YC, et al. The prolyl isomerase Pin1
in breast development and cancer
. Breast Cancer
21. Li H, Shen H, Xu Q, et al. Expression of Pin1
and Ki67 in cervical cancer
and their significance. J Huazhong Univ Sci Technolog Med Sci. 2006;26:120–122.
22. Wang H, Zhang J, Feng W, et al. PIN1
gene overexpression and beta-catenin gene mutation/expression in hepatocellular carcinoma and their significance. J Huazhong Univ Sci Technolog Med Sci. 2007;27:54–57.
23. Tan X, Zhou F, Wan J, et al. Pin1
expression contributes to lung cancer
: prognosis and carcinogenesis. Cancer
Biol Ther. 2010;9:111–119.
24. Balastik M, Lim J, Pastorino L, et al. Pin1
in Alzheimer’s disease: multiple substrates, one regulatory mechanism? Biochim Biophys Acta. 2007;1772:422–429.
25. Suizu F, Ryo A, Wulf G, et al. Pin1
regulates centrosome duplication, and its overexpression induces centrosome amplification, chromosome instability, and oncogenesis. Mol Cell Biol. 2006;26:1463–1479.
26. Segat L, Pontillo A, Annoni G, et al. PIN1
promoter polymorphisms are associated with Alzheimer’s disease. Neurobiol Aging. 2007;28:69–74.
27. Wulf G, Garg P, Liou YC, et al. Modeling breast cancer
in vivo and ex vivo reveals an essential role of Pin1
in tumorigenesis. EMBO J. 2004;23:3397–3407.
28. Driver JA, Logroscino G, Buring JE, et al. A prospective cohort study of cancer
incidence following the diagnosis of Parkinson’s disease. Cancer
Epidemiol Biomarkers Prev. 2007;16:1260–1265.
29. HCUP Clinical Classifications Software (CCS) for ICD-9-CM. Healthcare Cost and Utilization Project (HCUP). 1999-2008. Agency for Healthcare Research and Quality, Rockville, MD. 2017. Available at: http://.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp
. Accessed June, 2017.
30. Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques (SAS SUGI paper pp. 214–226). Proceedings of the 26th annual SAS Users’ Group International Conference
. Cary, NC: SAS Institute; 2001. Available at: http://www2/sas.com/proceedings/sugi26/p214-26.pdf
. Accessed September, 2016.
31. Lim J, Balastik M, Lee TH, et al. Pin1
has opposite effects on wild-type and P301L tau stability and tauopathy. J Clin Invest. 2008;118:1877–1889.
32. Zhou XZ, Kops O, Werner A, et al. Pin1
-dependent prolyl isomerization regulates dephosphorylation of Cdc25C and tau proteins. Mol Cell. 2000;6:873–883.
33. Pastorino L, Ma SL, Balastik M, et al. Alzheimer’s disease-related loss of Pin1
function influences the intracellular localization and the processing of AbetaPP. J Alzheimers Dis. 2012;30:277–297.
34. Blair LJ, Baker JD, Sabbagh JJ, et al. The emerging role of peptidyl-prolyl isomerase chaperones in tau oligomerization, amyloid processing, and Alzheimer’s disease. J Neurochem. 2015;133:1–13.
35. Lloret A, Fuchsberger T, Giraldo E, et al. Molecular mechanisms linking amyloid beta toxicity and tau hyperphosphorylation in Alzheimers disease. Free Radic Biol Med. 2015;83:186–191.
36. Kondo A, Shahpasand K, Mannix R, et al. Antibody against early driver of neurodegeneration cis P-tau blocks brain injury and tauopathy. Nature. 2015;523:431–436.
37. Lu KP, Kondo A, Albayram O, et al. Potential of the antibody against cis-phosphorylated tau in the early diagnosis, treatment, and prevention of Alzheimer disease
and brain injury. JAMA Neurol. 2016;73:1356–1362.
38. Snyder HM, Ahles T, Calderwood S, et al. Exploring the nexus of Alzheimer’s disease and related dementias with cancer
therapies: a convening of the Alzheimer’s Association & Alzheimer’s Drug Discovery Foundation. Alzheimers Dement. 2017;13:267–273.
39. Zhou XZ, Lu KP. The isomerase PIN1
controls numerous cancer
-driving pathways and is a unique drug target. Nat Rev Cancer
40. Neve RL, McPhie DL. The cell cycle as a therapeutic target for Alzheimer’s disease. Pharmacol Ther. 2006;111:99–113.