Lower Risk for Dementia Following Adult Tetanus, Diphtheria, and Pertussis (Tdap) Vaccination

Abstract

Background

Adult vaccinations may reduce risk for dementia. However, it has not been established whether tetanus, diphtheria, pertussis (Tdap) vaccination is associated with incident dementia.

Methods

Hypotheses were tested in a Veterans Health Affairs (VHA) cohort and replicated in a MarketScan medical claims cohort. Patients were at least 65 years of age and free of dementia for 2 years prior to index date. Patients either had or did not have a Tdap vaccination by the start of either of the 2 index periods (2011 or 2012). Follow-up continued through 2018. Controls had no Tdap vaccination for the duration of follow-up. Confounding was controlled using entropy balancing. Competing risk (VHA) and Cox proportional hazard (MarketScan) models estimated the association between Tdap vaccination and incident dementia in all patients and age subgroups (65–69, 70–74, and ≥75 years).

Results

VHA patients were, on average, 75.6 (SD ± 7.5) years of age, 4% female, and 91.2% were White. MarketScan patients were 69.8 (SD ± 5.6) years of age, on average and 65.4% were female. After controlling for confounding, patients with, compared to without, Tdap vaccination had a significantly lower risk for dementia in both cohorts (VHA: hazard ratio [HR] = 0.58; 95% confidence interval [CI]:0.54–0.63 and MarketScan: HR = 0.58; 95% CI:0.48–0.70).

Conclusions

Tdap vaccination was associated with a 42% lower dementia risk in 2 cohorts with different clinical and sociodemographic characteristics. Several vaccine types are linked to decreased dementia risk, suggesting that these associations are due to nonspecific effects on inflammation rather than vaccine-induced pathogen-specific protective effects.

Cohort, Dementia, Epidemiology, Tdap, Vaccination

Substantial evidence supports the conclusion that infections contribute to worsening cognitive impairment and incident dementia (1–4). Although the mechanisms are not firmly established, inflammation related to infection may generate neurotoxic molecules leading to neurodegeneration (2). Others suggest the mechanism is due to neurotropic viral, fungal, and bacterial infections that directly cause neurotoxic effects (4,5).

A growing number of studies have found vaccinations are associated with a lower risk of dementia. Two studies of patient cohorts from the Taiwan National Health Insurance database, one with chronic kidney disease and another with chronic obstructive pulmonary disease, reported reduced incidence for dementia following influenza vaccination (6,7). More recently, herpes zoster vaccination was found to be associated with a 30%–35% lower risk of dementia, and this association remained after accounting for the contribution of herpes zoster infection and antiviral therapy (8). Studies of the link between Tdap vaccination and incident dementia are limited to 2 longitudinal cohorts in which vaccination was measured via self-report. Cognitively intact participants in the longitudinal Manitoba Study of Health and Aging who reported a past influenza, polio, tetanus, or diphtheria vaccination at baseline had a significantly reduced risk of cognitive impairment at 5-year follow-up (9). In a similar design, Verreault et al.’s (10) results from the Canadian Study of Health and Aging revealed that self-reported history of diphtheria or tetanus vaccination was associated with a 60% lower risk of Alzheimer’s disease (AD) after adjusting for family history of dementia and other risk factors. However, both studies lacked measures of preventive health care or socioeconomic status which could explain the vaccination effect.

Among the infectious disease—dementia hypotheses, Bordetella pertussis infection is particularly intriguing because it colonizes the nasopharynx and has access to the central nervous system via co-located olfactory nerves (3). Yet there are no studies of pertussis, that is, adult Tetanus, Diphtheria, Pertussis (Tdap) vaccination and incident dementia, which have adequately controlled for use of preventive health services, current socioeconomic status, and other confounders while using objective measures of vaccine receipt. The current study was designed to first determine if individuals with an objective history of Tdap vaccination, compared to those remaining free of Tdap vaccination, had lower dementia incidence in Veterans Health Affairs (VHA) patients 65 years and older, controlling for significant confounders. Second, we determined if this association differed by age groups (65–69, 70–74, and ≥75 years). Lastly, we replicated analysis in a private sector claims database.

Method

De-identified VHA administrative medical record data from fiscal years (FY) FY09–FY19 (October 1, 2008–September 30, 2019) record data were used to create the primary study cohort. Medical records include ICD-9-CM and ICD-10-CM diagnoses codes, vaccinations, medication fills, laboratory results, vital signs, and demographic data. Medicare claims and Part-D pharmacy claims are linked to VHA administrative data and used to supplement information on diagnosis codes, laboratory values, and prescription fills.

The replication cohort was created from IBM MarketScan Commercial Claims and Medicare Supplemental databases from January 1, 2009 to December 31, 2018 (calendar year [CY] 2009–CY2018). The IBM MarketScan data contained de-identified patient-level health care claims related to inpatient and ambulatory care encounters at academic and nonacademic health care systems across the United States. Encounters could occur with any provider type. Patients with private and federal health insurance are included in the claims data.

Detailed variable definitions are shown in Supplementary Table 1. This research was reviewed by the Saint Louis University IRB and deemed nonhuman subjects research because data were de-identified.

Eligibility

Base samples used to create this study’s cohorts included patients with at least 3 well visits during the observation period who were 50 years and older at first well visit. Requiring frequent well visits limited confounding related to the correlation between the use of preventive health care and receipt of appropriate vaccinations (11). Two index dates were defined: FY2011 and 2012 for VHA (October 1, 2010 and October 1, 2011), allowing for 8–9 years of possible follow-up and CY2011 and 2012 for MarketScan (January 1, 2011 and January 1, 2012), allowing for 7–8 years of possible follow-up time. The choice of index dates was to ensure that there were enough observation years for dementia to occur. The addition of a second index date was to ensure we captured more vaccinations for our analytic cohort. New FDA and Advisory Committee on Immunization Practices recommendations expanded the use of Adecel and Boostrix in the second index dates in both cohorts.

Patients were at least 65 years of age at index and had at least one outpatient claim in the 2 years prior to index. For the 2 years leading up to index, patients were free of dementia diagnoses, any dementia treatment medications, and conditions that lead to cognitive impairment such as intracranial injury. The list of dementia diagnoses and medications excluded in the 2 years prior to index are given in Supplementary Table 1. Patients must have had more than 90 days follow-up after index and following a per-protocol approach, those patients without Tdap vaccination must have remained free of vaccination for the entirety of follow-up. Patients who did not meet eligibility in 2011 were used to sample for the 2012 index date.

After applying eligibility criteria and removing 2679 patients with missing marital status, 69 missing race, and 1 missing gender, there were 122 946 eligible VHA patients. There were 174 053 eligible MarketScan patients. The sampling process is shown in Supplementary Figures 1A and B. The left-hand columns show eligibility applied to the first index date; patients not eligible on the first index date could have been eligible on the second index date and are shown in the right-hand column of sampling figures. Detailed variable definitions for exposure, outcome, and covariates are given in Supplementary Table 1.

Exposure

Tdap vaccination was defined by CPT codes 90701 or 90715 and by product names Adacel and Boostrix. Vaccination status was defined at index date; patients with a Tdap vaccination must have had a CPT code or pharmacy record by index date, that is, the start of follow-up. Patients without a Tdap vaccination remained free of a CPT code or pharmacy record for Tdap vaccination until the end of follow-up.

Outcome

Incident dementia was defined by the presence of 2 or more ICD-9/ICD-10 codes for dementias on separate days within the same 12-month period. The date of onset was the first of the 2 codes. We have used this algorithm in other studies of incident dementia (12,13) as it has shown to have the best agreement with medical record Mini-Mental Status Examination and Saint Louis University Mental Status scores (14).

Follow-up time

Follow-up time was defined as months from index date to dementia or censoring. VHA patients were censored at the last available VHA health encounter or Medicare claim or death. MarketScan patients were censored at last available inpatient or outpatient claim. Mortality data were not available in MarketScan.

Covariates

Covariates, selected due to documented associations with risk for dementia or vaccination (15,16), were measured from the start of the observation period to index date. Covariates were the same in both cohorts with the exception of race, marital status, and VHA health insurance which were not available in MarketScan data. We controlled for demographics, geographic region, health services utilization, comorbid conditions, and medications.

Demographic data included age, race (VHA only), gender, and marital status (VHA only). Because vaccination rates vary across the United States, we controlled for region of the country (Northeast, Northcentral, South, West, and Unknown) in which patients lived. To address detection bias, we controlled for the number of well visits prior to index and the volume of total health care utilization. The latter was based on the distribution of the mean number of health care encounters prior to index, and the top 25th percentile was used to categorize high health care utilization (high utilization) versus not high.

Comorbid physical and psychiatric conditions were defined by ICD-9 and ICD-10 codes and included type 2 diabetes, obesity, hypertension, stroke, ischemic heart disease, congestive heart failure, atrial fibrillation, asthma, chronic obstructive pulmonary disease, traumatic brain injury, vitamin B12 deficiency, depression, anxiety disorder, nicotine dependence, and alcohol or drug abuse/dependence.

We controlled for sustained use, defined by 2 or more fills, of the following medications prior to index: anticholinergics, nonsteroidal anti-inflammatory drugs (NSAIDs), statins, steroids, antivirals, metformin, and sulfonylurea.

Analytic Approach

Receipt of vaccination is not random, therefore we reweighted data via entropy balancing (e-balance) to balance patient characteristics between those who did and did not receive the vaccine (17,18). This method was chosen over another commonly used method (propensity scores and inverse probability of treatment weighting) because better balance can be achieved without relying on correctly specifying the propensity score model. The e-balance method, in contrast to propensity score weighting, reweights the control (or nonvaccine) cohort by deriving weights so that specified covariate moments (eg, mean and standard deviation) will match the vaccinated cohort. Balance in the weighted sample was evaluated using the standardized mean difference percent (SMD% = 100 × SMD). Well-balanced covariates have an SMD of less than 10% (19).

Differences in patient characteristics between vaccinated and unvaccinated groups were first measured using unweighted bivariable analyses and chi-squared tests. The SMD% measured the effect size for differences in patient characteristics before and after e-balance weighting. In VHA data, competing risk survival regression models (20) were used to control for bias associated with a preceding, competing event (eg, death) that would prevent observing incident dementia diagnoses. Cox proportional hazard models were used to measure the association between vaccination and incident dementia in MarketScan data which did not have data on mortality.

In both cohorts, measures of association were expressed as hazard ratios (HRs) and 95% confidence intervals (CIs). Separate models were computed in unweighted and weighted data for the entire cohort and stratified by age groups (65–69, 70–74, and ≥75 years). Effect modification by age group was assessed with an age group by vaccination status interaction term in models using the entire cohort. Weighted models used robust, sandwich-type variance estimators to calculate CIs and p values (19).

The proportional hazards assumption was tested in all models by computing a time-dependent interaction term of Tdap vaccination and log (follow-up time). The assumption was met in all models (p > .05). Stata 16 (StataCorp, College Station, TX) was used to apply e-balance methods. All main analyses were performed with SAS v9.4 (SAS Institute, Cary, NC) at a two-tailed alpha = 0.05.

To determine whether unmeasured confounding may explain our results, we computed the e-value (21). The e-value is the magnitude of association between an unmeasured confounder and the exposure, and separately with the outcome, which would account for the association between Tdap and incident dementia.

Sensitivity Analysis

Mechanisms leading to AD versus other types of dementia may differ. Therefore, a sensitivity analysis was conducted to measure the association of Tdap and incident AD in overall samples to elucidate whether Tdap is associated with a specific type of dementia, rather than any dementia. See Supplementary Table 1 for AD definition.

Results

VHA and MarketScan cohort characteristics are given in Table 1. VHA patients were, on average, 75.6 (SD ± 7.5) years of age, 4% were female, 91.2% were White race, and 70% were married. MarketScan patients were 69.8 (SD ± 5.6) years of age and 65.4% were female. Physical and psychiatric conditions were more prevalent in the VHA cohort. Antihypertensive medications, statins, steroids, metformin, and sulfonylurea use were more prevalent in the VHA cohort.

Table 1.

Characteristics (%) of VHA and MarketScan Patient Cohorts

Covariates	VHA (n = 122 946)	MarketScan (n = 174 053)
Index fiscal year
2011	104 725 (85.2)	114 926 (66.0)
2012	18 221 (14.8)	59 127 (34.0)
Sociodemographic-related
Age, mean (±SD)	75.6 (±7.5)	69.8 (±5.6)
Age category
65–69	34 629 (28.2)	107 498 (61.8)
70–74	21 442 (17.4)	33 585 (19.3)
≥75	66 875 (54.4)	32 970 (18.9)
Female gender	4860 (4.0)	113 907 (65.4)
Race
White	112 108 (91.2)	–
Black	9502 (7.7)	–
Other	1336 (1.1)	–
Married	86 110 (70.0)	–
VHA only insurance	23 318 (19.0)	–
Region
Northeast	14 647 (11.9)	65 677 (37.7)
Northcentral	52 021 (42.3)	32 563 (18.7)
South	40 082 (32.6)	41 264 (23.7)
West	12 890 (10.5)	29 959 (17.2)
Unknown	3306 (2.7)	4590 (2.6)
High health care utilization	31 353 (25.5)	43 516 (25.0)
Number of well visits, mean (±SD)	1.1 (±2.5)	1.6 (±1.0)
Number of well visits, category
0	72 479 (59.0)	23 102 (13.3)
1–2	34 246 (27.9)	126 708 (72.8)
≥3	16 221 (13.2)	24 243 (13.9)
Comorbidities
Type II diabetes	40 482 (32.9)	25 662 (14.7)
Obesity	41 555 (33.8)	9258 (5.3)
Hypertension	101 821 (82.8)	103 872 (59.7)
Stroke	5974 (4.9)	2965 (1.7)
Ischemic heart disease	51 819 (42.2)	25 164 (14.5)
Congestive heart failure	18 430 (15.0)	5082 (2.9)
Atrial fibrillation	20 262 (16.5)	9300 (5.3)
Asthma	9253 (7.5)	12 621 (7.3)
COPD	25 609 (20.8)	10 198 (5.9)
Traumatic brain injury	3355 (2.7)	2921 (1.7)
Vitamin B12 deficiency	6876 (5.6)	4297 (2.5)
Depression	11 261 (9.2)	6188 (3.6)
Anxiety disorder*	9813 (8.0)	4956 (2.9)
Nicotine dependence	29 132 (23.7)	7614 (4.4)
Alcohol abuse/dependence	4933 (4.0)	717 (0.4)
Drug abuse/dependence	1415 (1.2)	224 (0.1)
Medications†
Anticholinergics	15 537 (12.6)	19 626 (11.3)
NSAIDS	16 112 (13.1)	16 089 (9.2)
Antihypertensives	88 017 (71.6)	84 663 (48.6)
Statins	71 336 (58.0)	65 658 (37.7)
Steroids	12 679 (10.3)	9378 (5.4)
Antivirals	1407 (1.1)	3316 (1.9)
Metformin	16 195 (13.2)	11 905 (6.8)
Sulfonylurea	15 376 (12.5)	5825 (3.4)

Note: COPD = chronic obstructive pulmonary disease; GAD = generalized anxiety disorder; NOS = not otherwise specified; NSAIDs = nonsteroidal anti-inflammatory drugs; OCD = obsessive compulsive disorder; VHA = Veterans Health Affairs.

*Anxiety disorders = panic disorder, OCD, social phobia, GAD, Anxiety NOS.

^†Medications = sustained use prior to index (at least 2 fills in a 6-month period).

In the VHA cohort, 7.8% (n = 9608) received a Tdap vaccination and 10.3% (n = 17 872) of MarketScan patients had a Tdap vaccination. In both cohorts, patients of at least 75 years were less common among the vaccinated group while those of 65–69 years were more prevalent among those who received a Tdap vaccination. Race (VHA only) was significantly (p = .021) associated with vaccination status but the effect size was small (SMD% <10). Similar small effects were observed for the association between marital status and vaccination in VHA patients. Those with only VHA health insurance were significantly more common among vaccinated patients, (p < .0001, SMD% = 10.1). In both cohorts, the geographic region was significantly (p < .0001) associated with vaccination.

In the VHA cohort, obesity was significantly more common in vaccinated versus unvaccinated patients (p < .0001, SMD% >10). Ischemic heart disease and atrial fibrillation were significantly more prevalent among unvaccinated versus vaccinated patients (p < .0001, SMD% >10). Depression, anxiety disorder, and nicotine dependence were significantly more common in the vaccinated patients (p < .0001, SDM% >10). The prevalence of anticholinergic and NSAID use was significantly (p < .0001) more prevalent among vaccinated patients in the VHA (p < .0001, SMD% >10). Among MarketScan patients, there were statistically significant differences in the prevalence of comorbid physical and psychiatric conditions and medication use by vaccination status, but these differences were small (SDM% <10; Table 2).

Table 2.

Characteristics (%) of 65 Years or Older Patients by Tdap Vaccine Status, VHA and MarketScan Cohorts

	VHA Patient Cohort (n = 122 946)				MarketScan Patient Cohort (n = 174 053)
Covariates	No Tdap Vaccine (n = 113 338)	Tdap Vaccine (n = 9608)	Unwtd p Value	Unwtd SMD%	No Tdap Vaccine (n = 156 181)	Tdap Vaccine (n = 17 872)	Unwtd p Value	Unwtd SMD%
Index fiscal year
2011	101 965 (90.0)	2760 (28.7)	<.0001	−159.45	108 909 (69.7)	6017 (33.7)	<.0001	−77.4
2012	11 373 (10.0)	6848 (71.3)		159.45	47 272 (30.3)	11 855 (66.3)		77.4
Sociodemographic-related
Age category
65–69	30 047 (26.5)	4582 (47.7)		44.9	94 268 (60.4)	13 230 (74.0)	<.0001	29.4
70–74	19 903 (17.6)	1539 (16.0)	<.0001	−4.1	30 899 (19.8)	2686 (15.0)		−12.6
≥75	63 388 (55.9)	3487 (36.3)		−40.2	31 014 (19.9)	1956 (10.9)		−24.9
Female gender	4377 (3.9)	483 (5.0)	<.0001	5.7	103 116 (66.0)	10 791 (60.4)	<.0001	−11.7
Race
White	103 420 (91.2)	8688 (90.4)		−2.9	—	—	—	—
Black	8691 (7.7)	811 (8.4)	.021	2.8	—	—	—	—
Other	1227 (1.1)	109 (1.1)		0.5	—	—	—	—
Married	79 774 (70.4)	6336 (65.9)	<.0001	−9.5	—	—	—	—
VHA only insurance	21 134 (18.6)	2184 (22.7)	<.0001	10.1	—	—	—	—
Region
Northeast	13 059 (11.5)	1588 (16.5)		14.4	60 876 (39.0)	4801 (26.9)		−26.0
Northcentral	47 898 (42.3)	4123 (42.9)	<.0001	1.3	28 828 (18.5)	3735 (20.9)	<.0001	6.2
South	37 868 (33.4)	2214 (23.0)		−23.2	37 706 (24.1)	3558 (19.9)		−10.2
West	11 274 (9.9)	1616 (16.8)		20.3	24 647 (15.8)	5312 (29.7)		33.7
Unknown	3239 (2.9)	67 (0.7)		−16.4	4124 (2.6)	466 (2.6)		−0.2
High health care utilization	27 832 (24.6)	3521 (36.6)	<.0001	26.5	39 291 (25.2)	4225 (23.6)	<.0001	−3.5
Number of well visits, category
0	67 212 (59.3)	5267 (54.8)		−9.1	22 093 (14.2)	1009 (5.7)		−28.8
1–2	31 672 (27.9)	2574 (26.8)	<.0001	−2.6	115 421 (73.9)	11 287 (63.1)	<.0001	−23.3
≥3	14 454 (12.8)	1767 (18.4)		15.6	18 667 (11.9)	5576 (31.2)		48.1
Comorbidities
Type II diabetes	37 336 (32.9)	3146 (32.7)	.691	−0.4	23 202 (14.9)	2460 (13.8)	<.0001	−3.1
Obesity	37 367 (33.0)	4188 (43.6)	<.0001	22.0	7962 (5.1)	1296 (7.3)	<.0001	9.0
Hypertension	93 943 (82.9)	7878 (82.0)	.026	−2.4	93 665 (60.0)	10 207 (57.1)	<.0001	−5.8
Stroke	5599 (4.9)	375 (3.9)	<.0001	−5.1	2618 (1.7)	347 (1.9)	.009	2.0
Ischemic heart disease	48 272 (42.6)	3547 (36.9)	<.0001	−11.6	22 850 (14.6)	2296 (12.8)	<.0001	−5.2
Congestive heart failure	17 279 (15.3)	1151 (12.0)	<.0001	−9.5	4653 (3.0)	429 (2.4)	<.0001	−3.6
Atrial fibrillation	19 031 (16.8)	1231 (12.8)	<.0001	−11.2	8377 (5.4)	923 (5.2)	.262	−0.9
Asthma	8515 (7.5)	738 (7.7)	.548	0.6	11 055 (7.1)	1566 (8.8)	<.0001	6.2
COPD	23 641 (20.9)	1968 (20.5)	.384	−0.9	9250 (5.9)	948 (5.3)	.001	−2.7
Traumatic brain injury	3104 (2.7)	251 (2.6)	.466	−0.8	2519 (1.6)	402 (2.3)	<.0001	4.6
Vitamin B12 deficiency	6389 (5.6)	487 (5.1)	.020	−2.5	3878 (2.5)	419 (2.3)	.258	−0.9
Depression	10 055 (8.9)	1206 (12.6)	<.0001	11.9	5305 (3.4)	883 (4.9)	<.0001	7.7
Anxiety disorder*	8635 (7.6)	1178 (12.3)	<.0001	15.6	4328 (2.8)	628 (3.5)	<.0001	4.3
Nicotine dependence	26 344 (23.2)	2788 (29.0)	<.0001	13.2	6719 (4.3)	895 (5.0)	<.0001	3.4
Alcohol abuse/dependence	4367 (3.9)	566 (5.9)	<.0001	9.5	600 (0.4)	117 (0.7)	<.0001	3.8
Drug abuse/dependence	1246 (1.1)	169 (1.8)	<.0001	5.6	182 (0.1)	42 (0.2)	<.0001	2.8
Medications†
Anticholinergics	13 931 (12.3)	1606 (16.7)	<.0001	12.6	17 445 (11.2)	2181 (12.2)	<.0001	3.2
NSAIDS	14 301 (12.6)	1811 (18.9)	<.0001	17.2	14 178 (9.1)	1911 (10.7)	<.0001	5.4
Antihypertensives	80 857 (71.3)	7160 (74.5)	<.0001	7.2	76 228 (48.8)	8435 (47.2)	<.0001	−3.2
Statins	65 333 (57.6)	6003 (62.5)	<.0001	9.9	58 634 (37.5)	7024 (39.3)	<.0001	3.6
Steroids	11 529 (10.2)	1150 (12.0)	<.0001	5.7	8322 (5.3)	1056 (5.9)	.001	2.5
Antivirals	1222 (1.1)	185 (1.9)	<.0001	7.0	2767 (1.8)	549 (3.1)	<.0001	8.5
Metformin	14 582 (12.9)	1613 (16.8)	<.0001	11.1	10 723 (6.9)	1182 (6.6)	.206	−1.0
Sulfonylurea	14 039 (12.4)	1337 (13.9)	<.0001	4.5	5316 (3.4)	509 (2.9)	<.0001	−3.2

Note: VHA = Veterans Health Affairs; NSAIDs = nonsteroidal anti-inflammatory drugs; COPD = chronic obstructive pulmonary disease.

*Anxiety disorders = panic disorder, OCD, social phobia, GAD, anxiety NOS.

^†Medications = sustained use prior to index (at least 2 fills in a 6-month period).

E-balancing effectively balanced differences in the distribution of covariates by vaccination status. As given in Supplementary Tables 2 and 3, for both VHA and Marketscan cohorts as a whole and within each age group, there were no meaningful differences (SMD% <10) in the distribution of covariates between patients who did compare to those who did not receive a Tdap vaccination.

The overall median follow-up time in the VHA was 95 months (interquartile range [IQR]: 54–106). VHA patients without Tdap vaccination had a median follow-up time of 94 months (IQR: 53–106) and those without Tdap vaccination had a median follow-up time of 95 months (IQR: 79–96). In the MarketScan cohort, the median follow-up time overall, and for both no Tdap and Tdap groups, was 36 months (IQR: 24–60). Cumulative incidence of dementia in follow-up was 17.0% in VHA and 1.8% in MarketScan. Observed incidence rates of dementia for Tdap compared to no vaccination were 144.0/10 000PY and 272.5/10 000PY, respectively, in VHA and 29.5/10 000PY and 51.1/10 000PY, respectively, in MarketScan. See Table 3 for cumulative incidence and incidence rate for Tdap and no Tdap overall and by age group.

Table 3.

Cumulative Dementia Incidence % and Dementia Incidence Rate per 10 000 Person-Years (PY), Patients 65 Years or Older

	Veterans Health Affairs Cohort				MarketScan Cohort
Age Group	Total n	Dementia Events	Cumulative Incidence %	Incidence Rate per 10 000PY	Total n	Dementia Events	Cumulative Incidence %	Incidence Rate per 10 000PY
All ages
Overall	122 946	20 890	17.0%	262.0/10 000PY	174 053	3144	1.8%	49.0/10 000PY
No Tdap vaccine	113 338	19 927	17.6%	272.5/10 000PY	156 181	2953	1.9%	51.1/10 000PY
Tdap vaccine	9608	963	10.0%	144.0/10 000PY	17 872	191	1.1%	29.5/10 000PY
Age 65–69
Overall	34 629	2363	6.8%	93.0/10 000PY	107 498	582	0.5%	15.0/10 000PY
No Tdap vaccine	30 047	2151	7.2%	97.6/10 000PY	94 268	525	0.6%	15.4/10 000PY
Tdap vaccine	4582	212	4.6%	61.6/10 000PY	13 230	57	0.4%	12.3/10 000PY
Age 70–74
Overall	21 442	2569	12.0%	168.0/10 000PY	33 585	658	2.0%	49.8/10 000PY
No Tdap vaccine	19 903	2455	12.3%	173.2/10 000PY	30 899	617	2.0%	50.8/10 000PY
Tdap vaccine	1539	114	7.4%	103.7/10 000PY	2686	41	1.5%	38.7/10 000PY
Age ≥75
Overall	66 875	15 958	23.9%	409.0/10 000PY	32 970	1904	5.8%	155.7/10 000PY
No Tdap vaccine	63 388	15 321	24.2%	415.2/10 000PY	31 014	1811	5.8%	158.1/10 000PY
Tdap vaccine	3487	637	18.3%	296.9/10 000PY	1956	93	4.8%	120.7/10 000PY

Survival modeling results are given in Table 4. Before controlling for confounding, Tdap vaccination, when compared with no Tdap vaccination, was significantly associated with a lower risk for dementia in the VHA cohort (HR = 0.53; 95% CI: 0.50–0.56) and in the MarketScan cohort (HR = 0.58; 95% CI: 0.50–0.66). This association remained nearly unchanged and significant after controlling for confounding in weighted data (VHA: HR = 0.58; 95% CI: 0.54–0.63 and MarketScan: HR = 0.58; 95% CI: 0.48–0.70).

Table 4.

The Association (HR [95% CI]) Between Tdap Vaccination Versus No Tdap Vaccination, Overall and by Age Groups (competing risk survival models in Veterans Health Affairs cohort and Cox Proportional hazard models in MarketScan cohort)

	Veterans Health Affairs Cohort		MarketScan Cohort
Age group	Crude	Weighted	Crude	Weighted
All ages	0.53 (0.50–0.56)	0.58 (0.54–0.63)	0.58 (0.50–0.66)	0.58 (0.48–0.70)
Age 65–69	0.64 (0.55–0.73)	0.68 (0.57–0.81)	0.80 (0.61–1.05)	0.77 (0.58–1.03)
Age 70–74	0.59 (0.49–0.71)	0.45 (0.36–0.56)	0.74 (0.54–1.02)	0.58 (0.37–0.91)
Age ≥ 75	0.71 (0.65–0.77)	0.56 (0.51–0.62)	0.73 (0.59–0.90)	0.52 (0.40–0.68)
Age × Tdap vaccine	p = .119	p = .014	p = .824	p = .120

Table 4 also displays significant effect modification by age after controlling for confounding (p = .014) in VHA data. In VHA weighted data, Tdap vaccination compared to no Tdap vaccination was significantly associated with lower risk for dementia in each age group (Table 4), and the magnitude of association was largest among patients 70–74 years of age (HR = 0.45; 95% CI: 0.36–0.56) and smallest among patients 65–69 years of age (HR = 0.68; 95% CI: 0.57 to −0.81). For MarketScan, age-stratified results for the association of Tdap vaccination and dementia were similar (p = .12).

Sensitivity analyses using incident AD as the outcome showed that in weighted survival models, Tdap vaccination was associated with a 46% reduced risk of AD in the VHA (HR = 0.54; 95% CI: 0.47–0.63) and a 53% reduced risk of AD in MarketScan (HR = 0.47; 95% CI: 0.34–0.63).

The e-value for the association of Tdap and incident dementia was 2.84. An unmeasured confounder would need at least this level of association with both Tdap and incident dementia to explain the results.

Discussion

In a large cohort of VHA patients 65 years and older, patients who received a Tdap vaccination, compared with those who had not been Tdap vaccinated throughout follow-up, had a 42% lower risk for incident dementia. This result was replicated in MarketScan patients of at least 65 years. Nearly identical results in 2 cohorts that differ greatly in clinical and demographic characteristics strengthen the evidence for a protective association between Tdap vaccination and incident dementia. In addition, both cohorts were selected to have 3 or more preventive health care visits in the observation period and our findings are independent of the number of patient well visits prior to baseline. In addition, the magnitude of the association between Tdap vaccination and incident dementia was quite similar before and after controlling for numerous potential confounding factors.

Unlike the VHA sample, in MarketScan patients, there was no significant interaction of age by Tdap vaccination. Yet, the point estimates from both cohorts indicate the weakest relationship among those 65–69 years. This could be due to the low risk of dementia in this age group relative to patients 70 years and older. Tdap vaccination may have a smaller effect on younger patients because they have an overall lower risk of dementia.

Our results are consistent with the analysis of data from the Canadian Study of Health and Aging (10). Specifically, patients 65 years and older who reported a history of tetanus or diphtheria vaccination at baseline versus those who did not self-report this vaccination had a 60% lower risk for AD at 5-year follow-up.

Our study adds to the growing and consistent literature on vaccinations and lower risk for incident dementia. Prior studies have reported a lower risk for dementia in patients with chronic health conditions who did versus did not have a medical claims documented influenza vaccination (6,7). Using a design similar to the present report, we observed patients with a medical record or medical claim for herpes zoster vaccination, compared to those who remained without this vaccination, were significantly less likely to develop dementia (8).

In the context of nearly identical point estimates from multiple studies of Tdap vaccination, and the consistent pattern of results across several different types of vaccinations, it is increasingly plausible that immune responses to vaccination lead to a lower risk for dementia. It is less likely that each specific vaccine type has a unique mechanism in reducing risk for dementia, or that protection against infection with all of these pathogens would have the same biologic effects on dementia risk. These conclusions are supported by evidence that vaccines have nonspecific effects enabling a general resistance to infectious disease (22) and by evidence that the Tdap vaccine has shown beneficial effects on mortality (22). Patients who received Tdap vaccination versus those who did not are probably more likely to have a lifetime history of appropriate vaccinations. Over time, repeated vaccinations may train the immune system and limit inflammation and oxidative stress.

Limitations

Results should be interpreted in the context of several limitations. Due to the very small number of patients who had an indication of tetanus, diphtheria, or pertussis infection in follow-up, we were not able to test if developing one of these infectious diseases was associated with increased risk of dementia, which could provide further evidence for nonspecific versus pathogen-specific mechanisms involved in reduced dementia risk. Only 130 (0.01%) VHA patients and 59 (0.003%) MarketScan patients had one of these conditions during follow-up. Our source data included patients with 3 or more well visits. While this helped control for confounding, it potentially reduces the generalizability of our results. We do not have neuropsychological assessments or neuroimaging to confirm whether Tdap vaccination was associated with a slower progression of cognitive decline. We did not have rearing household income nor measures of social support which are associated with vaccine uptake and dementia (23–25). But we found consistent results in cohorts that differ in socioeconomic status, and adjusted for marital status, a proxy for social support, in the VHA data. Retrospective cohort studies are limited by misclassification and unmeasured/residual confounding. Patients classified as not having a Tdap vaccination may have received one at an Urgent Care site outside of their health care system. Hazard ratios could be underestimated if vaccination status and dementia were misclassified. Unmeasured confounding is a potential limitation. However, it is unlikely that unmeasured confounding could explain our results given the e-value was 2.84. Last, some may question direct comparison between the competing risk model and Cox models computed in the VHA and MarketScan data, respectively. To address this, we reran weighted survival models treating the death as a censored event in VHA data. Weighted results controlling for confounding showed near identical results to competing risk models (HR = 0.57; 95% CI: 0.53–0.62).

Conclusions

Compared to patients who do not receive Tdap vaccination, vaccinated patients have a 42% lower risk for dementia. Tdap joins several other common vaccines that appear to decrease dementia risk. Receipt of appropriate vaccinations may be a cost-effective means to either prevent dementia or slow progression of cognitive decline. Prospective studies and clinical trials are needed to confirm this conclusion.

Funding

This research was supported by a Benter Foundation Grant (2020-01), Common Adult Vaccinations and Incident Dementia. The funder had no role in the study report and no role in the decision to submit for publication.

Declarations

Support for VA/CMS data is provided by the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development, Health Services Research and Development, VA Information Resource Center (Project Numbers SDR02-237 and 98-004). This material is the result of work supported with resources and the use of facilities at the Harry S. Truman Memorial Veterans’ Hospital.

Patient participation: Patients and the public were not involved in design, conduct, or reporting of this study. Dissemination directly to study participants is not possible.

Disclaimer: The views expressed do not necessarily reflect those of the Veterans Health Administration.

Transparency statement: J.F.S. affirms that the manuscript is an honest, accurate, and transparent account of the study being reported. J.F.S. and J.S. had full access to data.

Conflict of Interest

All authors report no conflicts of interest that could inappropriately influence this work.

Author Contributions

All authors met criteria for authorship including contributions to data analysis (J.S.), design (all authors), interpretation of results (all authors), drafting, and completing the final manuscript for submission (all authors).

Data Availability

Data are available with appropriate IRB approval and fees paid to the vendor of commercial claims data. Programming code available upon request.