Herpesviruses, antiviral treatment, and the risk of dementia – systematic review and meta-analysis

Abstract

Introduction

The aim of this systematic review and meta-analysis was to synthesize the evidence on the association between herpesviruses, antiviral treatment, and the risk of dementia. We also aimed to explore the impact of time between herpesviruses and dementia on the reported associations.

Methods

PubMed and Web of Science were searched along with reference lists of the included studies. We included studies that looked at clinical episodes or serology (IgG/IgM) of herpes simplex virus type 1/2 (HSV1/2) and/or varicella zoster virus (VZV), antiviral treatment and incident dementia (all-cause dementia, Alzheimer’s disease, and vascular dementia). Study results were pooled with random effect meta-analyses.

Results

We included 32 studies. The pooled hazard ratio for all-cause dementia was 1.36 [95% CI: 1.01, 1.83] following a clinical episode of HSV1/2, and 1.12 [95% CI: 1.00, 1.25] following a clinical episode of VZV. The pooled estimate for all-cause dementia following antiviral treatment and VZV was 0.88 [95% CI: 0.81, 0.96].

Conclusions

The present review of the scientific literature generally shows little evidence of an association between herpesviruses and risk of dementia. However, the review shows evidence of an association between antiviral treatment and a decreased risk of dementia. Because of considerable heterogeneity, future investigations could advantageously target certain subgroups.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13195-025-01838-z.

Introduction

In the past 30 years it has been debated whether pathogens are linked to dementia, especially Alzheimer’s disease (AD). Herpes simplex virus (HSV) and herpes zoster/varicella zoster virus (HZ/VZV) have gained most attention for their association with dementia [1].

Several epidemiological studies have been published on the association between herpesviruses, antiviral treatment, and dementia, and the research area has grown in interest. Within the past 5 years, seven systematic reviews have been conducted to synthesize evidence on the association between herpesviruses and dementia, but results remain conflicting. Two systematic reviews investigated the risk of AD following HSV1, and both found significantly increased risks [2, 3]. By contrast, another review found no association between HSV1 and all-cause dementia [4]. Five reviews investigated the risk of all-cause dementia following HZ [4–8]. Three of these found increased risks of dementia following HZ [5, 8], and herpes zoster ophthalmicus (HZO) [4], while two found no association [6, 7]. Three systematic reviews investigated the potential benefit of antiviral treatment, focusing on HZ only [5, 7, 8], and found significantly decreased risks of dementia in groups treated with antivirals for HZ. Notably, all the reviews pooled estimates from studies with different study designs and assessments of herpesviruses and outcome, which may have led to inconsistencies in the interpretation of findings. Further, none of the reviews investigating antiviral treatment looked at HSV and very few reviews differentiated by dementia types. Importantly, most of the human population (> 90%) becomes infected with herpesviruses from childhood onwards, and these viruses persist subclinically for a lifetime. For this reason, cases of ‘infection’ generally represent reactivation of latent virus attributed to stress, immune deficiency, or other causes, although reinfection with a new (possibly more virulent) substrain cannot be ruled out. We, therefore, avoid the term ‘infection’ and instead use the terms ‘clinical episode’ or ‘serology’ throughout, to refer to the clinical manifestation of viral disease and reflect past infections or present/reactivations. Most of the previous reviews have grouped studies without differentiating between clinical episodes and serology, which may have complicated interpretation of findings.

Moreover, several observational studies have shown that risk estimates attenuate when removing diagnosed dementia cases, 1–5 years following herpesvirus, indicating – among others – reverse causality. Owing to the long preclinical dementia phase (potentially 10 years before diagnosis), herpesviruses occurring during this period may be a consequence of dementia rather than leading to dementia, due to hygienic habits, declining immune system, or malnutrition. However, none of the seven systematic reviews considered the impact of time between herpesviruses and dementia.

Therefore, the aim of this systematic review and meta-analysis was to synthesize the evidence on the association between herpesviruses (HSV and HZ/VZV), antiviral treatment, and the risk of dementia according to herpesvirus type, assessment of herpesviruses and/or treatment, and dementia type (all-cause dementia, AD, and vascular dementia (VaD)). We also aimed to explore the impact of time between herpesviruses and dementia on the reported associations.

Methods

This systematic review and meta-analysis follows the PRISMA 2020 guidelines. PROSPERO ID number: CRD42024527253.

Search strategy

We searched PubMed and Web of Science for articles in the English language, published since 2013, investigating associations between herpesviruses, antiviral treatment, and incident dementia. The initial search was conducted on 17 March 2024, and updated on 23 June 2025. We used MeSH terms and text words to form the search string (Supplementary Table A1).

Study selection

Our inclusion criteria were human populations with no age limit, and any geographical and/or healthcare setting. We included original articles, excluding reviews, case reports, books, and other reviewed reports etc. Herpesviruses were defined by either a clinical episode or serology (IgG/IgM). Antiviral treatment was defined as prescribed antivirals or for herpesvirus treatment. The main outcome was incident dementia, which included all-cause dementia, AD, and vascular dementia (VaD). Assessments included any that led to a dementia diagnosis, such as medical records, antidementia drugs, neuropsychological assessments and tests conducted by health professionals in the study, or other. Two reviewers (JJ, NJD) screened titles and abstracts independently before selecting full-text papers for review, which were likewise reviewed in parallel. The screening of each study was done using a template created in Covidence – a web-based tool for data extraction [9]. Reasons for exclusion were recorded in Covidence, and the first author resolved disparities by comparing results from each author.

Data extraction

Two reviewers (JJ, NJD) independently and blinded to each other, extracted data from the included studies. We extracted a priori defined information presented in the Supplementary Table A2.

Risk of bias and quality assessment

The Newcastle-Ottawa Scale (NOS) tool was used to assess the quality of the included studies. Scores ranged from 0 to 9 points, where scores equal to or greater than 7 were considered high quality, and scores below 7 were considered low, as previously defined in the literature [10]. Two reviewers (JJ, NJD) assessed the risk of bias individually, blinded, and in parallel with all the studies.

Data analysis

Two main groups of meta-analyses were conducted:

1. Meta-analyses of the association between herpesviruses and risk of dementia. Separate meta-analyses were conducted, grouping by dementia type (all-cause dementia, AD and VaD) and herpesvirus type (HSV1/2 and VZV) because of potential differences in pathways involved between herpesviruses and dementia. We further grouped herpesviruses by clinical episodes and serology (IgG/IgM).

2. Meta-analyses of the association between antiviral treatment and the risk of dementia. We were interested in antiviral treatment among those with a herpesvirus, with reference to no treatment. Separate meta-analyses were conducted, grouping by dementia type (all-cause dementia, AD and VaD), and by herpesvirus type.

We stratified the analyses by sex and setting (study country). Where there was only one study for the analysis, this was not done.

Finally, we were interested in the impact of timing on the association because it could indicate, among others, reverse causality. We did this by gathering studies that excluded time between a clinical episode or serology (IgG/IgM) of a herpesvirus, and incident dementia. If a study used several periods, we included the estimates with the longest period. We gathered and presented these findings narratively due to small number of studies for the same herpesvirus type or dementia outcome.

The studies’ effect estimates were pooled, within the above defined groups, with random effect meta-analysis. We pooled both hazard ratios (HR) and odds ratios (OR)– here collectively named HR– because of the low incidence of dementia in the general population; an OR from a case-control study is similar to a HR in its background population. In the meta-analyses, the studies were weighted by the inverse of the variance of the (logarithm of the) effect measure in each study, giving more influence on studies with greater precision. These weights were further adjusted for between-study heterogeneity through the random effect. Weights for a study in a meta-analysis were presented as the percentage of the total weight of studies included.

The I² coefficient assessed heterogeneity, with I² > 50% indicating heterogeneity [11]. Additionally, a p-value for chi-squared test was presented, with p < 0.05 indicating heterogeneity.

Publication bias was assessed using visual evaluation of the funnel plots for the analysis of herpesviruses and dementia, and for antiviral treatment and dementia. Additionally, Egger’s test was used to quantify the assessment of publication bias.

If two or more studies were based on the same populations, we only included one and conducted sensitivity analyses replacing the study.

Results

Study characteristics

Our initial searches retrieved 1611 studies of which 1292 were screened by title and abstract, 54 were selected for full-text view, and 32 were included in this systematic review (Fig. 1). The final studies (26 cohort studies [12–37], three nested case control studies [38–40], and three case control studies [41–43]) were published between 2013 and 2025, most being published the past 5 years (n = 24). Studies varied according to geographical settings [Europe (n = 17) [15–22, 25, 28, 29, 34, 37–41], Asia (n = 7) [12, 14, 23, 24, 26, 27, 42], North America (n = 8) [13, 30–33, 35, 36, 43].

Fig. 1.

Prisma flow chart. Legend: Figure presents flow chart following PRISMA guidelines

Table 1 presents the characteristics of included studies (further detailed in Supplementary Table A3). Twelve studies assessed HSV1/2 [16, 17, 23, 24, 27, 28, 31–35, 38], six studies HSV1 [13, 19, 25, 36, 41, 43], and two studies HSV2 [36, 41]. Fifteen studies assessed VZV [12, 14, 20, 21, 23, 24, 29, 30, 32, 33, 36, 37, 40–42] and one study HZO [26]. Twelve studies investigated antiviral treatment [12, 14, 15, 18, 22, 27–31, 39, 43]. Of these, four studies investigated antiviral treatment among HSV1/2 only [27, 28, 31, 43], four studies among VZV only [12, 14, 29, 30], two studies both HSV1/2 and VZV [18, 22], and two studies investigated antiviral treatment without specifying virus type [15, 22]. The age range was 54–75 years for groups with herpesviruses, and 64–74 years for groups with antivirals. The majority of the groups in the studies included were women, except in two studies [27, 31]. Several studies lacked information on the age and sex of the groups [13, 16, 17, 19, 22, 32, 36, 39–43]. Twenty-six studies looked at all-cause dementia as outcome [12, 14, 15, 17–33, 35–37, 40, 41, 43], 20 studies AD [12, 13, 15–17, 19, 22–25, 27–29, 33, 34, 36, 38, 39, 42, 43], and seven studies VaD [12, 15, 22–24, 27, 29].

Table 1.

Characteristics of the studies included

Author, country (year), study design	Data source, sample size	Follow up, years	Assessment of herpesviruses and/or treatment			Outcomes assessed	Measure of association, adjustment	Overall quality
			n (%)	Age mean (SD), years	Sex, female, n (%)
Araya et al., US (2025), Cohort [35]	TriNetX database 33,236,890	NI	Clinical episode HSV1/2: 19,878 (0.1)	66.6 (± 7.6)	10,515 (52.9)	All-cause dementia	HR Propensity score matching (age at positive HSV test, sex, race/ethnicity, hypertension, diabetes, stroke, depression, heart disease, obesity, head injury)	7
Bae et al., Korea (2020), Cohort [12]	National Health Insurance Service 229,594	Total: 11	Clinical episode VZV: 34,505 (15)	60.3 (± 8.1)	21,135 (61)	All-cause dementia, AD, VaD	HR Age, sex, economic class, hypertension, diabetes, dyslipidemia, chronic lung disease, ischemic heart disease, stroke, heart failure, atrial fibrillation, valvular heart disease, chronic renal disease, carotid artery stenosis, peripheral vascular disease, chronic liver disease, rheumatic disease, inflammatory bowel disease, tumor, transplant, HIV, depression	9
			Treatment among VZV: 28,873 (84)	66.9 (± 8.1)	17,707 (61)
Barnes et al., US (2014), Cohort [13]	Rush Memory and Aging Project, Religious Orders Study, Minority Aging Research Study 849	Mean: 5	Serology (IgG) HSV1: 370 (44)	NI	NI	AD	HR Adjustment unclear	4
Beydoun et al., US (2024), Cohort [36]	National Health and Nutrition Examination Study (NHANES), Centers for Medicare & Medicaid Services 2975	Total: 19	Serology (IgG) HSV1: NI	NI	NI	All-cause dementia, AD	HR Age, sex, race/ethnicity, income, education, residence, household size, marital status, nutritional status, lifestyle factors, dentate status and social support.	8
			Serology (IgG) HSV2: NI	NI	NI
			Serology (IgG) VZV: NI	NI	NI
Blandi et al., Italy (2025), Cohort [37]	The Lombardy Region Welfare Directorate 132,968	Total: 22	Clinical episode VZV: 12,088 (9)	NI	6963 (57.6)	All-cause dementia	SHR Cancer, COPD, depression, immunosuppression, rheumatologic diseases	9
Chen et al., Taiwan (2018), Cohort [14]	National Health Insurance Research Database 78,410	Mean: 6.22 (± 4.05)	Clinical episode VZV: 39,205 (50)	63.5 (± 10)	21,422 (55)	All-cause dementia	HR Age, sex, residence, depression, autoimmune disease, ischemic stroke, traumatic brain injury, alcohol, antiviral treatments, HZ drug, HSV	9
			Treatment among VZV: 2131 (5)	NI	NI
Choi et al., Korea (2021), Case control [42]	National Health Insurance Service 11,445 dementia cases, 45,780 controls	Total: 12	Clinical episode VZV: cases 928 (8), controls 3929 11 (9)	NI	NI	AD	OR Age, sex, income, residence, medical history (matched), ischemic heart disease, stroke, depression	7
Hemmingsson et al., Sweden (2021), Nested case control [39]	Betula Cohort study 262 AD cases, 262 controls	Mean: cases 9.3 (± 6.1), controls 13.5 (± 6.9)	Treatment among HSV1: cases 6 (2), controls 20 (8)	NI	NI	AD	OR Age, sex, APOE4 (matched), length of follow-up	6
Linard et al., France (2020), Cohort [17]	Bourdeaux-3 C cohort study 1037	Mean: 7.4 (± 2.9)	Serology (IgG) HSV1/2: 844 (81)	NI	NI	All-cause dementia, AD /mixed dementia	HR Sex, education, marital status, APOE4, hypertension, diabetes, hypercholesterolemia, smoking	5
			Serology (IgM) HSV1/2: 47 (5)	NI	NI
Linard et al., France (2021), Cohort [16]	Three City study, Aging Multidisciplinary Investigation study 1599	Mean: 6.8 (± 2.6)	Serology (IgG) HSV1/2: 1327 (83)	NI	NI	AD	HR Sex, education, marital status, APOE4, hypertension, diabetes, hypercholesterolemia, smoking, cohort	5
Linard et al., France (2022), Cohort [15]	The “Echantillon Généraliste des Bénéficiaires” 68,291	Total: 8	Treatment among total sample: 6642 (10)	74 (± 6)	4331 (65)	All-cause dementia, AD, VaD	HR Age, sex, health insurance, hypertension, diabetes, hypercholesterolemia, heart disease, stroke, NSAID, glucocorticoids, medications before inclusion, outpatient medical consultations	9
Lindman et al., Sweden (2021), Cohort [18]	National Patient Register, Prescribed Drug Register 530,344	Mean: in exposed: 5.2 (± 3.4)	Treatment among HSV1/2, VZV: 29,593 (11)	72 (± 10)	17,371 (59)	All-cause dementia	HR Birth year, sex (matched), alcohol, COPD, congestive heart failure, myocardial infarction, Parkinson’s disease, stroke, antidepressants, antidiabetics, antihypertensives, antipsychotics, education	9
Liu et al., US (2025), Case-control [43]	IQVIA PharMetrics Plus 344,628 AD cases, 344,628 controls, 747,653 ADRD cases, 747,653 controls	Mean: AD: 47.86 (± 46.24), ADRD: NI	Clinical episode HSV1: AD analysis 2330 (0.7), ADRD analysis 5109 (0.7)	NI	NI	AD, ADRD (all-cause dementia)	OR Age, sex, residence, comorbidities (CCI)	7
			Treatment among HSV1: AD analysis 931 (40), ADRD analysis 2104 (41)	NI	NI
Lophatananon et al., UK (2021), Nested case control [40]	UK biobank 2378 dementia cases, 225,845 controls	NI	Clinical episode VZV: cases 417 (18), controls 34,699 (15)	NI	NI	All-cause dementia	OR Age, sex	6
Lövheim et al., Sweden (2015), Cohort [34]	Betula cohort study 3432	Mean: 11.3 (± 5.7)	Serology (IgG) HSV1/2: 3026 (88)	63.6	1661 (55)	AD	HR Age, sex	4
			Serology (IgM) HSV1/2: 100 (3)	63.9	61 (61)
Lövheim et al., Sweden (2015), Nested case control [38]	The Northern Sweden Health and Disease Study 360 AD cases, 360 controls	NI	Serology (IgG) HSV1/2: cases 338 (94), controls 324 (90)	NI	Cases 255 (75), controls 247 (76)	AD	OR NI	6
			Serology (IgM) HSV1/2: cases 27 (8), controls 20 (6)	NI	NI
Mekli et al., UK (2022), Case control [41]	UK biobank 85 dementia cases, 9346 controls	Mean: 8.9	Serology (IgG) HSV1: cases 71 (84), controls 6522 (70)	NI	NI	All-cause dementia	OR Age	4
			Serology (IgG) HSV2: cases 11 (13), controls 1514 (16)	NI	NI
			Serology (IgG) VZV: cases 83 (98), controls 8633 (92)	NI	NI
Murphy et al., The Netherlands (2021), Cohort [19]	Rotterdam Study 1915	Mean: 9.1 (± 3.4)	Serology (IgG) HSV1: 1518 (79)	NI	NI	All-cause dementia, AD	HR Age, sex, BMI, smoking, alcohol, education, hypertension, hypercholesterolemia, APOE4, coronary heart disease, diabetes, stroke	6
Omland et al., Denmark (2021), Cohort [20]	Nationwide registries 10,340	NI	Clinical episode VZV: 517 (5)	Median: 59 (31–77)	268 (52)	All-cause dementia	HR NI	6
Schmidt et al., Denmark (2022), Cohort [21]	Nationwide registries 1,483,195	Median: 6 (3–11)	Clinical episode VZV: 247,305 (17)	Median: 64 (54–74)	149,796 (61)	All-cause dementia	HR Birth year, sex (matched), autoimmune disease, chronic kidney disease, COPD, asthma, hematologic cancer, solid cancer, diabetes, glucocorticoids, HIV, lipid-lowering therapy, traumatic head injury	9
Schnier et al., Wales, Germany, Denmark, Scotland (2021), Cohort [22]	Wales: Secure Anonymized Information Linkage Databank, 434,689 Germany: IMS disease analyzer, 53,629 Denmark: Nationwide registries, 1,712,100 Scotland: Electronic Data Research and Innovation Service, 342,637	Median: Wales: 6.7 (3.3–11.3) Germany: 8.8 (4.5–14.5) Denmark: 7.4 (3.8–12.2) Scotland: 2.7 (1.4–4.2)	Treatment among HSV1/2: Wales: NI, Germany: NI, Denmark: NI, Scotland: NI Treatment among VZV: Wales: NI, Germany: NI, Denmark: NI, Scotland: NI Treatment among total sample: Wales: 39,997 (9), Germany: 10,352 (19), Denmark: 169,989 (8), Scotland: 8625 (3)	NI	NI	All-cause dementia, AD, VaD	HR Wales and Scotland: birth year, sex, multi deprivation Germany: birth year, sex, insurance Denmark: birth year, sex, education, civil status, CCI	8
Shim et al., Korea (2022), Cohort [23]	National Health Insurance Service 373,197	Total: 10	Clinical episode HSV1/2: 92,095 (25)	62.51 (± 9.1)	56,021 (61)	All-cause dementia, AD, VaD	HR Age, sex, hypertension, diabetes, hyperlipidemia, stroke	9
			Clinical episode VZV: 97,323 (26)	63.48 (± 9.4)	59,130 (61)
Shin et al., Korea (2024), Cohort [24]	National Health Insurance Service 752,205	Total: 11	Clinical episode HSV1/2: 91,990 (12)	56.9 (± 9.2)	NI	All-cause dementia, AD, VaD	HR Age, sex, HSV, VZV, depression, dyslipidemia, ischemic stroke, coronary heart disease, hypertension, diabetes, chronic kidney disease, BMI, income	9
			Clinical episode VZV: 113,645 (15)	59.5 (± 10.1)	NI
Torniainen-Holm et al., Finland (2019), Cohort [25]	Finnish Health Survey 8028	Total: 11	Serology (IgG) HSV1: 4656 (58)	Median: 54.4	2440 (52)	All-cause dementia, AD	HR Age, sex, marital status, education, residence, language	9
Tsai et al., Taiwan (2017), Cohort [26]	National Health Insurance Research Database 3384	Total: 5	Clinical episode HZO: 846 (25)	62.2 (± 12.5)	426 (50)	All-cause dementia	HR Age, sex, residence, income, hypertension, diabetes, hyperlipidemia, coronary heart disease, stroke	9
Tzeng et al., Taiwan (2018), Cohort [27]	National Health Insurance Research Database 33,448	Total: 10	Clinical episode HSV1/2: 8362 (25)	NI	3638 (44)	All-cause dementia, AD, VaD	HR Age, sex, residence, hospitals (local/regional), income	9
			Treatment among HSV1/2: 7215 (86)	NI	NI
Ukraintseva et al., US (2024), Cohort [33]	Health and Retirement Study 6001 for AD sample, 5548 for ADRD-sample	Mean: AD: 6.08, ADRD: 5.69	Clinical episode HSV1/2: (3) †	65–75	NI	AD, ADRD (all-cause dementia)	SHR Sex, race, marital status, education, veteran status, economic class, wealth, insurance, BMI, alcohol, smoking, self-reported health	6
			Clinical episode VZV: (9) †	65–75	NI
Vestin et al., Sweden (2024), Cohort [28]	Uppsala Seniors cohort 1002	Total: 15	Serology (IgG) HSV1/2: 825 (82)	70.2 (± 0.2)	418 (51)	All-cause dementia, AD	HR Sex, education, APOE4	3
			Serology (IgM) among (IgG) HSV1/2: 81 (10)	NI	NI
			Treatment among serology (IgG) HSV1/2: 51 (6)	NI	NI
Warren-Gash et al., UK (2022), Cohort [29]	Clinical Practice Research Datalink 884,045	Median: 4.6 (2.0-8.1)	Clinical episode VZV: 177,144 (20)	65.1 (± 12.9)	106,454 (60)	All-cause dementia AD, VaD	HR Age, sex, frailty index, prior consultation rate, alcohol, BMI, smoking, chronic kidney disease, asthma, autoimmune disease, COPD, depression, hypertension, ischaemic heart disease, immunosuppression, liver disease, stroke, traumatic brain injury, HSV, diabetes	9
			Treatment among VZV: 110,997 (63)	NI	NI
Weinmann et al., US (2024), Cohort [30]	Kaiser Permanente cohort 101,328	Total: 20	Clinical episode VZV: 25,332 (25)	Median: 64 (57–73)	15,556 (61)	All-cause dementia	HR Age, sex, race, general practice, specialty/inpatient visits, autoimmune disorder, immunosuppression, alcohol, blood disorder, cancer, depression, diabetes, hearing loss, heart failure, hyperlipidemia, hypertension, stroke, traumatic brain injury, smoking	8
			Treatment among VZV: 20,842 (82)	Median: 64 (57–72)	12,920 (62)
Young-Xu et al., US (2021), Cohort [31]	Veterans’ Health Administration and electronic medical records 305,582	Median: 8.9 (6.0-12.2)	Clinical episode HSV1/2: 87,687 (29)	NI	5794 (7)	All-cause dementia AD, VaD	HR Age, rurality, economic class, ischemic stroke, autoimmune disease, immunosuppression, alcohol	8
			Treatment among HSV1/2: 61,776 (71)	NI	NI
Zilli et al., US (2021), Cohort [32]	The Framingham Heart Study 1380	Total: 35	Clinical episode HSV1/2: 641 (46)	NI	NI	All-cause dementia	HR Age, sex, education	5
			Clinical episode VZV: 145 (11)‡	NI	NI

Legend: The table presents characteristics of studies included in the systematic review. We presented the total follow up time; if this was not available, we presented the mean or median. Means are presented with SD and medians with IQR. Age is presented as mean (SD), unless it is specified as median, then median (IQR). Antiviral treatment was assessed through a prescription record of drugs, detailed in Table A3. † % of the study population without specifying the sample. The study used three samples and only presented the % as reported. ‡ Medical record diagnosis based on self-reports through questionnaire of clinical history of herpesvirus, assessed together with medical examinations. Overall quality was assessed using the Newcastle-Ottawa Scale. Scores ranged from 0 to 9 points, where scores equal to or greater than 7 were considered high quality and scores below 7 were considered low, as previously defined in the literature [10]. Abbreviations: AD, Alzheimer’s disease; AD/ADRD, Alzheimer’s disease and Alzheimer’s disease related dementias; APOE4, apolipoprotein E4 allele; BMI, body mass index; CCI, Charlsons comorbidity index; COPD, chronic obstructive pulmonary disease; HIV, human immunodeficiency virus; HR, hazard ratio; HSV1, herpes simplex virus type 1; HSV2, herpes simplex virus type 2; HSV1/2, herpes simplex virus type 1 and 2; HZ, herpes zoster; HZO, herpes zoster ophthalmicus; IgG, immunoglobulin G; IgM, Immunoglobulin M; IQR, Interquartile range; NI, no information; NSAID, non-steroidal anti-inflammatory drugs; OR, odds ratio; SD, standard deviation; SHR, Sub distribution HR which takes competing risk into account; UK, United Kingdom; US, United States; VaD, vascular dementia; VZV, varicella zoster virus

Risk of bias and quality of studies

Of the cohort studies, 17 studies had high quality [12, 14, 17, 18, 21–27, 29–31, 35–37], and nine studies had low quality [13, 16, 17, 19, 20, 28, 32–34]. Of the nested case control and case control studies, two studies had high quality [42, 43], and four studies had low quality [38–41]. Results are presented in Table 1 and detailed in Supplementary Tables A4 and A5.

Meta-analyses

In the meta-analyses, we included 30 of the 32 studies. Shin et al. (2024) [24] and Shim et al. (2022) [23] analyzed populations from the same data source and the same herpesvirus and dementia type. Likewise for Schmidt et al. (2022) [21] and Omland et al. (2021) [20]. We included only Shin et al. [24] and Schmidt et al. [21] in the main analyses owing to higher study quality, longer follow-up, or larger sample sizes. These were replaced with Shim et al. [23] and Omland et al. [20] in the sensitivity analyses (Supplementary Figures A1-A3). Furthermore, two studies assessed HSV1 and HSV2 separately [36, 41] so we pooled the results for HSV1 in the main analysis and replaced them with HSV2 in a sensitivity analysis (Supplementary Figure A4).

Association between herpesviruses and risk of dementia

Figure 2 presents meta-analyses of the association between herpesviruses and risk of all-cause dementia. The pooled HR for all-cause dementia was 1.36 [95% CI: 1.01, 1.83] following a clinical episode of HSV1/2, 1.22 [95% CI: 0.87, 1.71] following HSV1/2 by serology (IgG), and 0.80 [95% CI: 0.44, 1.44] following HSV1/2 by serology (IgM). The pooled HR for all-cause dementia was 1.12 [95% CI: 1.00, 1.25] following a clinical episode of VZV, and 1.62 [95% CI: 0.60, 4.37] following VZV by serology (IgG).

Fig. 2.

Meta-analyses of the association between herpesviruses and all-cause dementia, by type and assessment of herpesviruses. Legend: The figure presents a forest plot of meta-analyses of studies on the association between herpesviruses and all-cause dementia. Meta-analyses were performed separately by herpesvirus type (HSV1/2 or VZV) and by clinical episode or serology (IgG/IgM). We pooled the overall HSV1/2 estimates, and if a study only looked at HSV1 we pooled this instead. We pooled VZV for each study, but in Tsai et al. (2017) the results were only conducted for HZO, and this was therefore pooled with VZV. Araya et al. (2025) based the clinical episode on LOINC codes that are codes for laboratory and clinical observations. Each meta-analysis presents a measure of heterogeneity (I²) between the studies. Schmidt(a) HRs with 95% CIs of VZV associated with dementia within 0–1 year of follow-up. Abbreviations: CI, confidence interval; HR, hazard ratio; HSV1, herpes simplex type 1; HSV1/2, herpes simplex virus types 1 and 2; HZO, herpes zoster ophthalmicus; IgG, immunoglobulin G; IgM, immunoglobulin M; VZV, varicella zoster virus

In the sensitivity analyses replacing studies from the same populations, the pooled estimate changed for clinical episodes of HSV1/2 and all-cause dementia to 1.33 [95% CI: 0.99, 1.80] (changes from statistically significant to insignificant). For clinical episodes of VZV and all-cause dementia, the pooled estimates changed to 1.07 [95% CI: 1.01, 1.13], when replacing Shin et al. with Shim et al., and to 1.19 [95% CI: 1.02, 1.39], when replacing Schmidt et al. with Omland et al. (changes from statistically insignificant to significant).

Figure 3 presents meta-analyses of the association between herpesviruses and the risk of AD and VaD. The pooled HR for AD was 1.45 [95% CI: 0.97, 2.17] following a clinical episode of HSV1/2, and 1.27 [95% CI: 0.76, 2.12] for VaD, following a clinical episode of HSV1/2. The pooled HR for AD was 1.04 [95% CI: 0.88, 1.24] following HSV1/2 by serology (IgG), 1.41 [95% CI: 0.99, 1.99] following HSV1/2 by serology (IgM), and 1.12 [95% CI: 0.93, 1.35] following a clinical episode of VZV. The pooled HR for VaD was 1.13 [95% CI: 0.92, 1.39] following a clinical episode of VZV.

Fig. 3.

Meta-analyses of the association between herpesviruses and AD and VaD, by herpesvirus type and assessment. Legend: The figure presents a forest plot of meta-analyses of studies on the association between herpesviruses and AD and VaD. Meta-analyses were performed separately by herpesvirus type (HSV 1/2 or VZV) and by clinical episode or serology (IgG/IgM). We pooled estimates for HSV1/2, and if a study only looked at HSV1 we pooled this instead. Each meta-analysis presents a measure of heterogeneity (I²) between the studies. Lövheim(a) Betula cohort study. Lövheim(b) The Northern Sweden Health and Disease Study Cohort. Abbreviations: AD, Alzheimer’s disease; CI, confidence interval; HR, hazard ratio; HSV1, herpes simplex type 1; HSV1/2, herpes simplex virus types 1 and 2; IgG, Immunoglobulin G; IgM, Immunoglobulin M; VaD, vascular dementia; VZV, varicella zoster virus

Figures A5 and A6 present meta-analyses of the association between herpesviruses and risk of dementia by sex and setting, respectively. Meta-analyses were only possible for all-cause dementia as the outcome, and not AD/VaD, because of few studies. The pooled HR for all-cause dementia following a clinical episode of VZV was 1.16 [95% CI: 0.84, 1.60] for males and 1.04 [95% CI: 0.91, 1.18] for females. The pooled HR for all-cause dementia following a clinical episode of HSV1/2 was 1.88 [95% CI: 1.02, 3.45] in Asia and 1.19 [95% CI: 0.88, 1.62] in North America. The pooled HR for all-cause dementia following a clinical episode of VZV was 1.02 [95% CI: 0.93, 1.13] in Europe, 1.44 [95% CI: 0.97, 2.14] in Asia, and 1.02 [95% CI: 0.94, 1.11] in North America.

Association between antiviral treatment and the risk of dementia

Figure 4 presents meta-analyses of the association between antiviral treatment and the risk of dementia. The pooled HR for all-cause dementia was 0.66 [95% CI: 0.33, 1.29] following HSV1/2, 0.88 [95% CI: 0.81, 0.96] following VZV, and 0.91 [95% CI: 0.82, 1.00] following unspecified type. The pooled HR for AD was 0.78 [95% CI: 0.33, 1.86] following HSV1/2 and 0.93 [95% CI: 0.82, 1.05] following unspecified type. Lastly, the pooled HR for VaD was 0.96 [95% CI: 0.80, 1.16] following unspecified type.

Fig. 4.

Meta-analyses of the association between herpesviruses, antiviral treatment and all-cause dementia, AD, and VaD. Legend: The figure presents a forest plot of meta-analyses of studies on the association between antiviral treatment and dementia. Meta-analyses were performed separately by dementia type (all-cause dementia, AD, and VaD). Each meta-analysis presents a measure of heterogeneity (I²) between the studies. Antiviral treatment was assessed from drug prescription records, as detailed in Table A3. All studies except Linard et al. (2022) and Schnier et al. (2021) studied antiviral treatment among patients with herpesvirus compared to herpesvirus without treatment (hence the herpesvirus type for these studies is not presented in the forest plot). In Lindman et al. (2021) the results were compared to controls, defined as people who had neither herpesviruses nor were treated. Schnier(a) The Secure Anonymised Information Linkage Databank, Wales. Schnier(b) The IMS Disease Analyzer, Germany. Schnier(c) Nationwide registries, Denmark. Schnier(d) Electronic Data Research and Innovation Service, Scotland. Abbreviations: AD, Alzheimer’s disease; CI, confidence interval; HR, hazard ratio; HSV1, herpes simplex type 1; HSV1/2, herpes simplex virus types 1 and 2; VaD, vascular dementia; VZV, varicella zoster virus

Figure A7 presents meta-analyses stratified by setting. Stratification by sex was not possible due to few studies. The pooled HR for antiviral treatment and all-cause dementia was 0.93 [95% CI: 0.90, 0.97] for HSV1/2 in Europe, 0.78 [95% CI: 0.72, 0.85] for HSV1/2 in North America, 0.67 [95% CI: 0.49, 0.91] for VZV in Asia, 0.92 [95% CI: 0.89, 0.96] for VZV in Europe, and 0.91 [95% CI: 0.82, 1.00] for unspecified herpesviruses in Europe. Furthermore, the pooled HR for antiviral treatment and AD was 0.72 [95% CI: 0.13, 4.09] for HSV1/2 in Europe, and 0.93 [95% CI: 0.82, 1.05] for unspecified herpesviruses in Europe. Lastly, the pooled HR for antiviral treatment and VaD was 0.96 [95% CI: 0.80, 1.16] for unspecified herpesviruses in Europe.

Timing between herpesviruses and dementia

Figure 5 presents eight studies investigating the impact of timing between herpesviruses and all-cause dementia, AD or VaD [14, 20, 21, 27, 29, 30, 34, 37]. The studies used different methods, including delayed entry into the study, time between herpesviruses and dementia diagnosis, exclusion of follow-up time after herpesviruses, and washout periods. Such approaches signal whether the impact of herpesviruses on dementia is mostly seen in the short term or if it persists in the long term. Time intervals in the eight studies varied from 1 year to 10 years. Of the seven statistically significantly increased dementia HRs (of five studies [14, 20, 21, 27, 37], most attenuated after applying timing, with two estimates (of two studies [20, 21]) no longer statistically significant. Of the two statistically significantly decreased dementia HRs (of two studies [21, 29]), both remained statistically significantly decreased after applying timing. Lastly, three estimates (of three studies [21, 30, 34]) showed no statistically significant associations before applied timing. Of these, one estimate [34] was statistically significantly increased after applying timing, another [21] was statistically significantly decreased. Finally, one estimate was increased but did not reach significance [30] (see changes in Fig. 5).

Fig. 5.

Timing between herpesviruses and dementia. Legend: The figure presents a forest plot of studies applying timing between herpesviruses and dementia. Main estimates from the meta-analyses are presented in black and estimates with the longest time excluded between herpesviruses and dementia are shown in purple. If a study looked at multiple timings, these were pooled and compared to the respective overall estimate. If a study looked at time excluded after baseline, this was highlighted. Otherwise, the excluded time between herpesviruses and dementia was presented. Schmidt et al. (2022) presented both an overall estimate, an estimate for hospitalization with VZV, and an estimate for prescription of drugs from VZV. Each identification method with applied timing is compared to the main estimate with the respective identification method. Lövheim(a) The Northern Sweden Health and Disease Study Cohort. Schmidt(a) Herpesviruses either identified through hospitalization or prescription of drugs. Schmidt(b) Herpesviruses identified through hospitalization. Schmidt(c) Herpesviruses identified through prescription of drugs. Abbreviations: CI, confidence interval; AD, Alzheimer’s disease; HR, hazard ratio; VaD, vascular dementia; VZV, varicella zoster virus

Heterogeneity and publication bias

Results on heterogeneity ranged from 0% to 99.8%, and most meta-analyses showed high heterogeneity (I² > 50%). In contrast, one meta-analysis with eight studies looked at AD following HSV1/2 by serology (IgG), and showed low heterogeneity (I² = 29.7%), suggesting a common underlying effect. The funnel plots showed that data points (studies) were placed asymmetrically around the pooled effect estimate (Supplementary Figures A8 and A9), indicating publication bias favoring significant results. However, Egger’s test showed no statistically significant evidence for publication bias.

Discussion

Our systematic review found no overall consistent, statistically significant associations between herpesviruses (HSV1/2 and VZV) and risk of all-cause dementia, AD, or VaD. The pooled estimate for clinical HSV1/2 was borderline significant for all-cause dementia, while clinical VZV showed significant associations only in sensitivity analyses, not in main analyses. No significant associations were found for serology-based measures, but antiviral treatment—mainly for VZV—was associated with reduced dementia risk. To our knowledge, this is the first systematic review to comprehensively synthesize evidence on HSV1/2, VZV, all-cause dementia, AD, VaD, and antiviral treatment risks, and it shows attenuation when excluding time between herpesviruses and incident dementia.

In our meta-analyses, most pooled estimates showed increased risk of dementia following herpesviruses, although not significant. One of our meta-analyses showed an insignificant decreased risk of all-cause dementia followed by herpesviruses (serology (IgM)). However, this meta-analysis was based on two studies. Our findings align with previous meta-analyses, and we discuss discrepancies in the following.

For HSV, our findings were not in line with two previous reviews that found an increased risk of AD following HSV1 based on case control studies [2, 3]. Our findings were, however, in line with a review that found no significant associations between HSV1/2 and risk of all-cause dementia [4]. Inconsistencies between our findings and the previous can be explained by differences in the studies included. In the present review, both case control and cohort studies were included, and our inclusion criteria ensured that herpesviruses preceded dementia, hence incident dementia. The results and conclusions in the previous reviews were based on numerous studies that, for example, looked at the presence of herpesvirus DNA in brains of AD patients rather than incidence of AD. In addition, in one of the two previous reviews, the pooled effect based on cohort studies (exploring AD incidence) showed insignificant associations [3], aligned with our findings.

For VZV, we found no overall association with all-cause dementia, AD, or VaD, which was in line with four previous meta-analyses [3, 4, 6, 7]. However, two other meta-analyses found significantly increased risk of all-cause dementia, but not of AD or VaD, following a clinical episode of VZV (pooled HR 1.11 [95% CI: 1.02, 1.21] and 1.15 [95% CI: 1.03, 1.28]) [5, 8]. Our findings on VZV and all-cause dementia risk were inconsistent, but high-quality, long-term studies showed no significant association, highlighting the need for further research.

Our finding of an effect of antiviral treatment on dementia risk (pooled HR 0.88 [95% CI: 0.81, 0.96]) aligns with three previous reviews that found significantly decreased risks of all-cause dementia among patients with VZV and antiviral treatment [5, 7, 8]. However, when the previous reviews looked at dementia types, no significant association was found for AD or VaD, which is in line with what we found.

The lack of consistently statistically significant results in our meta-analyses can have several explanations. First, it may reflect no true association between herpesviruses and dementia. This aligns with evidence suggesting infections in general, rather than specific pathogens, contribute to dementia risk through inflammation or immune decline. Studies show that severe infections requiring hospitalization significantly increase dementia risk [44], while common infections in primary care show no evidence [45]. One study in our meta-analysis found increased dementia risk after hospitalization with herpesviruses [21], but not in primary care (i.e., less severe cases). However, most studies did not differentiate by severity, potentially masking effects of severe herpesvirus infections.

Second, the overall lack of associations in our meta-analyses can reflect the difficulty of assessment of herpesviruses in observational studies. Herpesviruses differ from many other infections in that they can be latent and/or reactivated. Moreover, we recognize the mounting evidence from animal and in vitro studies, highlighting potential causal pathways between herpesviruses and dementia [46], which may be difficult to replicate in observational assessments. We find that such difficulties are emphasized by our findings where we saw borderline statistically significant associations that were inconsistent across several of our analyses. In addition, we cannot exclude that a proportion of dementia cases may be attributed to herpesviruses because of genetics (e.g., APOE4) [47]. This could mask potential subgroup associations, however, most studies in our review did not differentiate according to this.

As for the effect of antiviral treatment, we found a decreased risk of dementia. This was mainly observed for VZV, while significant HSV1/2 associations appeared only by setting, likely due to few North American studies and the disproportionate influence of one large study in Europe. While antiviral treatment may offer a promising therapeutic avenue and suggest herpesvirus involvement in dementia, our findings showed no increased risk for herpesviruses, which may emphasize the challenges in herpesvirus assessment in observational studies. Furthermore, antiviral treatment may be difficult to measure in observational studies due to health seeking behavior. As such, decreased risks may reflect such differences, rather than treatment effects. Different antiviral drugs may also have different effects on dementia risk, notably because of different dosage, routes of administration (e.g., oral vs. intravenous), and brain penetrance, but we were not able to separate our meta-analyses because, in most cases, the relevant parameters (specific drug/dosage/route) were not reported. One study in our analyses reported that antiviral treatment was associated with a strongly reduced risk of dementia, but the study focused mainly on severely affected patients (with more than three visits at the GP within 1 year) who received antivirals for an average of 12 weeks, sometimes 30 weeks [27]. The role of herpes vaccinations has also been investigated, and protective effects have been seen ‎with vaccines against VZV (e.g., [48]). Nevertheless, diverse vaccines have a significant impact on ‎dementia incidence, but this was ascribed to their ability to induce trained immunity, rather than to agent-specific protection, as exemplified by the reported protective effects of administering ‎adjuvant-only preparations [49, 50]. For this reason, vaccination does not reliably cast light on the specific identity of the infectious agent(s) that might predispose to dementia development, which is the topic of the current report.

Overall, these difficulties in measuring herpesviruses and antiviral treatment in observational studies may be due to (a) disease severity (discussed above), perhaps reflecting an underlying genetic predisposition, (b) geographic differences in genetics and/or viral (sub)strains, and/or (c) extended duration of antiviral medication. These complexities, mentioned in observational study designs, can lead to biases that complicate the assessment of true effects of herpesviruses and antiviral treatment on dementia risk. Furthermore, the lack of associations in our meta-analyses should be interpreted with the high heterogeneity in mind, as underlying differences in the studies may have attenuated the effects. Although we investigated a wide range of potential sources that could influence the association, some may remain unexplored.

Our findings highlight the importance of exploring the timing between herpesviruses and incident dementia, as associations were strongest—or only observed—in the short term, despite no overall link. This suggests herpesviruses may trigger or unmask dementia symptoms, rather than cause the disease long-term, emphasizing the need for studies to investigate timing to better understand underlying mechanisms.

Strengths and limitations

Strengths of our review include our comprehensive search strategy and a pre-specified plan, with consideration of herpesvirus and outcome assessments. We conducted thorough bias and publication bias assessments, included studies with confounder adjustments and focused on the past 12 years. Importantly, we note that several of the previous reviews, particularly those on HSV, conclude that there is an increased dementia risk. However, such conclusions were based on results from case control studies that assessed herpesviruses in dementia/AD patients, rather than assessing incident dementia/AD following herpesviruses. Hence, another strength of our current review is that we only included studies of incident dementia. Nevertheless, this review was limited by the small number of studies available for sex-stratified antiviral analyses and by the limited data for pooling in some meta-analyses by herpesvirus and dementia type. However, we consider it a strength that we assessed the risk based on herpesvirus type and dementia type. Regarding our analysis of timing, we could not pool studies in the meta-analysis, and a qualitative synthesis was employed to analyze the results instead.

Future directions/implication

Future meta-analyses should group studies by both herpesvirus type and the type of dementia, because the pathways may be different. We also encourage future studies to explore the timing between herpesviruses and dementia diagnosis, and attempt to assess the severity of each case, to give more clues on understanding the underlying mechanisms and how they potentially shape the bidirectional relationships between herpesviruses and dementia.

In conclusion, our systematic review and meta-analyses found no overall consistent associations between herpesviruses, and the risk of all-cause dementia, AD, or VaD. However, we found an association between antiviral treatment (mainly for VZV) and all-cause dementia. We discuss our findings in light of potential underlying mechanisms and highlight important future research directions that enhance such understandings.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

IgG

Immunoglobulin G

IgM

Immunoglobulin M

HSV1/2

Herpes simplex virus type 1/2

VZV

Varicella zoster virus

AD

Alzheimer’s disease

VaD

Vascular dementia

HSV

Herpes simplex virus

HZ

Herpes zoster

HSV1

Herpes simplex virus type 1

HSV2

Herpes simplex virus type 2

HZO

Herpes zoster ophthalmicus

NOS

Newcastle-Ottawa Scale

HR

Hazard ratio

OR

Odds ratio

CI

Confidence interval

APOE4

Apoliprotein E4 allele

AD/ADRD

Alzheimer’s disease and Alzheimer’s disease related dementias

BMI

Body mass index

CCI

Charlsons comorbidity index

COPD

Chronic obstructive pulmonary disorder

HIV

Human immunodeficiency virus

IQR

Interquartile range

NI

No information

NSAID

Non-steroidal anti-inflammatory drugs

SD

Standard deviation

UK

United Kingdom

US

United States

Author contributions

All authors were involved in the conceptualization of this work. NJD and VS conducted the statistical analyses. NJD and JJ conducted the literature searches and assessments. NJD, VS, RL, GW, and JJ all interpreted the findings and critically reviewed the manuscript.

Funding

Open access funding provided by Copenhagen University. The Danish Dementia Research Center is supported by funding from the Danish Ministry of Health, who had no role in this study.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.