Risk Factors in Late Adolescence for Young-Onset Dementia in Men: A Nationwide Cohort Study

Importance  Young-onset dementia (YOD), that is, dementia diagnosed before 65 years of age, has been related to genetic mutations in affected families. The identification of other risk factors could improve the understanding of this heterogeneous group of syndromes.

Objective  To evaluate risk factors in late adolescence for the development of YOD later in life.

Design  We identified the study cohort from the Swedish Military Service Conscription Register from January 1, 1969, through December 31, 1979. Potential risk factors, such as cognitive function and different physical characteristics, were assessed at conscription. We collected other risk factors, including dementia in parents, through national register linkage.

Participants  All Swedish men conscripted for mandatory military service (n = 488 484) with a mean age of 18 years.

Setting  Predominantly Swedish men born from January 1, 1950, through December 31, 1960.

Exposure  Potential risk factors for dementia based on those found in previous studies, data available, and quality of register data.

Main Outcomes and Measure  All forms of YOD.

Results  During a median follow-up of 37 years, 487 men were diagnosed as having YOD at a median age of 54 years. In multivariate Cox regression analysis, significant risk factors (all P < .05) for YOD included alcohol intoxication (hazard ratio, 4.82 [95% CI, 3.83-6.05]); population-attributable risk, 0.28), stroke (2.96 [2.02-4.35]; 0.04), use of antipsychotics (2.75 [2.09-3.60]; 0.12), depression (1.89 [1.53-2.34]; 0.28), father’s dementia (1.65 [1.22-2.24]; 0.04), drug intoxication other than alcohol (1.54 [1.06-2.24]; 0.03), low cognitive function at conscription (1.26 per 1-SD decrease [1.14-1.40]; 0.29), low height at conscription (1.16 per 1-SD decrease [1.04-1.29]; 0.16), and high systolic blood pressure at conscription (0.90 per 1-SD decrease [0.82-0.99]; 0.06). The population-attributable risk associated with all 9 risk factors was 68%. Men with at least 2 of these risk factors and in the lowest third of overall cognitive function were found to have a 20-fold increased risk of YOD during follow-up (hazard ratio, 20.38 [95% CI, 13.64-30.44]).

Conclusions and Relevance  In this nationwide cohort, 9 independent risk factors were identified that accounted for most cases of YOD in men. These risk factors were multiplicative, most were potentially modifiable, and most could be traced to adolescence, suggesting excellent opportunities for early prevention.

Dementia is a major public health concern currently affecting an estimated 35.6 million people worldwide.¹ With the increasing number of elderly individuals, the disability and cost associated with dementia are expected to rapidly increase in the next 40 years, affecting more than 115 million people by 2050.¹

Young-onset dementia (YOD) is usually defined as that occurring before 65 years of age. Because affected individuals are young, YOD is associated with severe consequences related to employment, social life, and even patients’ roles as parents and guardians. The diagnosis of YOD is very challenging because its initial symptoms are quite heterogeneous and often atypical; its pathological features differ from those of dementia in older individuals.^2,3 Early-onset Alzheimer dementia (AD) has also been suggested to have an inherited cause more often than late-onset AD, and mutations in several genes have been identified.^4-7

The identification of more generally applicable risk factors for YOD would be important for several reasons. First, identification might aid in the construction of preventive measures for the general population. Second, identification could increase understanding of the mechanism underlying this complex group of syndromes. Studies of late-onset dementia have identified lower cognitive performance as a risk factor,^8,9 whereas other studies have suggested that higher levels of education, physical activity, and engagement in social activities are associated with a reduced risk of dementia.^10,11 Whether these risk factors apply to the different subtypes of YOD is not known.

In this study, we evaluated premorbid cognitive performance, physical characteristics, socioeconomic factors, various diagnoses, use of medication, and the influence of heritable factors in young adulthood in relation to the risk of developing different YOD syndromes in a nationwide cohort of 497 844 Swedish men.

The cohort considered for inclusion consisted of all Swedish men conscripted for mandatory military service from September 15, 1969, through December 31, 1979 (N = 497 844). Of these men, 9360 (1.9%) were excluded; 9352 had unknown or extreme body weight (<40 or >170 kg) or height (<140 or >215 cm), and registry data were not useful for the remaining 8. Thus, 488 484 men were included in the present analyses. The Swedish Military Service Conscription Register has been described in detail previously.¹² During the years of conscription investigated in the present study (1969 to 1979), exemptions from conscription (2% to 3% of all Swedish men) were granted only to incarcerated men and those with severe chronic medical conditions or disabilities documented by a medical certificate.

During a 2-day baseline examination, all men in the cohort underwent standardized cognitive and physical examinations at 6 conscription centers in Sweden before being assigned to service in the Swedish armed forces. All men also underwent a physician-administered medical examination; medical disorders were diagnosed according to the Swedish version of the International Classification of Disease, Eighth Revision (ICD-8). Isometric knee extension strength was measured in the right leg using a dynamometer. Four tests were used to assess cognitive performance; they have been described in detail and evaluated previously.^13-15 Briefly, a logical test assessed the ability to understand written instructions; a word recollection test, the ability to determine which of 4 alternatives was the synonym of a given word; a visuospatial test, the capacity to identify a 3-dimensional object correctly using a series of 2-dimensional drawings; and a technical test, problem-solving capacity. Results of these tests were used to estimate overall cognitive performance by summing the normalized z scores of each test, expressed as a score ranging from 1 to 40 points. Detailed information about drinking habits and smoking (yes/no) was available for a subcohort of 23 696 men.

Information about diagnoses of dementia through December 31, 2011, was obtained from the Swedish National Hospital Discharge Patient Register (NPR). The NPR covered about 90% of all specialist inpatient care since 1971 and all outpatient specialist care since 2001. Potential risk factors for dementia were selected based on risk factors found in previous studies, data available, and quality of register data.¹⁶ The following ICD codes (from ICD-8, International Classification of Disease, Ninth Revision, and International Statistical Classification of Disease, Tenth Revision [ICD-10]) were used: F00.x, G30.x, and 290.x for the subjects’ parents (AD); F01.x (vascular dementia); F10.7A (alcohol dementia); F03.9 (dementia of unspecified type); F02.3 (dementia associated with Parkinson disease); G31.8A (Lewy body dementia); and F06.7 (mild cognitive impairment [MCI]). For all diagnoses of dementia except MCI, the diagnosis at each subject’s last clinical examination was used in further analysis. Subjects with MCI were excluded. Other diagnoses of interest included ICD codes 303 and F10.x, excluding F10.7A (alcohol intoxication); F11.x and 304 (other drug intoxication); F32.x and 311 (depression); I63.x, 433, and 434 (ischemic stroke); and I21.x and 410.x (myocardial infarction). The NPR has been validated previously with a positive predictive value of 85% to 95% for most diagnoses.¹⁶ We obtained information about drugs prescribed and expedited in Sweden since July 2005 through the National Drug Register using the Anatomical Therapeutic Chemical Classification codes N06A (antidepressants), N05A (neuroleptics), and A10 (antidiabetics). In later analyses, the presence of depression will include diagnosed depression or any drug therapy expedited for depression.

Information about total income and highest achieved educational level was collected 15 years after the conscription tests from the Statistics Sweden database. Highest achieved educational level was classified into the following 4 groups: elementary school only, 2 and 3 years of secondary school, and education at the university level. Subjects’ biological parents were tracked through the Statistics Sweden database, and information about diagnoses of dementia in parents was obtained using the ICD codes specified above. Information about deaths, immigration, and emigration occurring during the study period was obtained through record linkage with the National Cause of Death Register and the Statistics Sweden database. All information was then linked to subjects using the unique social security numbers assigned to all Swedish citizens. The study protocol was approved by the regional ethics board in Umeå and the National Board of Health and Welfare in Sweden.

Two authors (P.N. and L.-O.W.) retrieved total journal records of all specialized care, including radiographic examinations and laboratory tests, in 79 men diagnosed with YOD. Personal history, including first symptoms of the disease and radiographic examination results, was obtained in all cases; a Mini-Mental State Examination was administered in all but 1 case; and more extensive neuropsychological examinations were performed in 56 cases. In 51 cases, markers of dementia (tau and amyloid protein) were measured in cerebrospinal fluid. On the basis of information provided in the journal records, misdiagnosis was identified for 4 patients in a nonblinded comparison using the ICD-10 criteria. The remaining 75 diagnoses (95%) were judged to be valid, which is higher than the percentages reported for most other diagnoses in a previous evaluation of the NPR.¹⁶

Differences at baseline based on the diagnosis of dementia during the follow-up period were examined using analysis of variance with a Bonferroni post hoc test. Cox proportional hazards regression models were used to evaluate independent risk factors for YOD and death. For the outcome of YOD of any subtype, or death, all covariates in Table 1 were included in the models. For the outcome of YOD of each subtype, only covariates that were significant in univariate analysis according to Table 1 were included. The study end point for models evaluating risk factors for YOD was the first date of a registered dementia diagnosis, date of emigration, date of death, or December 31, 2011, whichever came first. The study end point for the outcome of death was date of emigration, date of death, or December 31, 2011, whichever came first. The proportional hazards assumption was checked graphically using Kaplan-Meier curves and log minus log plots. Population-attributable risk (PAR) was calculated from a Cox proportional hazards regression model including continuous variables entered as above/below the median using the PAR macro of Spiegelman et al¹⁷ for SAS (version 9.3 for Windows; SAS Institute, Inc). We used SPSS software (version 20.0 for PC; SPSS, Inc) for all other statistical analyses. P < .05 was considered significant.

The study cohort included 488 484 men with a mean age of 18 years who were followed up for a median of 37 (range, 0-41) years. During the follow-up, 506 men were diagnosed as having some type of YOD and 36 men were diagnosed as having MCI during a total of 1808 (per-subject mean, 3.3; median, 3; range, 1-30) hospital visits (median age at diagnosis, 54 years). In addition to cases of MCI, 19 cases of dementia associated with Parkinson disease (n = 13) and Lewy body dementia (n = 6) were excluded, leaving 487 cases of YOD analyzed in the present study. At the study end point (December 31, 2011), the prevalence of all cases of YOD per 100 000 men was 91.7. Subjects’ baseline characteristics are presented in Table 1 according to a diagnosis of YOD during follow-up and in the Supplement (eTable 1) according to the subtype of YOD. The cumulative incidence of different forms of YOD during follow-up is illustrated in Figure 1.

Nine independent risk factors for YOD (n = 487) were identified (Table 2), including alcohol intoxication (hazard ratio [HR], 4.82 [95% CI, 3.83-6.05]; PAR, 0.28), stroke (2.96 [2.02-4.35]; PAR, 0.04), use of antipsychotic drugs (2.75 [2.09-3.60]; PAR, 0.12), depression (1.89 [1.53-2.34]; PAR, 0.28), father’s dementia (1.65 [1.22-2.24]; PAR, 0.04), drug intoxication other than alcohol (1.54 [1.06-2.24]; PAR, 0.03), low cognitive function at conscription (1.26 per 1-SD decrease [1.14-1.40]; PAR, 0.29), high systolic blood pressure at conscription (0.90 per 1-SD decrease [0.82-0.99]; PAR, 0.06), and low height at conscription (1.16 per 1-SD decrease [1.04-1.29]; PAR, 0.16). The PAR associated with these 9 independent risk factors was 68% (95% CI, 39%-85%). Independent risk factors for the subtypes of YOD are presented in the Supplement (eTable 2). The risk of alcohol intoxication during follow-up was strongly associated with subjects’ drinking habits at conscription in a subcohort of 23 696 men (eg, 6-fold higher in men who drank beer daily than in men who never drank beer) (Table 3). In the same subcohort, smoking at conscription was higher (74.6% vs 59.4% [P = .01]) in those later diagnosed with YOD (n = 59).

We analyzed the joint effects of the 9 independent risk factors identified (Table 2) using men in the highest tertile of overall cognitive function with no risk factor as a reference (Figure 2). Height and systolic blood pressure were analyzed using the lowest and highest deciles, respectively, as risk factors. Men in the lowest tertile of overall cognitive function and with at least 2 risk factors were found to have a 20-fold increased risk of YOD during follow-up (HR, 20.38 [95% CI, 13.64-30.44]). Furthermore, the effects of an increased number of risk factors on the risk of YOD were significant for all 3 strata of overall cognitive function (P < .001 for trend for all), and the effects of increased overall cognitive function were significant for all 3 strata of risk factors (P < .01 for trend for all).

During follow-up a total of 26 105 men died. After including all covariates according to Table 1, a diagnosis of YOD was associated with the second highest risk for death (HR, 3.54 [95% CI, 2.77-4.53]), whereas alcohol intoxication was associated with the highest risk (HR, 5.95 [95% CI, 5.71-6.20]). Other strong independent risk factors (P < .001 for all) for death included myocardial infarction (HR, 2.88), stroke (HR, 3.44), and other drug intoxication (HR, 1.86).

In the present nationwide cohort study, the following 9 independent risk factors for YOD were identified in young men: alcohol and other drug intoxication, stroke, high blood pressure, low overall cognitive function, low height, dementia in the father, depression, and use of neuroleptics. These factors accounted for most of the YOD cases identified (68%). The effects of these risk factors were multiplicative. Most could be traced to adolescence and were potentially modifiable, suggesting excellent opportunities to identify subjects at risk for YOD early in life. Such efforts would be of great importance, given the consequences of YOD for the patient, family, and society, including a high risk of early death as demonstrated in the present study.

Based on risk estimates and PAR, hospital-treated alcohol intoxication was the single most important risk factor for YOD identified. Alcohol intoxication was not only associated with alcohol dementia but also an independent risk factor for vascular dementia and dementia of unspecified type. This increased risk could be traced to behaviors in adolescence because drinking habits at conscription were strongly linked to the risk of alcohol intoxication later in life. The impact of alcohol abuse was clearly not reflected in the number of cases diagnosed with alcohol dementia in the present study, which represented about 15% of all YOD cases, as in other studies.^18-20 This finding may have several explanations. One important issue is the lack of uniform diagnostic criteria for alcohol dementia.^21,22 The diagnosis of alcohol dementia is also controversial, partly owing to the potential reversal of neurologic and cognitive symptoms. Furthermore, whether alcohol dementia is caused directly by the toxic effects of alcohol²³ or is predominantly related to secondary thiamine deficiency resulting in Wernicke encephalopathy²⁴ is not clear. Altogether, our data suggest that the effect of alcohol abuse on the overall risk of YOD in men is at present severely underestimated.

Other risk factors that increased the overall risk of YOD independently included drug intoxication other than alcohol, depression and neuroleptic use, stroke, high systolic blood pressure, low cognitive function, low height, and dementia in the father. The risk for YOD increased steeply with increasing number of independent risk factors. Thus, men in the lowest tertile of cognitive function and with at least 2 other risk factors had a 20 times higher risk of YOD than did men in the highest tertile of cognitive function with no other risk factors. Many of the most important risk factors, such as alcohol abuse and cognitive function, could be traced to subjects’ characteristics in adolescence. Thus, the results of the present study suggest that men at high risk for YOD could be identified already in young adulthood based on the risk factors identified in the present study.

The diagnoses of dementia were retrieved from registers, and we were able to validate complete journal records for 79 patients in a nonblinded fashion with a high positive predictive value. This result was expected because all diagnoses were made at specialist departments and patients with diagnoses underwent a median of 3 investigations, suggesting proper follow-up in most cases. Recent studies have suggested that the initial symptoms of early-onset AD are often atypical and that its neuropathological characteristics differ from those of late-onset AD, increasing the diagnostic challenge.^3,25 The results of the present study may support these findings because many cases of YOD were diagnosed as dementia of unspecified type. However, the risk factors of these men were similar to those of subjects diagnosed with vascular and alcohol dementias, including frequent diagnoses of alcohol and other drug intoxication, cerebrovascular disease, and low cognitive function in young adulthood. Thus, the results of the present study suggest that typical cases of AD may represent a lower proportion of YOD compared with dementia diagnosed in later life.

Only a few risk factors were found for AD and frontotemporal lobe degeneration. Based on the findings of previous studies suggesting that early-onset AD and frontotemporal lobe degeneration have strong genetic components,^4,5,7,26,27 we hypothesized that these conditions would be linked to dementia in the parents more often than would other YOD subtypes. This hypothesis could not be confirmed in the present cohort, and the risk associated with dementia in the parents was generally low. For all cases of YOD, the PAR associated with dementia in the father was only 4%, and we found no association with dementia in the mother. Given the high PAR associated with the remaining 8 independent risk factors identified, our results support the view that heritable factors have a small influence on YOD.

Given that the present cohort included only men, our findings are not applicable to women, whose risk profiles (eg, early cerebrovascular disease and substance abuse) likely differ. Most likely other risk factors that might contribute to the risk of YOD are not captured in the present study, for example, traumatic brain injuries and physical activity. Furthermore, given the observational study design, we cannot disentangle whether certain covariates, such as depression, represent early symptoms of YOD or a true risk factor. The incidence of YOD seems to correspond to those of previous reports,^18-20 bearing in mind that men who were excluded from conscription (2%-3%), might be at higher risk for YOD owing to genetic disorders, such as Down syndrome. The strengths of the study include a nationwide cohort of men consisting of almost 500 cases of YOD; the extensive evaluation of risk factors, including any dementia diagnosis in the parents; and the use of national registers, resulting in no loss to follow-up, all of which increased the study’s external validity.

In summary, we identified 9 independent risk factors for YOD in a nationwide cohort of men. Collectively, these factors accounted for 68% of the YOD cases identified. The effects of these risk factors were multiplicative, and most were potentially modifiable and could be traced to adolescence, suggesting the existence of excellent opportunities for prevention. Such efforts would be of importance given the consequences of YOD for the patients, family, and society. The results of the present study also suggest that YOD may include a lower proportion of typical AD cases compared with later-onset dementia.

Accepted for Publication: May 22, 2013.

Corresponding Author: Peter Nordström, PhD, Department of Community Medicine and Rehabilitation, Geriatric Medicine, 90187 Umeå University, Umeå, Sweden (peter.nordstrom@germed.umu.se).

Published Online: August 12, 2013. doi:10.1001/jamainternmed.2013.9079.

Author Contributions: Dr P. Nordström had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: P. Nordström, A. Nordström.

Acquisition of data: P. Nordström, A. Nordström, Wahlund.

Analysis and interpretation of data: All authors.

Drafting of the manuscript: P. Nordström, A. Nordström.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: P. Nordström, A. Nordström, Eriksson.

Obtained funding: P. Nordström, A. Nordström.

Administrative, technical, and material support: P. Nordström, A. Nordström, Gustafson.

Study supervision: P. Nordström, A. Nordström, Wahlund.

Conflict of Interest Disclosures: None reported.

Funding/Support: The study was supported by grants from the Swedish Research Council and the Swedish Dementia Foundation.

Correction: This article was corrected on August 28, 2013, to fix the x-axis labels in Figure 2.