Study participants were part of the Study of Dementia in Swedish Twins (HARMONY) [30
]. The HARMONY study population included all twins in the Swedish Twin Registry [31
] who were aged 65 and older and alive during the telephone-screening phase. Although this study uses data from Swedish twins, it is not a twin study per se.
Dementia was ascertained through a two-step procedure that entailed an initial cognitive screening with a subsequent diagnostic workup of each suspected case. In brief, the TELE [32
] and the Blessed Dementia Rating Scale (BDRS) [34
] were used to screen for cognitive dysfunction. Twins who screened positive for suspicion of dementia (and their twin partners) were evaluated in person by a physician and nurse. Final diagnoses of dementia were determined by a multidisciplinary consensus board. Dementia was diagnosed according to the criteria in the Diagnostic and Statistical Manual of Mental Disorders IV [35
] and differentially diagnosed for AD versus vascular dementia using NINCDS/ADRDA [36
] and NINDS-AIREN criteria [37
]. The HARMONY study used two sources of information to estimate the age of dementia onset: informant reporting during an in-depth semistructured interview and medical records [38
The sample for the present study (N = 238) consisted of all HARMONY twins who (1) were diagnosed as having dementia, (2) had a hard copy of their CT scan (performed as part of the clinical phase of dementia assessment) that could be assessed by the CT raters, and (3) did not meet any of the exclusion criteria (see ). Exclusion criteria included missing information on timing of depressive episodes (N = 4) or first depression onset more than six months after the scan date (N = 2).
Of the 238 CT scans scored by the raters, 56 were excluded during the rating process. Four scans were excluded due to the poor quality of the scan, 45 because there was evidence of major stroke (i.e., middle cerebral artery stroke, posterior cerebral artery stroke), 2 because of hydrocephalus, and 5 because of other major brain problems (e.g., evidence of major brain surgery or traumatic brain injury). Those with evidence of a minor stroke (i.e., small lacunar infarct) were not excluded. The final sample included white matter hypodensity and striatal hypodensity ratings for 182 individuals.
Diagnostic neuroimaging utilized CT scans because at the time of the clinical diagnostic assessment insufficient numbers of participants lived close to MRI centers. Twins completed CT scans at their most convenient participating CT center. All CT technicians were given standardized instructions to perform a CT of the brain with standard slices and noncontrast enhancement. The slices were to be four to five millimeters from the base of the skull and eight to ten millimeters from the pars petrosa ossis temporalis. Contrast enhancement was performed if needed for clinical reasons. The diagnostic protocol allowed individuals who had a CT within six months prior to the clinical workup to provide a copy of that scan instead. After obtaining a clinical read, CT scans were deidentified and made available for research purposes.
2.3. Visual Rating of CT Scans
B.M.S. and J.L.G. served as the CT raters. Raters were blind to clinical diagnosis and any demographic information, including which scans were from twin pairs, as well as the age, gender, and zygosity of the scanned individual. If either rater determined that an image quality was unacceptable, it was excluded from the analyses. If there was evidence of a major stroke, such as middle cerebral artery or posterior cerebral artery stroke, the CT was not included in the analyses. Individuals with small lacunar infarcts, however, were not excluded. CT scans that indicated hydrocephalus or other severe neuropathology, such as evidence of a brain tumor, major brain surgery, or traumatic brain injury, were also excluded from the analyses.
2.3.1. Frontal Deep White Matter Hypodensities
Deep white matter hypodensities in the frontal lobe were rated on a modified version of the Age-Related White Matter Changes (ARWMC) Scale [39
], where 0 = an absence of hypodensities, 1 = one focal hypodensity (≥5
mm), 2 = more than one focal hypodensity, 3 = confluent hypodensities, and 4 = confluent hypodensities with additional discrete focal hypodensities. Focal hypodensities were defined as discrete hypodensities greater than five millimeters in size. Confluent hypodensities were present when discrete hypodensities could not be separately defined. Deep white matter hypodensities (DWMH) were defined as WMH that were located medially from the sulci. The ratings of frontal DWMH in the right and left hemisphere were highly correlated (Spearman's rho = 0.92, P
< 0.0001). Therefore, summary scores for frontal DWMH used the maximum score for the right and left hemispheres.
2.3.2. Subcortical White Matter Hypodensities
The density of the white matter in the internal capsule and subinsular region (comprised of the external capsule, claustrum, and extreme capsule) was compared to the density of homogenous areas of white matter in the frontal lobe to determine if there were relative hypodensities evident in either of these components. The location, side (left versus right hemisphere), and rating of the hypodensities within these regions were indicated, with hypodensities rated on the modified version of the ARWMC Scale. The ratings of subcortical WMH in the right and left hemisphere were significantly correlated (Spearman's rho = 0.79, P < 0.0001); a summary subcortical WMH score was created using the maximum score for the right and left hemispheres.
2.3.3. Basal Ganglia-Thalamic Hypodensities
The raters also examined each scan for gray matter hypodensities in the basal ganglia and thalamic regions. The homogeny of the gray matter in the striatum (caudate nucleus and lentiform nucleus), substantia nigra, and thalamus were compared to one another to identify hypodensities. The total number of hypodensities in the striatum on both right and left hemispheres was summed.
2.3.4. Interrater Reliability
Each CT scan was scored separately by both raters. Interrater reliability for bilateral WMH was substantial (weighted kappa, right WMH = 0.92; weighted kappa, left WMH = 0.89) and slightly lower though adequate for bilateral basal ganglia-thalamic hypodensity ratings (weighted kappa, right = 0.73; weighted kappa, left = 0.62). In cases of disagreement, the scans were rerated conjointly, and the consensus rating was used for analysis.
History of depression, comorbid depression, and estimated age of depressive episodes were determined using information from four sources: (1) the national computerized Inpatient Discharge Registry (IDR), (2) the national registry of inpatient psychiatric hospital services, (3) medical history provided by an informant, and (4) medical records.
The Swedish Twin Registry is linked to the national computerized Inpatient Discharge Registry that records all inpatient hospital discharges in Sweden. Discharge diagnoses use International Classification of Disease (ICD) codes. If the discharge date was prior to 1969, the ICD-7 coded depression diagnoses as 302 (involutional melancholia), 314 (depressive neurosis), and 790.2 (other recurrent depressive disorder). If the discharge date was 1969–1986, the ICD-8 coded depression diagnoses as 296.0 (involutional melancholia), 298 (reactive depressive psychosis), 300.4 (depressive neurosis), and 790.2 (other recurrent depressive disorder). If the date of discharge was 1987–1996, the ICD-9 coded depression diagnoses similar to the ICD-8, with 296.2 (depressive psychosis), 296.3 (recurrent depressive psychosis), 296.82 (atypical depression), 300.4 (dysthymia), and 311 (depression, NOS). From 1997, the ICD-10 depression diagnoses included F32 (depressive episode), F33 (recurrent depressive disorder), and F34.1 (dysthymia). There were 11 participants with discharge diagnoses of depression in the Inpatient Discharge Registry.
The Swedish Twin Registry has also been linked to a national registry of inpatient psychiatric hospital services that was maintained between 1967 and 1983. For each person entered in this registry, there is a record of the discharge diagnosis and the date of hospitalization. All diagnoses are given in terms of an ICD-8 diagnosis (see above). There were two participants with diagnoses of depression in the inpatient psychiatric hospital services registry, both of whom also had a diagnosis of depression in the Inpatient Discharge Registry.
Thus, a total of 11 participants in this study had at least one depression-related discharge diagnosis in the IDR between the years of 1964 and 2004, a span of 40 years, or between 1967 and 1984 in the inpatient psychiatric hospital services registry. The most common depression-related discharge diagnosis was depression not otherwise specified, followed by dysthymia, recurrent depressive disorders, and depressive episodes of mild, severe, or other characteristics. Discharge diagnoses in both the IDR and the psychiatric hospital discharge registry that were regarded as not depression-related included bipolar affective disorder and manic-depressive reaction, manic or unspecified type, schizoaffective disorder, and unspecified mood disorders.
Medical history reported by an informant was collected during the clinical evaluation. The history included whether the individual had any history of “major depressive disorder” or “reactive depression,” and if so, the date or dates of onset. The first depressive episode according to the medical history used the earliest date recorded for the onset of a depressive disorder and the most recent episode of depression used the latest date recorded for a depressive episode.
Medical records, ordered during the clinical evaluation phase, were coded by the assessment team to reflect whether the twin had been diagnosed with depression. Records typically go back approximately ten years before the clinical evaluation for dementia. Thus, these records were most helpful in determining whether late-life depression was present. Data included the onset and dates of depression.
Use of antidepressant medications was also coded from the medical records and medical history. A total of 31 individuals were prescribed antidepressant medication, but did not have any other information indicating that they had ever received a diagnosis of depression. These individuals were coded as not having a history of depression, as it is possible that these medications were prescribed for reasons other than depression.
The age of the first episode of depression was determined using the age at the earliest reported occurrence of depression across all sources. Individuals with a first episode of depression at age 60 or older were considered to have late-onset depression (LOD). Individuals who had prior episodes of depression including at least one episode of depression at age 60 or over were considered to have a history of late-life depression (LLD). In sum, a total of 132 (72.5%) individuals did not have a history of a depression diagnosis and 50 (28.5%) individuals had a history of a depression diagnosis. Of these individuals, 36 had LOD, having their first depressive episode after age 60, with a mean (SD) age of first depressive episode of 74.2 (7.0) years. Of the 14 individuals who had their first depressive episode before age 60, nine individuals had an episode of depression occurring after age 60. Thus, 45 of the 182 participants were considered to have LLD. Those with an early episode of depression that did not recur in later life (N = 5) were not included in further analyses.
2.5. Cerebrovascular Disease
Cerebrovascular disease risk factors, as indicators of CVD risk, were examined as potentially important covariate. Data on hypertension, diabetes, atrial fibrillation, peripheral artery disease, transient ischemic attack, and coronary artery disease (CAD) indicators were extracted from the participant's medical records and coded by the assessment team to reflect whether the individual had a history of any of these risk factors. Coronary artery disease was considered present if the individual had a history of myocardial infarction, angina, or heart failure.
2.6. Other Covariates
Additional covariates included age at the time of the CT scan, dementia duration (i.e., age at the time of the scan compared with age of dementia onset), gender, zygosity, and number of years of education completed. Because of the cross-sectional study design, particular attention was paid to the duration of dementia at the time of the CT scan (calculated by subtracting age of dementia onset from the age at the CT scan).
Associations between history of LOD (no depression versus LOD) and all potential demographic and medical confounders were initially examined to determine factors to include as covariates in multivariate models. Chi-square tests were used to examine whether gender, zygosity, or risk factors for CVD (hypertension, diabetes, atrial fibrillation, peripheral artery disease, transient ischemic attack, and coronary artery disease indicators) differed by history of LOD. One-way ANOVA was used to determine whether age at CT scan, age of dementia onset, duration of dementia at the time of the CT scan, or level of education differed by presence versus absence of LOD.
Kruskal-Wallis chi-square was used to examine whether frontal deep white matter or subcortical white matter score varied by age at CT scan or dementia duration. Fisher's Exact Test was used to determine whether frontal deep white matter or subcortical white matter score differed by gender. Simple linear regression was used to test whether number of striatal hypodensities varied by age, gender, or dementia duration. We took a conservative approach to determining which potential confounders should be included in the final model, including covariates with an association of P < 0.15.
Logistic regression for dichotomous outcomes controlling for gender and history of TIA was used to examine the association between LOD and the presence of any frontal lobe deep WMH (focal and confluent WMH versus no WMH), as well as between LOD and the presence of confluent frontal lobe deep WHM (confluent WMH versus no WMH and focal WMH only). To analyze the relationship between LOD and level of subcortical WMH (0 = no WMH, 1 = one WMH, 2 = more than one WMH, 3 = confluent WMH), ordinal logistic regression was used, controlling for age at CT scan and zygosity. The association between numbers of striatal hypodensities and LOD was examined using one-way ANOVA; there were no demographic or medical covariates associated with striatal hypodensities. All of the analyses testing white matter associations with LOD were reanalyzed using LLD. Finally two sets of analyses were performed, the first using all participants with dementia and the second using only individuals with Alzheimer's disease.
Seven complete twin pairs were included in these analyses. To account for the possibility that these seven pairs had correlated data, we also used generalized estimating equations (GEE) to examine the association between the presence of frontal deep WMH and LOD, the amount of subcortical WMH and LOD, and the number of striatal hypodensities and LOD for both the total sample and the Alzheimer's disease only sample. GEE accounts for the lack of independence between twin observations.