Study design and sample
Data were derived from the Leipzig Longitudinal Study of the Aged (LEILA75+), a population-based study on the epidemiology of dementia and MCI. Initially, a total of 1692 individuals aged at least 75 years residing in the Leipzig-South district were selected for participation. Of these, 1500 individuals were identified by systematic random sampling from an age-ordered list from the local registry office. In addition, 162 institutionalized individuals were included by systematic random sampling from an age-ordered list by the four institutions in the study area. Study details have been published elsewhere .
Of the 1692 invited individuals, 242 (14.2 %) refused, 57 (3.4 %) had died, 15 (0.9 %) were not traceable, and 113 (6.7 %) were shielded by relatives. Finally, the LEILA75+ cohort comprised 1265 (74.8 %) individuals. Non-participants did not differ from participants regarding age (U = 263553, p = .46), gender (χ2 = 0.40, p = .53), or marital status (χ2 = 5.03, p = .17).
Data were collected over a total observation period of eight years between January 1997 (begin of baseline) and April 2005 (end of follow-up 5). Follow-up assessments took place on average every 1.4 years. Structured clinical interviews at baseline and follow-up were conducted at participants’ homes by trained psychologists and physicians. In addition, structured interviews were held with proxies.
Assessment instruments and procedures
The main assessment instrument was the Structured Interview for the Diagnosis of Dementia of Alzheimer type, Multi-infarct Dementia and Dementia of other Aetiology according to DSM-III-R, DSM-IV, and ICD-10 (SIDAM) . The SIDAM comprises a neuropsychological test and a section for clinical judgment and third-party information on psychosocial impairment, including a 14-item scale to assess activities of daily living (SIDAM-ADL Scale). The SIDAM neuropsychological test consists of 55 items, including all 30 items of the Mini-Mental State Examination (MMSE) . Six areas of neuropsychological functioning are covered: 1) orientation (time and place); 2) memory (delayed verbal recall, delayed visual reproduction, questions on biography and history); 3) intellectual abilities (abstract thinking) and judgment (plausibility judgments, describing pictures representing actions); 4) verbal and calculation abilities (calculating serial sevens, spelling backward, backward digit span); 5) visual-spatial constructional abilities (copying figures); and 6) aphasia and apraxia (naming objects, reading and obeying a sentence, writing a sentence, and performing a three-stage command).
If it was not possible to administer the SIDAM, a structured proxy interview was offered including the Clinical Dementia Rating scale (CDR) .
SCD was evaluated prior to cognitive testing by asking the participant: “Do you have problems with your memory?”.
We identified depressive symptoms using the German version of the 20-item Center of Epidemiologic Studies Depression Scale (CES-D) .
A standardized interview provided information on sociodemographic characteristics.
Death dates were obtained from relatives or the registry office.
Definition of cases
SCD was assumed if participants were cognitively unimpaired and stated to have memory problems unrelated to an event or condition explaining the memory problems according to recent research criteria . Consequently, we excluded participants who met the following criteria: 1) a diagnosis of MCI or dementia, 2) a MMSE score below 26 points, 3) presence of a major psychiatric (e.g., major depression, anxiety), neurological (e.g., Morbus Parkinson) or medical condition (e.g., stroke) that could affect cognitive functioning.
We then built two subgroups of SCD: stable vs. unstable SCD. Stable SCD was assumed if SCD was consistently reported at every assessment until progression to MCI or dementia or, in case of non-progression, the last completed follow-up. By contrast, unstable SCD was assumed if SCD was not consistently reported at every assessment, but occasionally, until progression to MCI and dementia or, in case of non-progression, the last completed follow-up.
Individuals who never reported SCD at baseline and follow-up until progression to MCI or dementia or last completed follow-up without progression were considered controls (CON).
Diagnosis of MCI was based on Winblad criteria . They comprised absence of dementia, at most minimal impairment in instrumental activities of daily living, and evidence of cognitive decline in objective cognitive tests at least one standard deviation below age- and education specific norms  on one or more main domain of cognitive functioning as assessed by the SIDAM. We refrained from the criterion of the presence of a memory complaint to be able to consider any case of cognitive impairment in differentiation to CON and SCD .
Dementia status at any assessment was agreed at consensus conferences with physicians and psychologists according to DSM-IV criteria . The diagnosis was based on SIDAM results or, if proxy interviews only, on CDR data.
We dichotomously defined the outcome variable in progression (to MCI or dementia) and non-progression. Progression was assigned if participants were first-time diagnosed with incident MCI or dementia at follow-up. Non-progression was assumed if individuals had remained CON or SCD at every completed follow-up.
Group differences in socio-demographic and health characteristics at baseline between individuals with stable vs. unstable patterns of SCD and CON were analyzed applying Kruskal-Wallis-tests for continuous variables and χ2 tests for categorical variables.
We developed univariate (model 1) and multivariate Cox proportional hazards models (models 2–4) to assess the association of progression to MCI and dementia in stable vs. unstable patterns of SCD in reference to CON.
Multivariate Cox models were stepwise adjusted for age, gender, education (categorized into low, middle and high according to the CASMIN criteria , categorical) (model 2), depressive symptoms (continuous, CES-D) ) (model 3), and cognitive functioning (continuous, MMSE ) (model 4).
Kaplan-Meier survival analyses were applied to determine time to progression in regard to SCD stability. Time to progression was defined as the interval from baseline to first follow-up of progression to MCI or dementia or, in case of non-progression, date of last completed follow-up. Individuals without progression by follow-up 5 were treated as censored data. A Log-rank test was performed to assess the unadjusted difference in time to progression between individuals with stable vs. unstable patterns of SCD vs. CON. Furthermore, a stratified Cox regression-based test for equality of survival was performed adjusted for age, gender, and education.
For all analyses a significance level of α = 0.05 was applied. We used Stata/SE, version 13.0 (StataCorp LP, College Station, Texas/USA).