Are Late- and Very-Late-Onset Schizophrenia Precursors of Dementia?

Abstract

Background/Objective: To examine whether patients with late-onset schizophrenia (LOS) and very-late-onset schizophrenia-like psychosis (VLOS) are at a higher risk of developing dementia than patients with early-onset schizophrenia (EOS). Methods: A random sample of incident cases of patients with schizophrenia with an age of onset after age 40 (LOS) and age 60 (VLOS) was selected from a psychiatric case register covering all psychiatric services within the city of The Hague, The Netherlands, between 1997 and 2012. Schizophrenia diagnosis and age of onset were audited by systematic review of all case notes up to 2019. Patients who were initially misclassified as LOS/VLOS in the registry but, after this audit, had an age of onset before age 40, were classified as EOS. The risk of developing dementia was compared between the groups using a logistic regression model with correction for age, gender, and stroke as possible confounders. Results: Our study groups consisted of 74 patients with EOS, 81 with LOS, and 25 with VLOS. Dementia was present in 10 patients at follow-up, 7 out of 106 (6.6%) patients in the combined LOS and VLOS groups, and 3 out of 74 (4.1%) patients in the EOS group. However, after correction for confounders, logistic regression showed no statistically significant higher change in the development of dementia in LOS nor VLOS patients compared to EOS patients, nor with age of onset as a continuous variable. Conclusions: Patients with LOS and VLOS are not at a higher risk of developing dementia compared to patients with EOS. These results do not support the hypothesis that LOS and especially VLOS, are precursors of dementia.

1. Introduction

Schizophrenia most commonly presents itself in early life, however at least 20% of patients have an onset after the age of 40 [1]. An international conference meeting agreed to define late-onset schizophrenia (LOS) with an onset between the ages of 40 and 60 [2]. It was disputed whether schizophrenia can have an age of onset after 60, mainly because some authors suggested that psychotic symptoms arising at that age are early symptoms of dementia. Therefore, this conference meeting could only agree on the definition of very-late-onset schizophrenia-like psychosis (VLOS) with onset after the age of 60.

We are aware of eight articles that studied the relation between age of onset of schizophrenia and the risk of developing DSM- or ICD-defined dementia. The first study found no difference in the development of dementia between 28 patients with late-onset schizophrenia (defined as age of onset after 44 years) and 48 patients with depression [3]. A prospective study found a significantly higher 5-year-incidence of dementia of almost 50% (9 out of 19 patients who completed a 5-year follow up) among patients with late-onset schizophrenia (defined as age of onset after 50 years) compared with 0% (0 out of 34) among healthy control subjects [4]. A retrospective nation-wide register study included 1206 patients with LOS (defined as the age of first contact as patient after 40 years) and 409 patients with VLOS (first contact after 60 years) and found that the rate ratio of developing dementia compared to the general population was raised significantly in both LOS (2.36; 95% CI 1.54–3.62) and VLOS patients (2.21; 95% CI 1.93–5.14) [5]. A population-based cohort study with a follow-up period of 18 years included more than 20,000 individuals with schizophrenia and found an adjusted incidence rate ratio (IRR) for the risk of developing dementia in all patients with schizophrenia compared to the general population of 1.71 (95% CI: 1.60–1.82) after adjusting for age, sex, calendar period, comorbidities, and substance abuse [6]. This was not confirmed in a prospective community-based cohort study including 435 men with a non-organic psychotic disorder, with no significant differences in risk of dementia between age of onset before 65 years (adjusted hazard ratio 2.73; 95% CI 2.34–3.18) and age of onset after 65 years (HR 2.22; 95% CI 1.74–2.84) [7]. Another prospective study in the community included 61 patients with VLOS (age of onset > 60 years) and found an adjusted HR of 2.67 (95% CI 1.82–3.91) for developing dementia compared to the group without incident schizophrenia) [8]. The largest cohort study included 15,409 participants diagnosed with VLOS (onset after 60 years of age) and found a HR for developing dementia of 4.22 (95% CI 4.05–4.41) compared to 154,090 matched participants without VLOS [9]. Finally, the most recent and largest cohort study included 39,022 patients with a diagnosis of psychosis and found that the risk of dementia varied with age at psychosis onset, with the rate being highest at late-onset (adjusted HR 5.16; 95% confidence interval 4.59–5.81) [10].

We conclude that the literature on the relation between (V)LOS and dementia is inconclusive, with only one study including patients with both LOS and VLOS defined according to the international consensus statement criteria, and that study found no difference in the risk of developing dementia between LOS and VLOS patients [5]. Most studies found an elevated risk of developing dementia in all patients with late-onset schizophrenia with a higher risk compared to patients without schizophrenia. In contrast to larger cohort studies based on registry data where the validity of the age of onset is not checked, we rigorously verified age of SSD onset in patient records. We also checked the validity of the dementia diagnosis in patients’ records, in contrast to large registry studies mentioned above. Diminished cognition is a core feature of schizophrenia, and therefore the boundary between degenerative dementia and cognitive deficits due to severe forms of schizophrenia is often difficult to determine.

The objective of the present study was to compare the incidence rate of dementia among late-onset schizophrenia and very-late-onset schizophrenia-like psychosis to early-onset schizophrenia. We chose EOS as reference group because of the debate on whether LOS, and especially VLOS, is not schizophrenia but a precursor of dementia. We hypothesized that both late-onset schizophrenia and very-late-onset schizophrenia-like psychosis increase the risk of later development of dementia when compared to early-onset schizophrenia. Cut-off points are arbitrary and we also analyzed our data with age of onset as a continuous variable.

2. Materials and Methods

2.1. Case-Identification

Since 1980, all patients in the municipality of The Hague, The Netherlands, who received psychiatric care, have been registered in the Longitudinal Psychiatric Register (LPR). The LPR is a database uploaded from the patient registration systems of the Parnassia Group, the largest psychiatric teaching hospital in The Netherlands. Records of all inpatient, outpatient, day, and psychiatric residential care in the catchment area of The Hague are kept in the LPR. It contains information on demographical data, DSM classifications, and all psychiatric service contacts for each patient in the LPR from 1997 onwards.

All patients in the LPR with a first DSM main classification of schizophrenia spectrum disorder (SSD) after the age of 40, between 1 January 1997 and 1 October 2012, were included, as part of an ongoing study. Diagnoses considered as part of SSD were schizophrenia, schizophreniform disorder, schizoaffective disorder, and delusional disorder. For the present study, we only selected patients with a diagnosis of schizophrenia, as described later. Only incident cases from 1997 onwards were included due to records between 1980 and 1997 being incomplete and therefore not usable. The administrative systems of the Parnassia Group were reorganized in October 2012 and data beyond that date have not yet been imported into the register. Our query of the register over the 1997–2012 period yielded 1058 cases of patients aged 40–60 years (LOS) and 507 cases of patients aged 60 or older (VLOS) with a first administrative diagnosis of SSD. This dataset constituted a complete, population-based list (n = 1566) of the administrative incidence of SSD among citizens of The Hague aged 40 and older. From this dataset, we extracted a random sample of 25% (n = 259) of the LOS cases and 50% (n = 255) of the VLOS cases. This dataset constituted an age-stratified random sample for which all (digital) case notes were requested from the Parnassia Group’s electronic case file system (these data are not included in the register’s data warehouse). From the random sample, digital case notes were available for 156 (LOS) and 88 (VLOS) cases respectively (total n = 244), mainly persons treated in the years after 2004, when the Parnassia Group implemented electronic case notes. This dataset was constructed in 2018 and digital notes were checked up until 2019.

Our study complied with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975 and later revisions and was approved by the Institutional Scientific Committee. According to national law, informed consent of patients was not necessary as the data were collected in routine care.

2.2. Diagnosis and Age of Onset

To identify true incident cases of SSD among the list of cases (n = 244) identified in the data warehouse and list their true age of onset (defined as first-in-lifetime clinical diagnosis of SSD), we examined all digital case notes (including scanned medical correspondence summarizing treatment before the electronic file was created) according to a pre-specified protocol.

All 244 cases were reviewed by two researchers (BC, BB) and inconclusive cases were discussed with a psychiatrist (RK) with extensive expertise in geriatric psychiatry. Interobserver agreement was defined as (1) agreement on the presence of schizophrenia and other psychiatric disorders according to DSM-IV criteria, (2) agreement on the categorization of a patient in the same age of onset subgroup, and (3) agreement on the presence of dementia as a comorbid disorder. Interobserver agreement was reached in all cases concerning the SSD diagnosis, age of onset, and the presence of dementia. We determined age of onset as the youngest age at which, in retrospect, a patient fulfilled DSM-criteria for any of the SSDs based on all available information in the electronic patient records.

The case-note analysis revealed a substantial number of cases for whom the year of onset had been recorded incorrectly in the register. Many cases had developed SSD at a much younger age than appeared in the administrative records but had been invisible to the register, possibly because they lived outside our catchment area, or because they were not under psychiatric treatment in the catchment area. Using the case notes to reclassify such cases resulted in a number of cases who appeared to have (V)LOS, but in fact were EOS cases from before 1997. Our EOS group consisted therefore of patients with delayed contact with psychiatric services in our region.

The EOS group was not formed by a separate dataset from the LPR because it was presumed that patients with EOS would, by definition, be younger than 40 years old in 1997, and therefore would be 60 years old at most at the time of inclusion in our study. We extracted a small random group of 50 patients with EOS from the LPR, and almost all were indeed under the age of 60 at the time of inclusion in our study, with no patients with a final diagnosis of dementia. Patients within the random samples of SSD who had a definite diagnosis of schizophrenia and onset before the age of 40 formed the EOS group. The other schizophrenia patients were divided into LOS, with onset between 40 and 60 years and VLOS with onset after age 60, according to the international consensus statement [2]. The diagnosis of dementia was made by or under supervision of a psychiatrist, based on DSM-IV criteria for dementia. MRIs and a neuropsychological investigation by a clinical psychologist where found in a limited number of patients’ case notes, and therefore information of the type of dementia was very often lacking. We did record a history of stroke but this is not enough to distinguish between vascular and neurodegenerative dementia.

2.3. Statistical Analysis

We used χ² tests, Fisher’s exact test, and ANOVA analyses, including post hoc analyses, to compare the demographic and clinical variables between the three different age-of-onset groups. The risk of dementia among patients with EOS, LOS, and VLOS was calculated using a logistic regression model with dementia as the outcome of interest. In order to be able to compare our results to other studies that used just two age-of-onset groups, we also analyzed our results with different ages as cut off. In addition, a logistic regression model was used with age of onset included as a continuous variable instead of a categorical one. Because of this, patients were no longer separated into three age of onset subgroups. In all regression analyses, gender, age, and stroke were included as possible confounders. Age was particularly important since it is a well-known risk factor for dementia, as is stroke [11]. We could not control for important variables such as antipsychotic treatment or substance abuse, as these were not registered systematically in the patients case notes. Analyses were performed using IBM SPSS for Windows version 25.

3. Results

Of the 244 patients, 64 patients were excluded due to the absence of a schizophrenia diagnosis (schizoaffective disorder (n = 44), bipolar disorder n = 10, delusional disorder n = 3, drug abuse n = 3, psychotic depression n = 2, other n = 2). This ultimately resulted in three schizophrenia subgroups consisting of 74 patients with EOS, 81 with LOS, and 25 with VLOS, respectively. The characteristics of the three age-of-onset groups are presented in Table 1.

Table 1. Baseline demographical and clinical characteristics.

	EOS	LOS	VLOS	EOS vs. LOS	EOS vs. VLOS	LOS vs. VLOS
	(N = 74)	(N = 81)	(N = 25)
Gender, Male ‡	35 (47.3%)	47 (58%)	9 (36%)	Χ2 = 1.786 (df = 1)	Χ2 = 0.966 (df = 1)	Χ2 = 3.719 (df = 1)
				p = 0.181	p = 0.326	p = 0.054
Age †	64.6 ± 11.2	65.8 ± 8.4	82.8 ± 8.1	t = −0.745 (df = 135)	t = −8.898 (df = 104)	t = −7.504 (df = 97)
				p = 0.458	p < 0.001	p < 0.001
Age at onset schizophrenia †	28.6 ± 7.9	46.9 ± 5.0	68.3 ± 7.7
Diagnosed with dementia ‡	3 (4.1%)	1 (1.2%)	6 (24%)	Χ2 = 1.223 (df = 1)	Χ2 = 8.996 (df = 1)	Χ2 = 16.053 (df = 1)
				p = 0.349	p = 0.007	p < 0.001
Stroke ‡	3 (4.1%)	1 (1.2%)	3 (12%)	Χ2 = 1.223 (df = 1)	Χ2 = 2.072 (df = 1)	Χ2 = 6.097 (df = 1)
				p = 0.349	p = 0.167	p = 0.040

N = number, EOS = early-onset schizophrenia, LOS = late-onset schizophrenia, VLOS = very-late-onset schizophrenia-like psychosis, df = degrees of freedom. † Student’s t-test; ‡ X²-test or Fishers exact test.

As can be seen in Table 1, there was no significant difference regarding gender between the three age-of-onset groups, but the mean age (at time of first analyses in 2018) was statistically significant higher in the VLOS group versus both other age-of-onset groups. In total, 10 patients were diagnosed with dementia, 4 out of 155 (2.6%) patients in the combined EOS and LOS groups and 6 out of 25 (24%) patients in the VLOS group. In these 10 patients, the time between the diagnoses of schizophrenia and dementia varied between 1 and 44 years, with a median of 8 years.

In the first logistic regression analysis, we used age of onset as a categorically defined variable (EOS, LOS, VLOS) and corrected for differences in age, gender, and stroke. We found that the crude OR for developing dementia in VLOS patients, compared to EOS patients, was 7.5 (95% CI: 1.7–32), but after correction for confounders the direction reversed and the risk became non-significant (0.42, 95% CI: 0.06–2.7). The crude OR for developing dementia in VLOS patients compared to LOS patients was 25.3 (95% CI: 2.9–222.4), after correction for confounders the direction also reversed and became also nonsignificant (0.15, 95% CI: 0.01–1.98).

A second logistic regression analysis was based on age of onset as a continuous variable. This is presented in Table 2. There was a significant association between the age at which the first schizophrenia diagnosis was made and the development of dementia (p = 0.015) before correction for confounders. When corrected for age of onset, age, gender, and stroke, however, it became non-significant (p = 0.850). As expected, there was a significant association between age and the development of dementia (p = 0.027). Contrary to our expectations, a history of suffering with stroke significantly lowered the risk of developing dementia (OR 0.075; p = 0.009). No significant association was demonstrated between age of onset and gender with the risk of subsequently developing dementia.

Table 2. Logistic regression model with age of onset as continuous variable.

Factor	B	SE	Wald	p-Value	OR	95% CI
Gender	1.19	0.916	1.695	0.193	3.30	0.55–19.9
Age	0.094	0.042	4.870	0.027	1.10	1.01–1.19
Age of onset	0.004	0.023	0.036	0.850	1.004	0.96–1.05
Stroke	−2.58	0.984	6.900	0.009	0.08	0.011–0.52

B = regression coefficient, SE = standard error, OR = odds ratio, 95% CI = 95% confidence interval.

To be able to better compare our study to the aforementioned studies, we performed a sensitivity analysis with the same logistic regression model using different ages of onset to divide EOS from LOS. In these analyses, there was no distinction between LOS and VLOS (i.e., they were considered as one group). As shown in Table 3, the results of these analyses did not change the outcome of this study.

Table 3. Results of the logistic regression models with dementia as outcome with different cut-offs for ages of onset defining LOS.

Age Cut-Off	B	SE	Wald	p-Value	OR	95% CI
40	7.32	1.386	0.279	0.597	2.08	0.14–31.5
45	1.07	1.480	0.526	0.468	2.93	0.16–53.2
50	−4.01	0.814	0.242	0.174	3.57	0.57–3.3
60	−8.62	0.834	1.067	0.302	0.42	0.08–2.2

B = regression coefficient, SE = standard error, OR = odds ratio, 95% CI = 95% confidence interval.

4. Discussion

This study showed that patients with LOS or VLOS are no more likely to develop dementia than patients with EOS. At first instance the percentage of patients developing dementia was almost 10-fold higher in the VLOS group when compared to the EOS and LOS groups. VLOS patients were significantly older (82.8 years) than LOS/EOS (≈65 years) and this large age gap strengthens the clinical importance of adjustment for age. After correction for age and other confounders, the risk reversed direction and became nonsignificant. Age of onset as a continuous variable was also initially significantly associated with a higher risk of developing dementia (p = 0.015) but also became clearly non-significant (p = 0.850) when corrected for age, gender, and stroke. This result suggests that LOS and VLOS may not be early manifestations of a neurodegenerative disease. The correlation between VLOS and dementia is strong but fully explained by age.

4.1. Comparison with Other Studies

VLOS is only separated from LOS in one of the 10 aforementioned studies with the age of onset of schizophrenia as a variable [5]. In that study, both LOS and VLOS had a higher risk of dementia compared to the general population without significant differences in dementia risk between LOS and VLOS. This is in line with the results of our study, which also compares the same age-of-onset groups. Another large population-based cohort study found evidence that patients with LOS have a significantly higher risk of developing dementia compared to EOS [6]. The largest cohort study only analyzed a cut-off of 45 years for onset of a psychotic disorder and found a higher elevated risk with onset after 45 years, compared to onset before 45 years [10]. Two large studies aimed at VLOS patients only, found a significantly increased risk for dementia compared to a community population without SSD, but did not include a group with LOS or EOS patients [8,9]. This means that the only two studies looking specifically at VLOS cast serious doubts about the idea of VLOS being an early manifestation of dementia, in contrast to LOS. Two other studies found an elevated risk for dementia in all older patients with a psychotic disorder, compared to community populations [7,10]. A meta-analysis of six large cohort studies including more than 5 million participants with schizophrenia, of which over 200,000 developed dementia, found that patients with schizophrenia were associated with a significantly greater risk of dementia incidence (RR 2.29; 95% CI 1.35–3.88) than those without [12]. Another large cohort study including 74,170 individuals with schizophrenia, published after the aforementioned meta-analysis, also found that patients with schizophrenia had an increased risk of receiving a diagnosis of dementia compared to patients without a severe mental illness [13]. Increased incidence of dementia in patients with schizophrenia showed the greatest increase at younger ages but was apparent across the older adult life span. Our results are also in line with the following: a meta-analysis that did not find a decline of cognitive deficits over a 1–6-year period among older individuals with schizophrenia [14] and reviews of longitudinal studies which showed either stability in cognitive impairment during 1 year follow-up or lack of evidence of deterioration [15,16].

If VLOS would be an early manifestation of dementia by definition, we would expect the diagnosis of dementia being made within a few years after the diagnosis of schizophrenia was made. In both studies with a high percentage of patients developing dementia of (almost) 50%, this percentage was achieved after 5 or even 10 years of follow-up [3,4]. The mean age of the schizophrenia patients was 73–75 years in both studies and, at that age, a high risk of developing dementia also exists in non-schizophrenia patients. The 8-year incidence of developing dementia was only 1.7% in the LOS group and 4.4% in the VLOS group in the Kørner et al. study [5], suggesting that even in VLOS patients after 8 years, more than 95% of all patients did not develop dementia. The largest study with VLOS defined according to the international consensus criteria (≥60 years of age) found the strongest risk for dementia in the first year following VLOS diagnosis, suggesting that psychotic symptoms in patients with dementia may have strong similarities with schizophrenia symptoms and both conditions may be difficult to separate [9]. In this study, the risk for dementia attenuated over time but dementia rates remained higher in participants with VLOS for up to 20 years of follow-up.

Although it was not surprising that our study confirmed age as a predictor of dementia, we were surprised by the lower risk of dementia in stroke patients. Stroke is a well-known risk factor for dementia, as we stated before [11]. We had only 10 patients suffering from dementia and 7 patients with a history of stroke, so this may just be an artefact.

4.2. Strengths

Strengths of our study included the validation of the schizophrenia diagnosis and age of onset by checking the electronic patient records. To the best of our knowledge, we are the only large study regarding this subject that checked the validity of age of onset. After the study by Kørner et al. [5], the Danish Psychiatric Central Research Register has been validated for schizophrenia diagnoses with correct diagnoses in 89.7% to 97.5% of cases [17]. However, it has never been validated for age of onset which has proven to be very different between the registration in the LPR and the true date according to the electronic patient records in our study. The complete attenuation of the LOS/VLOS–dementia association after adjusting for age is a unique result of our study.

As it turned out, many cases in the VLOS and LOS datasets in our study had an earlier age of onset than their registration date in the LPR suggested, with only 37.8% of patients having a correct age of onset status of LOS and 28.4% having a correct age of onset status of VLOS in the register. This was partly the result of the LPR only containing registrations from 1980 onwards. As such, patients who had already been diagnosed with schizophrenia before 1980 were only registered on a date after 1980, on which they had their next contact with a psychiatrist. Another factor was relocation into the municipality of The Hague of patients who had already been diagnosed with schizophrenia elsewhere. Reviewing patient files corrected both types of information bias.

Another strength is that, in contrast to all but one other study [5], we had separate groups for LOS and VLOS defined by ages of onset according to an international consensus statement [2]. This allowed us to draw conclusions for VLOS independently of LOS, which is valuable considering VLOS is still the area of most debate. In order to be certain to correctly identify someone as having schizophrenia or dementia, only DSM-IV criteria were used. Because cognitive deficits are considered relatively common among patients suffering from schizophrenia, the distinction with dementia may be difficult. By only diagnosing dementia if that person fulfilled the DSM-criteria after formal neuropsychological investigation or according to or under supervision of a psychiatrist, the risk of wrongly diagnosing someone with dementia has been minimized in our study.

4.3. Limitations

Our study also had some limitations, the main one being its retrospective nature with smaller amounts of patients included than the other case register studies regarding this subject [5,6,9,10]. However, as mentioned, other case register studies did not validate age of onset. Because of the retrospective nature, we were of course limited by the data entered in the electronic patient records by clinicians. For example, the type of dementia was lacking, resulting in a heterogenous group of DSM-IV-based dementia diagnoses.

Another limitation is the lack of a control group outside of schizophrenia patients, representative of the general population. However, patients with other diagnoses than schizophrenia are usually less likely to receive life-long treatment and therefore are not easily comparable for their risk of developing dementia. We also did not assess the risk for dementia in patients with other psychiatric diagnoses, as this has been found in other studies [18]. In addition, our EOS group was selected in a different way compared to the LOS and VLOS groups, because that method would have resulted in a group that was too young (<60 years) to have a significant risk of dementia. This may carry specific biases (e.g., greater stability, delayed treatment, or interrupted care outside our region) that could influence subsequent dementia risk.

In addition, diagnoses in our study were solely based on clinical diagnoses according to the DSM-criteria. No structured diagnostic interviews or validated questionnaires were used to diagnose schizophrenia, and MRI’s or structured neuropsychological assessments to diagnose dementia were exceptions. This also means that we cannot exclude the possibility that in some patients the cognitive deficits associated with schizophrenia were severe enough to fulfil DSM-criteria for dementia.

Lastly, correction for some possible confounding factors (e.g., antipsychotic treatment history, comorbidities, drug and alcohol abuse, education level, marital status, and living situations) was not possible because these data were not systematically collected for all patients in the electronic data file.

5. Conclusions

In summary, our study suggests that patients with LOS and VLOS are not at a higher risk of developing dementia compared to patients with EOS. All patients with schizophrenia have a higher risk of developing dementia. These results do not support the hypothesis that LOS and especially VLOS, are precursors of dementia.