In this study, informants reported neuropsychiatric symptoms in about one third to one half of patients with sCAA and D-CAA. Apathy and irritability were present in one third of CAA-patients. Apathy was associated with sCAA and D-CAA, but results for irritability were inconclusive. Our data suggest that NPS might be an early clinical sign of D-CAA. Having more NPS was associated with decreased processing speed and executive functioning. We found no association between NPS-count and overall or frontal CAA disease burden on MRI. Men had NPS more often than women, but there were no clear sex-differences for apathy or irritability.

Patients with symptomatic D-CAA and sCAA had a highly similar neuropsychiatric profile, with increased frequency and severity of NPS in patients with a history of ICH or cognitive decline. The four most frequently reported NPS in sCAA (irritability, apathy, agitation and depression) in our study resembled a previous study, albeit with lower prevalence estimates of agitation and depression [7]. The near absence of hallucinations, delusions and motor disturbances also resonates with the literature.

The apathy frequency of 30–37%, measured by the SAS, resembled previous findings in CAA-literature, where 35% of 43 patients with sCAA had NPI-Q based apathy [7]. Our estimate was also in line with the results of another study (n = 37, 43% apathetic according to the SAS) in patients with sCAA without previous ICH [8]. We note that informants might confuse apathy with depression. However, while this was not assessed in regression analysis for statistical efficiency, the overlap of apathy and depression on the NPI-Q, or on the CESD and SAS was limited. Considering the similarities between our findings in sCAA and the previous literature, we expect that our informant-based findings are robust across all patient groups.

Approximately one third of patients had high scores for irritability on the Irritability Scale and NPI-Q. In sCAA, this is slightly less than reported in previous literature (37%, NPI-Q based) [7]. In controls, the prevalence of irritability was similar to CAA on the Irritability Scale, but lower on the NPI-Q. As the Irritability Scale more extensively measures the underlying construct than the NPI-Q, we expect the true proportion of control participants with irritability to be approximately one third, suggesting that there was no clear difference between patients and controls. We note that this might be due to measurement properties of the irritability scale, or because relatively fewer irritability scales than NPI-Qs were analyzed. Thus our results for irritability remain inconclusive.

We observed higher odds of having NPS, but not specifically of apathy or irritability, in presymptomatic D-CAA, compared with controls. We note that patients with presymptomatic D-CAA were younger than controls, but also that their odds were similar to the other CAA-groups. No previous literature for D-CAA was available for comparison. However, previous studies have identified WMH to be correlated with development of NPS in CSVD patients, possibly as a consequence of damage to reward systems and structures related to decision making [39,40,41]. As non-hemorrhagic tissue injury is thought to be one of the early MRI markers in D-CAA to become abnormal, this might explain the onset of NPS in the presymptomatic phase of D-CAA [42]. Genetic disease-related psychological factors might also explain part of the NPS in (pre)symptomatic D-CAA.

We found that total NPS-count was associated with decreased processing speed and executive function. Interpreting these results is challenging. While in AD both domains were previously associated with total NPS-count, a previous CAA-study did not observe this association [7, 39, 43]. Our discrepancy with the CAA-literature might be due the use of different neuropsychological instruments. However, our sensitivity analysis yielded comparable results, despite adding another instrument and restricting sample size. Another explanation might be that answering strategies, such as speed-accuracy trade-off for errorless performance, are not captured in quantitative analysis of neuropsychological tests [44, 45]. This increases the influence of sampling error and random variation when comparing studies. Information to assess this trade-off was unavailable in our data. Further, differences in executive function between in- and excluded participants might have caused effect-underestimation, but might also be attributed to random variation. Overall, our findings suggest that NPS may be associated with processing speed and executive function, but our results should be interpreted with caution.

In our study, NPS and total (or frontal) CAA-burden score on MRI were not associated. This was unexpected, as we hypothesized that accumulation of vascular damage amounts to disruption of regulating circuits (such as but not limited to the limbic system, the reward systems, or those related to the orbitofrontal cortex) and cognitive decline, thereby causing NPS. Moreover, in patients with ICH, apathy and hyperactivity were previously associated with a CAA MRI-profile [9]. Our findings align with one previous sCAA study, but contrasts with another, in which high SAS-score correlated with increased CAA-burden and with disruption of white matter tracts on diffusion tensor imaging [7, 8]. Unfortunately, we were not able to assess white matter volume or microstructural integrity. However, our result might be a consequence of using the CAA-burden score, a measure that gives less weight to ischemic burden, and where different patterns of cSS, CMB, WMH and CSO-EPVS can accumulate to the same score. The same is true for the frontal burden, which only comprised hemorrhagic lesions (presence/absence of cSS and number of CMBs). This might have been an insufficiently sensitive measure for quantifying the frontal burden, which might explain the lack of an association with NPS. Future research might best focus on incorporating non-hemorrhagic burden with quantitative measures, and microstructural integrity to study this further.

In our study, men seemed to be more prone to have NPS than women, despite similar age profiles. This contrasts a recent meta-analysis in AD-patients (50% NPI-Q based) that did not find sex-differences for NPS, apathy or irritability, although men exhibited more severe apathy [5]. In a previous CAA-study, the NPS-incidence was similar between sexes [7]. Thus it remains unclear if our finding is generalizable. Nevertheless, clinicians should be aware that sex-differences in NPS may exist, as these symptoms can sometimes be treatable and have implications for patient and caregiver education.

A strength of our study is that we were able to study symptomatology, cognition and sex-differences in a patient sample that equals the combined previous literature in size. We confirmed findings for patients with sCAA without a history of ICH and added new knowledge about those with previous ICH. Also, we were able to investigate NPS in a unique hereditary cohort of patients with D-CAA, a relatively pure form of CAA, in both the presymptomatic and symptomatic phases. Further, by using the same instruments as in the previous literature, but also using more extensive validated questionnaires, we were able to study apathy and irritability in more depth, while retaining comparability and reliability despite using data from different source cohorts.

Our study has other limitations. First, we measured NPS rather than clinical (neuro)psychiatric diagnoses: the questionnaires do not differentiate between clinical and non-clinical complaints [3]. Additionally, NPS were only evaluated by informants, which might differ from clinical ratings. Therefore, our results should be interpreted as informant-reported symptomatology rather than as clinical diagnoses. Second, our questionnaire study is inherently susceptible to motived-responder bias. Specifically, informant responses might be biased in those with higher caregiver burden. Unfortunately, no reference group with patients with non-neurologic disease was available for comparison. Replication could assess this. Third, the controls are a selected sample. However, we believe that this group provides a useful reference for the NPS-profile in absence of CAA, or (subjective) cognitive decline. Fourth, data was collected from separate source cohorts and thus is susceptible for systematic differences. However, the neuropsychological testing protocols and study assessments of the source cohorts were designed to be comparable, thereby minimizing impact on the generalizability of our findings. Fifth, the CAA-burden score is a simplified compound score that quantifies the CAA-lesion load by combining multiple rating scales. Due to its ceiling effects (i.e., for cSS and WMH quantity, or number of CMBs), the loss of information by categorization, and its inherent omission of other (micro)structural lesions, the clinical applicability of our burden-related findings was restricted. In extension, unfortunately data on cortical thinning of the different lobes was not available. As cortical thinning might be associated with apathy, future research might investigate this further for patients with CAA [46]. Finally, because we considered depression and anxiety best measured by patient-reporting and no previous NPS-related CAA-literature existed at study conception, we did not collect data on these symptoms or assess them in-depth in our informant-based study.

In conclusion, neuropsychiatric symptoms, specifically apathy, are common in patients with sCAA and D-CAA. These symptoms are already present before the occurrence of ICH and men might be more susceptible to developing them. Worse cognition, but not evidently CAA MRI-burden score, may be associated with having more NPS. Neurologists should inform patients and caregivers of these disease consequences and treat or refer patients who are burdened by these symptoms appropriately.