Article Text

Download PDFPDF

Research paper
Hippocampal subfield atrophy of CA1 and subicular structures predict progression to dementia in idiopathic Parkinson’s disease
  1. Audrey Low1,
  2. Heidi Foo1,
  3. Ting Ting Yong1,
  4. Louis C S Tan1,2,
  5. Nagaendran Kandiah1,2
  1. 1 Department of Neurology, National Neuroscience Institute, Singapore, Singapore
  2. 2 Duke-NUS Medical School, Singapore, Singapore
  1. Correspondence to Dr Nagaendran Kandiah, Department of Neurology, National Neuroscience Institute, Singapore 308433, Singapore; nagaendran_kandiah{at}


Background Global hippocampal atrophy is a hallmark of Alzheimer’s dementia and has been similarly reported in Parkinson’s disease dementia (PDD). However, there is limited literature on the differential involvement of hippocampal subfields in predicting conversion to PDD. This study is an extension of previous findings on progression to mild cognitive impairment in Parkinson’s disease (PD).

Methods This cohort study recruited 73 non-demented participants with idiopathic PD (age 65.80±8.17, 75.3% male) from an outpatient neurology clinic. All participants underwent clinical assessment, neuropsychological testing and 3T MRI scans at baseline and 18 months while on prescribed dopaminergic medication. Hippocampal subfield volumes were obtained using automatic segmentation in FreeSurfer V.6.0. Participants who progressed to PDD and those who did not were compared on hippocampal subfield atrophy and cognitive change (episodic memory, attention, executive functions, language, visuospatial abilities). Subfields were further examined for their abilities to predict PDD conversion and distinguish PDD from non-demented PD using receiver operating characteristic analysis.

Results Smaller baseline global hippocampal volume, cornu ammonis (CA) subfield CA1, subiculum and presubiculum volumes were observed in participants who went on to develop dementia, and predicted PDD conversion. Those who progressed to PDD saw greater decline in global hippocampal volume, granule cell layer of the dentate gyrus, presubiculum, parasubiculum and fimbria. Decline in subiculum and fimbria volume corresponded to cognitive decline in attention and executive functions, respectively.

Conclusions Early atrophy of CA1, subiculum and presubiculum preceded, and predicted, PDD conversion. Differential patterns of subfield atrophy were also observed among those who progressed to PDD and were associated with impaired executive functions.

  • parkinson’s disease dementia
  • cognitive impairment
  • neuroimaging
  • hippocampus

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Parkinson’s disease (PD) is a debilitating neurodegenerative disorder affecting one in every hundred individuals above the age of 60 globally, making it the second most common neurodegenerative disorder after Alzheimer’s disease (AD).1 While PD is traditionally characterised by motor impairments, its definition has expanded to incorporate non-motor characteristics in the recent years, such as cognitive impairment. Cognitive impairment in PD has been found to be associated with poorer clinical outcomes, such as lower quality of life,2 higher rates of institutionalisation,3 higher rates of depression and early mortality,3 4 as compared with patients with PD without cognitive impairment. By identifying early and accurate biomarkers of presymptomatic Parkinson’s disease dementia (PDD), clinical care planning and disease management can be initiated earlier, hence alleviating the burden on patients, caregivers and public health systems.

Recent advancements in the use of MRI have identified the hippocampus as an important predictor of cognitive impairment in PD, perhaps unsurprising given its vast connections with other important cerebral structures5 and its involvement in a multitude of cognitive processes.6 However, although majority of these studies regard the hippocampus as a homogeneous structure, it is increasingly being recognised that the hippocampus in fact comprises several subfields with distinct morphologies.7 Importantly, hippocampal subfield volumes have been proposed to be better biomarkers for early detection of dementia, compared with global hippocampal volumes.8 Indeed, earlier work by Foo et al 9 found that baseline hippocampal subfield volumes, namely the granule cell layer of the dentate gyrus (GC-DG), CA4, parasubiculum and hippocampus–amygdala transition area (HATA), were predictive of conversion from PD without cognitive impairment to PD with mild cognitive impairment (PD-MCI). However, it remains unclear as to whether these early hippocampal changes are further implicated in the development of dementia in PD. Despite evidence of the differential vulnerability of hippocampal subfields to neuropathological changes, including atrophy patterns corresponding with the Braak staging of AD,7 10 and similar findings in other dementias such as semantic and frontotemporal dementia,8 11 no study has investigated the role of subfield-level hippocampal atrophy in PDD progression.

As cognitive deficits observed in PD are predominantly frontal-based, that is, attention, working memory and executive functions,12 we hypothesised that participants with PD who progressed to dementia would display greater loss of volume in the subfields involved in working memory and executive functions: the subiculum (including presubiculum and parasubiculum), the cornu ammonis (CA) subfield CA1, the GC-DG and the fimbria. We further hypothesised that subfield analysis would be a better predictor of PDD progression compared with global hippocampal volume, and that the loss of volume in these subfields would be associated with greater risk of progression to PDD, and greater decline in the domains of attention and executive functions.



This cohort study recruited 97 non-demented participants with mild idiopathic PD between August 2011 and September 2013 from a specialist outpatient neurology clinic (National Neuroscience Institute, Singapore). Five participants did not have complete data at baseline, and 19 withdrew from the study before the follow-up visit. Only participants with complete baseline and follow-up data on cognition and neuroimaging were included in this study (n=73).

All participants underwent clinical and neuropsychological assessments, APOE genotyping, as well as MRI imaging at baseline and at follow-up. Clinical evaluations included medical history, and an assessment of motor and functional abilities (Unified PD Rating Scale) and stage of disease (Hoehn and Yahr; H&Y). All participants were on dopaminergic medication, and all motor and neuropsychological assessments were conducted in an ‘on’ state. Levodopa equivalency daily dosage (LEDD) was calculated based on Tomlinson and colleagues’ formula.13 Diagnosis of PD was made by movement disorder specialists according to the National Institute of Neurological Disorders and Stroke criteria.14 Only participants with mild PD, as defined by H&Y stages between 1 and 2.5, and participants with idiopathic PD were recruited, hence excluding patients with dementia with Lewy bodies and other PD-plus syndromes. Participants with serious medical and/or psychiatric comorbidities were excluded from the study. Diagnosis of PDD was based on the Movement Disorder Society (MDS) criteria for PDD.15 The Strengthening the Reporting of Observational Studies in Epidemiology guidelines were used to ensure the reporting of this study.

MRI acquisition and hippocampal subfield volumetry

All participants underwent both baseline and repeat MRI after 18 months. Both MRI acquisitions were performed on a 3 Tesla GE scanner system. Structural scans using high-resolution T1-weighted Magnetization Prepared Rapid Acquisition Gradient Echo (MPRAGE) (axial acquisition, 176 slices, matrix size=256×256, voxel size=1.0×1.0×1.0 mm3, echo time=3.2 ms, repetition time=7 ms, inversion time=850 ms, flip angle=8°, field of view=256×256 mm2) was acquired for all subjects.

Briefly, this study used FreeSurfer V.6.0 ( to obtain baseline and follow-up hippocampal subfield volumes. This automated segmentation tool is based on a probabilistic statistical atlas built on ultra-high-resolution ex vivo MRI data from autopsy brains—technical details have been previously described.16 Thirteen subfield volumes for each hemisphere were obtained, namely the alveus, parasubiculum, presubiculum, subiculum, CA1, CA2–3, CA4, GC-DG, HATA, fimbria, molecular layer, hippocampal fissure and hippocampal tail. Total intracranial volume (ICV) and cerebral grey and white matter volumes were also obtained using FreeSurfer V.6.0. Details of the segmentation and analysis procedures have been previously described.9

Neuropsychological assessments

A standardised neuropsychological battery was administered at baseline and follow-up by trained assessors. Global cognitive function was evaluated using the Mini Mental State Examination and a local adaptation of the Montreal Cognitive Assessment. As recommended by the MDS Task Force,17 participants were also assessed on five specific cognitive domains. As per the methods by Foo and colleagues,9 the tests for each domain included the following measures: Episodic memory was measured using the delayed word recall and word recognition tasks from the Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog). Attention was assessed using Color Trails Test 1 and Digit Span Forward. Executive functioning was measured using Color Trails Test 2, Digit Span Backward and Sunderland’s Clock. Visuospatial function was assessed using the figure copy task and the maze tasks of the ADAS-Cog. Language was assessed using the naming subtest and comprehension of test instructions from the ADAS-Cog. To compute cognitive domain scores, raw scores of each neuropsychological test were converted to z-scores and averaged to yield composite scores. Changes in cognitive functioning in each domain were computed using the following formula: (zDomainfollow-up – zDomainbaseline).

APOE genotyping

Blood samples were collected from participants and DNA was extracted. APOE genotyping was performed using the TaqMan SNP genotyping assay and ABI 7900HT PCR system (Applied Biosystems).

Statistical analysis

Statistical analysis was performed using SPSS V.20.0. χ2 tests of independence were used for group comparisons of categorical variables such as gender, ethnicity and vascular risk factors, while Mann-Whitney U tests were used to compare PD-progressors and PD-stable on motor scores and LEDD. Hippocampal subfield volumes and cognition were compared at baseline and percentage change at follow-up using analysis of covariance (ANCOVA), controlling for gender, age, education, vascular risk factors and duration between visits, and adjusted for multiple comparisons using Bonferroni correction. Additionally, ICV was controlled for in volumetric analyses to account for interindividual differences in head sizes.

Separate multiple linear regression models were conducted to investigate whether hippocampal subfield volumes were associated with cognitive scores at baseline and whether subfield volume reduction corresponded with a worsening of cognitive function. To determine whether baseline hippocampal atrophy predicted conversion to PDD, logistic regression analysis was performed with the inclusions of gender, age, education and ICV as covariates. Receiver operating characteristic (ROC) analysis was used to assess the diagnostic performance of hippocampal subfield atrophy in differentiating between PD-stable and PD-progressors. The area under the curve (AUC) of global hippocampal atrophy was compared against the subfields to test the hypothesis that subfield atrophy is more accurate than global hippocampal volume in differentiating between PDD and non-demented PD. Significance was set at a two-tailed probability value of 0.05.


Characteristics of study population

A total of 73 participants were included in the analyses. Demographic and clinical characteristics are summarised in table 1. The mean age of the cohort was 65.8 years (SD 8.17), while the mean years of education was 10.9 years (SD 3.36). Majority of participants were Chinese (90.4%) and male (75.3%). At 18 months of follow-up, 11 (15.1%) had progressed to PDD.

Table 1

Patient characteristics

PD-progressors and PD-stable did not differ in terms of gender, ethnicity or education (p>0.05). However, participants who progressed to PDD were older (mean=67.62 years) compared with those who remained stable (mean=65.1 years). PD-progressors and PD-stable were similar in terms of motor severity and use of dopaminergic medication both at baseline and follow-up, with the exception of baseline LEDD (p=0.048). However, baseline LEDD differences became non-significant after controlling for baseline motor score (p=0.387) using ANCOVA. In terms of cardiovascular risk factors, the two groups did not differ on prevalence of diabetes or history of smoking. However, those who progressed to PDD were more likely to have a history of hypertension and hyperlipidaemia.

Hippocampal subfield volumetry

At baseline, controlling for gender, age, education, vascular factors and ICV, and corrected for multiple comparisons using Bonferroni correction, smaller global hippocampal volume (p=0.040) and subfield volumes of the CA1 (p=0.049), subiculum (p=0.037) and presubiculum (p=0.038) volumes were observed in participants who went on to develop dementia. Although significant group difference was observed in the subiculum and presubiculum, there were no significant differences in the parasubiculum (p>0.05) at baseline (table 2).

Table 2

Hippocampal subfield volume differences at baseline and changes after 18 months

The percentage change in hippocampal subfields were compared at follow-up using ANCOVA, controlling for gender, age, education, vascular factors, ICV and interval between scans, and adjusted for multiple comparisons using Bonferroni correction. Those who progressed to PDD saw greater atrophy in global hippocampal volume (p=0.006), presubiculum (p=0.027), parasubiculum (p=0.008), GC-DG (p=0.040) and fimbria (p=0.019) (figure 1).

Figure 1

Atrophy of hippocampal subfields in a study participant who converted from non-demented idiopathic Parkinson’s disease to PDD after 18 months. Participants who converted to PDD saw greater volume loss in the parasubiculum, presubiculum, GC-DG and fimbria. CA, cornu ammonis; GC-DG, granule cell layer of the dentate gyrus; HATA, hippocampus–amygdala transition area; PDD, Parkinson’s disease dementia.

Further analysis was conducted to evaluate the diagnostic utility of the subfields, in terms of differentiating between PDD and non-demented PD at 18 months using ROC analysis. All subfields attained an AUC significantly above 0.5, except the parasubiculum. Global hippocampal volume (AUC=73.8% [58.6%, 88.9%]) was more predictive than any other individual subfield (AUC=55.0%–72.0%), with the exception of the presubiculum (AUC=74.6% [58.1%, 91.2%]). However, on combining the volumes of the GC-DG, presubiculum, parasubiculum and fimbria, this composite volume measure was superior to global hippocampal volume at distinguishing non-demented from demented patients with PD (AUC=76.2% [62.1%, 90.4%]).

To examine if baseline hippocampal subfield volumes predicted conversion to PDD, logistic regression analysis was conducted, controlling for gender, age, education and ICV. Smaller baseline volumes of the whole hippocampus (p=0.023), CA1 (p=0.021), subiculum (p=0.023) and presubiculum (p=0.039) increased the likelihood of conversion to PDD. The parasubiculum, GC-DG and fimbria did not significantly predict conversion to PDD.

Further ROC analysis was conducted to evaluate the diagnostic accuracy of longitudinal hippocampal subfield volume loss in differentiating between PD-stable and PD-progressors. The AUC was significantly higher than 0.5 for the global hippocampal volume (AUC=76.8% [63.5%, 90.2%]), presubiculum (AUC=74.2% [59.2%, 89.2%]), parasubiculum (AUC=71.0% [55.3%, 86.7%]), GC-DG (AUC=71.6% [55.8%, 87.3%]) and fimbria (AUC=74.6% [61.1%, 88.1%]). None of the subfields had an AUC higher than global hippocampal volume loss.

Neuropsychological assessment and associations with hippocampal subfield volumetry

At baseline, PD-progressors and PD-stable performed similarly on global cognition, episodic memory, attention, executive function and language (p>0.05). However, PD-progressors performed worse at baseline visuospatial function (p=0.015). At the follow-up, PD-progressors saw a greater decline in episodic memory, compared with those who did not develop dementia. Notably, changes in global cognition, attention, executive function, visuospatial ability and language were not significantly different between the two groups (table 3).

Table 3

Baseline cognition and change scores

Multiple regression analysis controlling for gender, age, education and ICV revealed that, at baseline, total hippocampal volume was not associated with poorer cognition. However, significant associations were observed within specific subfields. Smaller CA1 volume was associated with poorer language (p=0.028), while smaller presubiculum (p=0.028) and parasubiculum volume (p=0.037) were associated with poorer attention. The main subiculum volume itself showed no significant associations with cognition.

The associations between decline in hippocampal subfield volume and worsening cognition were also investigated, controlling for gender, age, education and ICV. Decline in subiculum volume was associated with worsening attention (p=0.042), and reduced fimbria volume was related to greater decline in executive function (p=0.028).


This study demonstrated that smaller volumes in the CA1, subiculum and presubiculum were evident at baseline among patients with PD who went on to develop dementia—these subfields were also able to predict conversion to PDD 18 months in advance. Those who progressed to PDD experienced greater volume loss in the presubiculum, parasubiculum, GC-DG and fimbria, compared with those who remained cognitively stable. At follow-up, although individual subfields were less effective than global hippocampal volume at distinguishing between PDD and non-demented patients with PD, combining the presubiculum, parasubiculum, GC-DG and fimbria formed a composite measure that was superior to global hippocampal volume in detecting dementia in PD. This study also found that greater volume loss in the subiculum and fimbria was related to a corresponding worsening of attention and executive functions, respectively.

To our knowledge, this is the first study to assess changes in hippocampal subfield volume in PDD conversion, expanding on previous work by Foo and colleagues9 on cognitively healthy PD conversion to PD-MCI. Using a state-of-the-art hippocampal segmentation approach, greater loss of volume was observed in the presubiculum, parasubiculum, GC-DG and fimbria of participants who converted to dementia, compared with those who did not. Anatomically, these structures occupy the border of the hippocampal fissure and are situated at the utmost medial portions of the hippocampus, indicating a medial-to-lateral pattern of hippocampal degeneration among patients with PD who converted to dementia.

The fimbria is a major input–output pathway of the hippocampus and has been found to be involved in visuospatial functions and working memory,18 19 both of which are known to be dominant deficits in PDD. In this study, atrophy of the fimbria indeed corresponded to a worsening of executive function, suggesting that the executive dysfunctions often observed in PDD could be a result of fimbria atrophy.

Additionally, this study also found greater pathological alterations in the presubiculum and parasubiculum among those who converted to PDD. The two structures play vital roles in cognitive processing and visual spatial functioning. Due to the strong inputs received from the anterior thalamic nuclei, and its substantial intrahippocampal connections, the parasubiculum is considered an important pathway through which early processing of incoming information occurs.20 As atrophy of the parasubiculum disrupts these pathways, this might account for the cognitive decline leading to PDD conversion.

Although the GC-DG is one of the few sites in the brain where new neurons continue to be generated throughout adulthood,21–23 neither the dentate gyrus nor neurogenesis has received much attention with regard to cognitive impairment in PD. This present study found that patients who converted to PDD experienced a greater decline in GC-DG volume, compared with counterparts who did not. Earlier postmortem evidence has demonstrated reduced neurogenesis in the GC-DG of patients with PD24—this may be explained by the well-documented deficits in the dopaminergic system observed in PD, which plays an important role in neurogenesis.25 The importance of adult-born GC-DG neurons should also be emphasised—compared with mature neurons, new neurons generated by the GC-DG in adulthood (1) are preferentially recruited into the dentate gyrus circuits compared with older neurons,26 (2) have lower thresholds for plasticity27–29 and (3) make unique contributions to memory processing in the dentate gyrus.21 Taken together, this study suggests the potential of therapies aimed at enhancing neurogenesis of newborn neurons via the GC-DG, as a source of endogenous repair. For instance, lifestyle factors such as exercise and blood sugar control have been found to affect the cytoarchitecture of the dentate gyrus, neurogenesis, and in turn cognitive functions.30 31 Other therapies aimed at improving cognition via neurogenesis include deep brain stimulation,32 33 which has been found to improve learning and memory in mouse models—similar human clinical trials are also currently under way.34

To test the hypothesis that subfield analysis can be more accurate than global hippocampal volume at detecting dementia in PD, ROC analysis was conducted. However, contrary to studies done in AD populations,7 8 global hippocampal volume was better than all individual subfields at distinguishing between non-demented patients with PD and patients with PDD, with the exception of the presubiculum. As this study observed that the presubiculum, parasubiculum, GC-DG and fimbria saw greater volume loss in PD-progressors, we combined these regions and found that this composite measure was superior to global hippocampal volume in detecting PDD.

This study also demonstrated the predictive value of baseline hippocampal subfield volumes in predicting conversion to PDD. Apart from global hippocampal volume, the CA1, subiculum and presubiculum were able to identify individuals at greater risk of converting to PDD. This was in contrast to the findings of Foo et al,9 who found the GC-DG, CA4, parasubiculum and HATA to be predictive of MCI in patients with PD.9 Instead, our findings were more closely aligned with findings on AD conversion, which have demonstrated the CA1 and subicular structures to be predictive of conversion to AD,35–38 suggesting an underlying AD pathology in patients with PD who go on to develop more advanced cognitive impairment—future research incorporating amyloid imaging should be done to elucidate these mechanisms.

In terms of cognition, this study found that baseline CA1 volume was positively associated with language, while the presubiculum and parasubiculum were significantly related to executive function. Although there were no significant associations at baseline, a decline in subiculum volume was accompanied by a corresponding dip in attention. These findings are largely in line with previous research which has demonstrated that the CA1 and subiculum play vital roles in attentional processes—CA1 as a novelty detector or attentional gate, and the subiculum for its involvement in working memory.9 39 Given the central role of attention in a broad spectrum of cognitive tasks, baseline atrophy of the CA1 and subicular structures and the corresponding worsening of attention could explain the subsequent overall cognitive decline and conversion to PDD.

The strengths of the present study include the use of a longitudinal design, which contributes significantly to the existing literature comprising largely of cross-sectional studies. Another advantage of our study is the use of a full neuropsychological battery covering the cognitive domains recommended by the MDS Task Force, that is attention, executive function, language, memory and visuospatial function.17 However, findings could be strengthened by following a larger sample size of non-demented patients with PD over a longer duration of time, especially given the small number of participants who converted to PDD within the follow-up period. Furthermore, as participants were not followed up with further neuropsychological testing, we are not able to ascertain the stability of the PDD diagnosis. This study chose to employ automated segmentation of the hippocampal subfields, as opposed to manual delineation, in order for the findings to be transferable to the clinical context for prognostic and therapeutic purposes. The use of automatic hippocampal segmentation may be considered a limitation, due to concerns about the accuracy of mapping subfield boundaries. However, as the technique employed in this study was based on ultra-high-resolution ex vivo MRI training data,16 accuracy of subfield delineation is far more reliable compared with past methods, producing comparable findings with manual delineation. In addition, automatic procedures have the advantage of being bias-free, less labour-intensive and more easily reproducible.

This present study demonstrates the clinical relevance of studying the patterns of selective hippocampal subfield atrophy in creating a deeper understanding of the pathophysiology of cognitive deficits and dementia in PD. A direct clinical implication is the identification of subfield atrophy patterns which are predictive of conversion to PDD, thus enabling clinicians to identify patients at greater risk of conversion. Furthermore, findings suggest that therapeutic interventions aimed at neurogenesis may be a promising avenue to slow or prevent further cognitive deterioration in PD, which in turn has implications on research in the field of disease-modifying therapeutics for PD.


To date, few studies have investigated the patterns of hippocampal atrophy at the subfield level in PDD. This present study found that conversion from non-demented PD to PDD was accompanied by greater atrophy of the fimbria, presubiculum, parasubiculum and GC-DG. Findings of the predictive value of the CA1, subiculum and presubiculum in predicting conversion to PDD were consistent with previously reported findings on conversion to AD, suggesting a possible underlying AD pathology in cases of PDD. Therapies targeting neurogenesis in the GC-DG could be an effective route to preserving cognitive functions in patients with PD.



  • Contributors AL contributed to study design, statistical analysis, interpretation of data, and drafting and revision of the manuscript for intellectual content. HF contributed to study design, analysis, and drafting, revision and final approval of the manuscript. TTY contributed to data collection, statistical analysis and revision of the manuscript for intellectual content. LCST contributed to interpretation of data, and revision and final approval of the manuscript. NK contributed to study design, interpretation of data, and revision and final approval of the manuscript.

  • Funding The research was supported by the National Neuroscience Institute (Singapore) in the design and conduct of the study. This study was funded by the National Medical Research Council (NMRC/CSA/0063/2014).

  • Competing interests AL, HF and TTY report no disclosures. LCST has received research support from Singapore Millennium Foundation and funding for conference travel from GlaxoSmithKline. NK has received CME sponsorship from Lundbeck, Novartis, Pfizer and Eisai, and research funding from the SingHealth Foundation and the National Medical Research Council of Singapore.

  • Patient consent for publication Not required.

  • Ethics approval This study was approved by the SingHealth Institutional Ethics Review Board, and informed consent was obtained from all participants prior to data collection.

  • Provenance and peer review Not commissioned; externally peer reviewed.