Study procedure

Patients and carers participating in the study visited the Dementia Research Group (DRG) for completion of the baseline study assessment- this took approximately four hours and comprised:

1. Explanation of the project and an opportunity for patients and carers to ask any questions they had regarding the study

2. Consent was obtained and the documentation signed 3. Neuropsychiatrie, functional and behavioural assessments 4. Volumetric MRI scan

5. Carers were then interviewed separately to complete measures of carer burden and resource utilisation and to clarify the history.

2.4.1 Patient clinical assessment

All patients and carers were seen at baseline at the DRG. All interviews and assessments were carried out by the author and repeated at follow up after approximately 1 year. If patients had deteriorated and were unable to travel, they were visited at home by the author for reassessment. The domains of cognitive impairment, daily function and behavioural psychopathology were examined using the following structured assessments:

Mini-Mental State Examination (MMSE)

This is a very widely used measure o f cognitive function in dementia. It was not intended to be used as a diagnostic test but to assist in the differentiation of functional and organic cognitive impairment and measure change over time (Folstein e t a I., ^ 975). It is well validated in AD and test- retest reliability is 0.89 (Tombaugh and McIntyre, 1992). The MMSE has not been validated for use in FTD. It does not include tests for frontal lobe dysfunction and in FTD, patients' short -term memory and praxic skills may be intact. Gregory and Hodges (1996b) found in 15 FTD patients that the majority scored above the cut off of 24 required for a diagnosis of dementia. A third obtained a maximum score of 30/30. The MMSE is also dependent on language function and this may give disproportionately low scores in patients with language impairment (Mesulam, 2001).

Clinical Dementia Rating Scale (CDR)

The CDR is a semi-structured interview of patient and carer and provides a more global assessment of function and dementia severity (Hughes et al., 1982). It has become a gold standard in the rating of patients with AD but has not been validated for FTD. It examines six domains of functioning: memory, judgement and problem solving, personal care, orientation, community affairs, home and hobbies. Data can be derived in two ways. A categorical score can be calculated using the algorithm provided this gives five categories ranging from 0- healthy to 3-severe dementia. A continuous variable can be calculated by summing boxes in each area of function and this gives the CDR-sb score (Morris, 1993). The CDR relies on accurate carer report and it is acknowledged that distressed and burdened carers may overestimate the level of functional disability.

MOUSEPAD

The Manchester and Oxford University Scale for the assessment of Psychopathology in Dementia (MOUSEPAD) measures the frequency and severity of psychiatric and behavioural disturbance. It includes 59 items covering psychiatric symptoms of hallucinations, delusions and misidentifications and behavioural changes in eating, walking, sleeping, sexual behaviour, aggression and disinhibition. Test retest reliability is variable (from 0.43-0.93 depending on the symptom groups), inter-rater reliability is 0.56-1.0 and the scale is sensitive to change (Alien et a/., 1996). There are no published studies using the MOUSEPAD in FTD and it was selected because it investigates in detail frontal lobe symptoms such as aggression and d ietary c hanges. 11 has b een validated f or u se i n A D a nd h as b een u sed t o investigate determinants of carer burden (Donaldson eta!., 1998).

Clinical Anxiety Schedule

The Clinical Anxiety Scale (CAS) was developed from the anxiety items of the Hamilton Anxiety Scale. It is completed by the rater whilst examining the patients' current mental state. The scale focuses on symptoms of tension and psychic anxiety thus excluding features that could also be ascribed to depression. There are seven items such as “ability to relax” and “apprehension” which are rated on a five point scale, the maximum score is 28. It has not been validated in studies of dementia

and no previous studies have used it in FTD. It was selected for this project as there are no specific anxiety scales available for use in cognitively impaired patients (Snaith et al., 1982). It is acknowledged that patients with FTD rarely have insight and may not report subjective symptoms of anxiety.

Cornell Scale for Depression In Dementia

This was selected because the MOUSEPAD does not include any depression items. It is based on clinical observation and informant report of symptoms over the previous two weeks (apart from the item on weight loss which is scored for the preceding month) and thus can be used in patients with severe dementia. It comprises of 19 items covering symptoms such as weight loss and suicidal ideation. Each is scored on three defined grades. 0=absent, 1=mild or intermittent, 2=severe. Depression is diagnosed at a cut off score of eight (Alexopoulos et a!., 1988).Items were first discussed with the patients and then separately with the carer. Any discrepancies between patient and carer report were clarified with the carer.

Interview to Determine Deterioration of Daily Function In Dementia (IDDD) This is completed with the caregiver and has 33 items divided into simple non-verbal activities such as dressing and using the toilet and complex activities such as writing or using the telephone. It has the advantage of assessing tasks performed equally by males and females. Previous scales tended to rely more on female dominated household chores. The scale also assesses the initiative of the patient to perform activities, a useful dimension to explore in FTD, as this is more likely to be impaired in frontal lobe problems than the ability to perform the task itself. The minimum score is 0 and maximum score 66 (Teunisse et a/., 1991). As with the CDR, the carers’ report of function on the IDDD may be influenced by the level of burden and stress they are experiencing.

2.4.2 Neuroimaging

Volumetric imaging was performed in the coronal plane using the MRI scanner at the Queen Square Imaging Centre. All scans were assessed qualitatively using a standard DRG rating sheet (Appendix 2) and reported on by a neuroradiologist blind to any clinical details. This enabled the predominant pattern of atrophy to be identified.

Scans were performed at baseline and after approximately one year and then positionally matched (registered). The registration methodology uses the complexity of brain structure for accurate positional matching of Images. This is achieved through iteratively rotating and translating the scans (this compensates for the differences in head position within the scanner at the time of scanning) in order to minimise the variance of the ratio between corresponding points in the image pairs.

An automated subtraction algorithm is then used to measure the differences In total brain volume between two brain scans. Brain volume changes are measured by calculating the integral of the shift in the brain-CSF boundary (the brain-boundary shift integral) taken over the entire brain surface. Whole brain volumes can be registered to within a fraction of a voxel of each other, which allows even small volume changes to be detected by image subtraction. Registration analysis is well suited to longitudinal studies of disease progression in neurodegenerative disease. (Fox et al., 1996a).

Rates of atrophy were calculated from the registered image sets through the measurement of boundary shifts from the difference images and all brain volumes were corrected for total intracranial volume (TIV) to allow for inter-individual variation in head size:

1. The mean TIV was calculated for the whole group at baseline

2. A correction factor (C) was calculated for each individuals' TIV: C= mean croup TIV

Individual TIV

3. % Brain volume loss= brain volume at year 1-brain volume at vear 2 x100 brain volume at year 1

4. % Brain volume loss corrected for TIV (Bjiv) =% brain volume loss x C

5. Rate of volume loss= (Bjiv) x ___________ 365_________ interval between scans in days

2.4.3 Carer assessment

The first part of the interview with the carer completed the assessment of the patient. This included the history and past medical history and completion of carer ratings of the CDR, Cornell, IDDD and quality of life. The second part of the interview focussed on the carer. Five dimensions of care-giver well-being were examined after Colerick and George (1986a).

1. Physical health 2. Mental health 3. Economic status 4. Caregiver burden

2.4.3.1 Physical health

This was assessed by a visual-analogue scale (VAS) of perceived physical well­ being comprising of a 100mm scale which was anchored at 0 (My health has significantly deteriorated as a result of caring for someone with dementia) and 100 (my health and wellbeing has significantly improved as a result of caring for someone with dementia). I he central point at 5 0mm indicates t hat t here was no change. Details of frequency and type of medical consultations and treatments prescribed are documented in the carer assessment part of the Resource Utilisation in Dementia (RUD) instrument.

2.4.3.2 Mental health

This was measured using the 28 item GHQ and the Hospital Anxiety and Depression Scale (HAD). Both scales enable psychiatric “caseness" to be rated. The definition of caseness is that above a particular cut-off score, should the subject be assessed by a psychiatrist they would have a high probability of being diagnosed with a psychiatric disorder.

General Health Questionnaire (GHQ-28)

The GHQ is the most widely used scale for assessing psychiatric caseness. It is self administered and was designed for use in community settings. It is scored in two ways; firstly as a multiple response Likert scale (0-1-2-3). This yields a measure of overall psychological distress giving scores between 0 and 84; and secondly as a bimodal scale (0-0-1-1) which gives a total score of 28. A cut-off score of 4/5 on the bimodal scale was used to derive “psychiatric caseness” for individuals. This gives a sensitivity of 80% and specificity of 88.8%. The 28 -item version used in the study is quick to complete and was developed for population screening (Goldberg and Hillier, 1979). The GHQ has been used previously to examine psychiatric morbidity in carers of patients with AD, PSNP (UttI et al., 1998) and multiple sclerosis (Knight et al., 1997). Studies using the GHQ suggest that GHQ morbidity in carers is closely related to patient psychopathology (Marriott et al., 2000).

Hospital Anxiety and Depression Scale (HAD)

This was developed to screen for clinically significant anxiety and depression in patients attending general medical clinics (Zigmond and Snaith, 1983). Ail items on the scale are concerned with the psychological symptoms of anxiety and depression. It therefore excludes items that may be influenced by physical illness such as sleep disturbance. It is therefore relevant to use it in carers who are known to have increased levels of physical illness or disturbed sleep. The HAD has been used in studies of carers of people with dementia (Gold et al., 1995; Welleford at a/., 1995). Using a Likert scale (3-2-1-0) it gives specific information about the degree of anxiety and depression (maximum score is 21). Using a cut off score off score of 10/11 gives “caseness” for depression and anxiety; the depression scale has a sensitivity of 67% and a specificity of 94% ; the anxiety scale has a sensitivity of 87% and specificity of 76% (Snaith and Zigmond, 1986).

2.4.3.3 Carer burden

Screen for Caregiver Burden (SOB)

This is a 25 item self -rating scale that gives separate scores for objective and subjective burden (Vitaliano eta!., 1991a). Subjective burden (SB) measures overall carer distress whilst objective burden (OB) measures care giver experiences such as problems with feeding or housekeeping which occur independent of their distress. Internal consistency of the scale is 0.85 and it is sensitive to change over time. One potential drawback of the scale is that it has also been found to correlate with the carers’ mood (Burns et ai., 2002) and thus carer depression may confound measurement of burden. The SOB has been used in previous studies of caregiving in AD (Vitaliano et a!., 1993) the frail elderly (Thompson, Jr. et a/., 1993) and multiple sclerosis (Knight eta!., 1997).

Burden 1&2

This is a self- report assessment interview that was developed for a study of gender comparison in caregiving (Pruchno and Resch, 1989). It consists of two scaies:

Burden 1-The carer responds to the question “When caring for another person, some people experience a sense of burden. Overall, how burdened do you feel in caring for the person you are looking after” on a five point scale ranging from “not at all” to “greatly”.

Burden 2- This is a 17 item index of burden with each item describing psychological responses or feelings in the carer that are related to the carer experience. Responses are rated as never, sometimes and often and are scored as 0,1,2, giving a continuous variable of burden.

2.4.3.4 Economic status and health service utilisation

The Resource Utilisation in Dementia Questionnaire (RUD)

This comprehensive tool measures the use of statutory services by the patient. It also allows costing of informal care provided by carers and documents their levels of medical and mental health service utilisation (Wimo et al., 1998). At each assessment this semi-structured interview was used to identify benefits received, and all sources of health and social service support including inpatient admission, home care and nursing services and residential care.

2.4.3.5 Marital relationship

Locke-Waiiace Marital Adjustment Scale

Marital adjustment is defined as the accommodation of a husband and wife to each other at a given time. In the case of married (or equivalent) carers assessment of relationship quality was made using the Locke-Wallace marital adjustment scale (Locke and Williamson, 1958). The responses to the 24 items on the questionnaire are weighted according to the sex of the respondent using an equation derived from factor analysis of the scale. It is short, but reliable and well validated. Test-retest correlations ranged between 0.69-0.78 suggesting it can measure stable and enduring characteristics of the marital relationship (Locke and Williamson, 1958).

The scale is usually completed by both partners but this was not thought to be possible in patients with dementia so only the care giver completed the

questionnaire. There are 23 items in total, twelve items have a multiple choice response, nine items ask the extent of agreement and disagreement on relationship issues and one item describes 22 possible sources of marital conflict (i.e. gambling, infidelity). Dementia will alter the quality of the marital relationship and so care givers were asked to complete the test measuring the relationship at a point shortly before the onset of the patients’ dementia. The onset of FTD can be very insidious and that this may therefore not be a completely reliable measure.

2.4.3.6 Quality of life

The measurement of quality of life in dementia is subjective, complex and methodologically difficult. There may be large discrepancies between the patient and their carer’s perception of the effect of dementia on the patient’s quality of life, this issue is particularly pertinent in FTD where lack of insight is a cardinal feature. Quality of life estimates are correlated with functional abilities and the presence of depression in Alzheimer’s patients (Jonsson at al., 2000) but there is no data on their use in FTD.

Quality of Life in Alzheimer’s Disease: Patient and caregiver report (QOL-AD) This 13 item scale examines separate domains of quality of life ranging from “physical health” to the “living situation” and “ability to do things for fun”. Patients and their carers independently rate these as poor, fair, good or excellent and can point to answers if they are unable to verbalise (Logsdon at al., 1999; Logsdon at al., 2002). A small validation study found it to be acceptable to a group of patients with young onset dementia and their carers (Selai at al., 2001).

2.4.4 Data management and statistical analysis

Each patient was allocated a number and q uestionnaires and rating scales were collated in paper folders only identified by the number to ensure preservation of confidentiality. The data were entered onto an Excel spreadsheet and the STAT Transfer package was used to move the spreadsheet to SPSS version 10.0 (SPSS,

2.4.4.1 Subgroups for analysis

Patients were grouped according to both their clinical presentation and area of atrophy on neuroimaging. It was decided not to use a taxonomicai for a number of reasons. There have been previous studies that have used such an approach (Snowden et a/., 2001a; Bozeat at a/., 2000) and neither of these found clear distinctions between subgroups of FTD patients. This study was planned as naturalistic study of FTD and its course over one year. We felt that patients should be analysed in groups that reflected how they present in normal clinical practice. A taxonomicai approach would also not allow investigation of how the clinical or neuroimaging presentation affected the carer.

Clinical subgroups

Patients were divided into two groups by clinicians (Professor M Rossor and Dr N Fox) who were independent to the study using the recent work group criteria (McKhann at a/., 2001). They examined the (anonymised) patient clinical notes made at the first visit to the cognitive disorders clinic, before neuroimaging had been performed.

• Behavioural (denoted by “B” in the results section) • Language (denoted by “L” in the results section)

Neuroimaging subgroups

The baseline MRI scans were rated by a consultant neuroradiologist who was blinded to the clinical details and the cohort divided according to the predominant areas and extent of atrophy:

Normal scan on visual inspection (denoted by “N” in the results section) Predominant frontal lobe atrophy (denoted by “F” in the results section) Predominant temporal lobe atrophy (denoted by “T” in the results section) Fronto-temporai atrophy (denoted by “FT” in the results section)

2.4.4.2 Power calculation

This study has a range of different outcome measures. Exploratory power calculations were performed using a number of outcome measures. For example:

• From clinical experience we would expect 25% of patients with a language presentation to develop abnormal eating behaviours compared to 70% of those with a language presentation.

Analyses assumed Type 1 error at 1% and Type II error at 10%. Power was set at 90% and significance at 5%. The results suggested a sample size of 35-56 would be required.

2.4.4.3 Statistical analysis

Simple descriptive statistics were used in the first instance to describe the cohort as a whole. The size of the data set indicated that non-parametric methods were the most appropriate. Differences between means of unpaired data (the clinical behavioural and language sub groups) were explored using the Mann-Whitney U test. Where comparison of more than two groups was required (i.e. the neuroimaging subgroups), the Kruskall -Wallace test was used. Categorical data were analysed using Fischer’s exact test.

Multivariate analysis techniques were not used as this was primarily an exploratory and descriptive study, the sample size was relatively small and some of the variables were not normally distributed. Data for the MOUSEPAD were particularly skewed and this was not remediable on logarithmic conversion, precluding the use of parametric and multivariate techniques.

The neuroimaging also provided continuous data on within subject changes in volume of brain tissue. Differences between paired data (baseline and year 1 scores) were analysed using the Wilcoxon signed rank test. Carer Assessments gave both categorical and quantitative data on indexes of burden and support. This data was analysed according to carer characteristics such as gender or work status but also according to the patient subgroups as described above.

2.4.4.4 Predictors of outcome in patient deciine and caregiver burden

The analysis was primarily exploratory, using a range of different models. We