Abstract
Diagnosis of atypical/unclear dementia is often difficult and this delays treatment initiation. Several authors have shown that beyond standard dementia workup, 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) reduces the number of unclear diagnoses, leads to earlier treatment, and has a beneficial impact on families. However, it is not uncommon that the FDG-PET findings are equivocal in this setting. For those cases, a repeat FDG-PET may clarify the diagnosis and prevent treatment delay. We retrospectively assessed the clinical impact of a repeat FDG-PET in 59 patients with atypical/unclear dementia syndromes and inconclusive initial FDG-PET. Changes in primary diagnosis, diagnostic confidence, and management following the second FDG-PET were examined. Conducting a second FDG-PET reduced the number of unclear diagnoses from 80% to 34% , led to diagnostic change in 24% of cases, and treatment modification in 22% of patients. Overall, the clinical impact was higher when initial diagnostic confidence was low and the second FDG-PET repeated ≥12 months after the first one. In tertiary care memory clinic settings, when diagnostic incertitude persists despite extensive evaluation and an equivocal FDG-PET, repeating the FDG-PET 12 months later can greatly clarify the diagnosis and improve management.
INTRODUCTION
Dementia diagnosis can be challenging, particularly in the early stages of the disease, in younger patients, in atypical/unclear presentations, or in patients with comorbid neuropsychiatric symptomatology [1–3]. Delay in treatment due to diagnostic incertitude, particularly frequent in atypical/unclear dementias, has important clinical and psychological consequences for patients and their families [4, 5].
In tertiary care memory clinics, where the most complex patients are seen, a significant proportion of cases remain unclear despite a comprehensive clinical evaluation. For such atypical/unclear dementias, further investigation is often undertaken to obtain a clear diagnosis. These complementary evaluations include a detailed neuropsychological evaluation, blood tests, cerebrospinal fluid analysis, magnetic resonance imaging (MRI) as well as molecular imaging with 18F-fluorodeoxyglucose positron emission tomography (FDG-PET).
A growing body of evidence indeed supports the value of FDG-PET in the diagnosis of patients with atypical/unclear dementias [6–14]. Importantly, FDG-PET can improve diagnostic accuracy, lead to earlier treatment, better planning for future care, and less suffering for patients and their families. In 2010, a retrospective memory clinic study evaluating the value of FDG-PET in mild cognitive impairment (MCI), typical and atypical/unclear dementias [9] showed that the addition of FDG-PET to the routine memory clinic workup significantly lowered the number of unclear diagnoses from 39% to 16% . FDG-PET was also associated with a change in diagnosis in 29% of patients and a 64% increase in the use of cholinesterase inhibitors (ChEIs), the current drug approved for the treatment of Alzheimer’s disease (AD). These results were recently corroborated by a prospective study on the clinical utility of FDG-PET in 194 patients, which showed remarkably concordant results [6]. In this study, FDG-PET changed the clinical diagnosis in 35% , altered the use of ChEI medication in 17% , and reduced uncertain diagnoses from 30% to 18% . Altogether, recent EFNS guidelines [15] recommend the use of FDG-PET for dementia cases showing atypical features and diagnostic uncertainty.
However, in a number of cases, equivocal or incongruent FDG-PET results may leave clinicians in an even greater dilemma. Indeed, despite the diagnostic clarification FDG-PET allows in most cases, 18% of patients in Elias et al., and 16% in Laforce et al. respectively remained with an uncertain diagnosis after a first FDG-PET and an extensive diagnostic work-up [6, 9]. In these highly difficult cases, very few studies provide guidance about the next steps to take in the investigative process. Given the availability of FDG-PET technology in many centers, some clinicians order a second FDG-PET to look for a progression or clarification of metabolic findings. Whether this practice is useful and associated with improved outcomes is unclear and, to our knowledge, has never been clearly addressed. Furthermore, the most appropriate repeat delay has never been studied. The literature reports one conference abstract [16] showing that repeating FDG-PET one year after a first indeterminate scan accurately predicts progression to dementia within three years. However, this study was restricted to assessing FDG-PET accuracy in predicting MCI progression to dementia related to AD, and not to other atypical syndromes. Moreover, it did not test different inter-PET time lapses, or evaluate the practical impact of FDG-PET repetition in a tertiary center setting.
The aim of this study was therefore to explore the clinical benefits of a second FDG-PET on the differential diagnosis and management of patients with atypical presentations in a tertiary care memory clinic setting. Secondarily, we sought to identify the optimal delay when FDG-PET scans should be repeated. This information is important to either support the use of repeated FDG-PET in atypical dementia or to prevent patients from undergoing unnecessary radiologic exams.
METHODS
Patient selection
We retrospectively reviewed the files of all patients followed at our tertiary memory clinic who underwent an FDG-PET between January 2007 and June 2014. 550 FDG-PET were found from a total of over 5,000 patients seen at the memory clinic. Among these, 55 patients were found to have undergone two FDG-PET and 4 to have undergone three FDG-PET. All patients had an atypical presentation, for which the clinician still faced diagnostic uncertainty after extensive clinical evaluation, blood tests, neuropsychological testing, structural imaging, and a first inconclusive FDG-PET. ‘Atypical dementia’ is defined in the context of this study as any clinical syndrome whose diagnosis remains unclear after detailed tertiary-care diagnostic work-up. These atypical dementia cases revealed underlying frontotemporal dementia, primary progressive aphasia, Alzheimer’s young-onset focal variants (frontal, visuospatial, language), corticobasal syndrome, progressive supranuclear palsy, etc. Table 1 shows the patient characteristics.
FDG-PET imaging
Patients fasted for at least 4 h before administration of 18F-FDG. The serum glucose level was measured for all patients. The 18F-FDG brain PET was obtained either with a dual-head Philips Vertex MCD-AC coincidence camera (Philips Healthcare, Best, Netherlands), Siemens Biograph 6 (Siemens Healthcare, Erlangen, Germany) or GE Discovery VCT-16 (GE Healthcare, Cleveland, Ohio, USA) PET/CT scanners. In a dimly lit room, between 111 and 370 MBq (3–10 mCi) 18F-FDG were injected via a venous catheter. Between 30 to 60 min uptake time was observed before starting the acquisition (gamma camera: 64×64×16 matrix, 64 steps, 25 seconds/steps; PET/CT: 3D mode, 10 or 15 min). Measured attenuation and scatter corrections were applied to the iterative reconstruction method. For image analysis, dementias were classified mainly by visual rating according to generally accepted criteria: for recent review see Brown RK et al. [17]. Main categories included 1) AD (or AD variant): metabolic reduction primarily involving parieto-temporal regions and posterior cingulate gyrus/precuneus [14, 18]; 2) Dementia with Lewy bodies: parietal and occipital hypometabolism with sparing of posterior cingulate gyrus [19, 20]; 3) Frontotemporal lobar degeneration: hypometabolism in frontal and anterior temporal regions in the behavioral variant [21], in the anterior temporal lobe in semantic dementia [22] and in the left dorsolateral and dorsomedial prefrontal cortex in progressive non-fluent aphasia [23]; 4) Corticobasal degeneration: asymmetric dorsal frontal and anterior parietal metabolic reduction with or without ipsilateral striatum and thalamus hypometabolism, greatest on the side of the brain contralateral to the most affected limbs [24]. Semi-quantitative analysis using Neurostat/3D-SSP software was available to corroborate visual findings. Since cerebral FDG-PET criteria have greatly evolved in the last years, scans interpreted in 2008-2009 could not as efficiently distinguish for instance primary progressive aphasia variants, but could still very well distinguish AD variants from non-AD pathologies on the basis of posterior cingulate and temporoparietal hypometabolism [14, 18]. Finally, since this was a retrospective effort, images were evaluated by a number of different readers with variable experience who were not blind to the clinician’s diagnostic hypotheses. This reflects the usual challenges faced inreal-life practice.
Clinical impact of the second FDG-PET
Two experienced behavioral neurologists (RWB, LV) retrospectively reviewed the medical records of each patient included in the study. They were given the medical notes of the last clinical evaluation before the second FDG-PET (prePET2) and were asked to provide a differential diagnosis, associated level of confidence on a scale of 0–100% (very low: 20% , low: 40% , moderate: 60% , high: 80% , very high: 100%). They were then given the result of the PET2 as well as clinical notes for subsequent evaluations and asked again to provide their diagnosis and associated level of confidence postPET2 (see Fig. 1 for a schematic representation). Both evaluations (prePET2 and postPET2) were compared to determine the impact of the second FDG-PET on the diagnosis. Clinicians were also asked to rate the results of the second FDG-PET as ‘useful’ or ‘useless’ to their differential diagnosis. They also had to determine which new data (FDG-PET or MRI, neuropsychological evaluation, clinical evolution, etc.) was the most useful for their differential diagnosis between prePET2 and postPET2. Both clinicians were blind to each other’s rating. Changes in primary diagnosis and treatment refinement (prescription or discontinuation of ChEI treatment) following FDG-PET2 were also assessed. Results were further analyzed according to three criteria in order to explore optimal conditions for FDG-PET repetition: 1) InterPET delay, 2) PrePET2 diagnostic confidence, and 3) PrePET2 Mini-Mental State Evaluation (MMSE) score.
Bowker’s test was used to determine if the global increase in diagnostic certitude following PET2 was significant. Inter-rater reproducibility was calculated using simple Kappa coefficient. The impact of interPET time lapse, prePET2 diagnostic confidence, and prePET2 MMSE score on the outcomes was analyzed using two-sided t-test. Regression analyses were conducted to evaluate the relation between interPET delay and FDG-PET2 clinical impact. Sensitivity analyses were conducted to evaluate the relation between pre-to-postPET2 delay on FDG-PET2 clinical impact. The study has been approved by the institutional review board.
RESULTS
Overall results
As outlined in Table 2, there was a significant increase in the clinician’s diagnostic confidence from 63% ±10 to 79% ±16 following the second FDG-PET (p < 0.001), reducing the number of uncertain diagnoses (≤moderate) from 46 (80%) to 20 (34%). Out of 59 scans, 43 (73%) of PET2 studies were considered useful by the clinicians, with an inter-rater reproducibility of 0.83. In 14 patients (24%), FDG-PET2 was associated with a change in primary diagnosis. FDG-PET2 also allowed ChEI initiation in 10 patients (17%) whose FDG-PET2 suggested AD pathology and treatment cessation in 3 patients (5%) whose PET2 showed a frontotemporal dementia pattern of hypometabolism. Details about diagnostic changes are presented in Table 3.
Impact of interPET delay
The mean interval between both FDG-PET was 20.8 months (SD: 12.4), ranging from 3 months to 5 years. InterPET duration markedly affected outcomes. Indeed, an FDG-PET repeated less than 12 months after the first (n = 17) was much less likely to be considered useful by clinicians (47%) than if it was repeated more than 12 months later (83%) (two-sided t-test, p < 0.0034) (see Table 2). Early FDG-PET repetition was also much less likely to result in a change of diagnosis (0% for <12 months versus 33% for ≥12 months) (two-sided t-test, p < 0.0259) or a change in treatment (0% for <12 months versus 31% for ≥12 months) (two-sided t-test, p < 0.03612). Regression analysis showed that interPET delay is statistically associated with FDG-PET2 utility (Chi-square test, p = 0.0034). However, when performing such analysis only on FDG-PET2 repeated ≥12 months after the first (n = 42), there is no longer significant association between interPET delay and FDG-PET2 utility (Chi-square test, p = 0.3514).
Impact of prePET2 diagnostic confidence and MMSE Score
When prePET2 diagnostic confidence was low (≤moderate, n = 46), FDG-PET2 was more likely to impact on patient care (24% treatment change versus 16% when prePET2 diagnostic confidence was high) and diagnosis (29% change versus 15% , respectively). Repeated FDG-PET were also substantially more likely to impact on treatment plan of mildly affected patients (MMSE ≥24:35% modification) than in more advanced patients (MMSE ≤23:7%). However, these differences did not reach statistical significance.
FDG-PET versus other complementary investigations
We asked our clinicians-raters to determine which test was the most useful in their differential diagnosis between prePET2 and postPET2 (including FDG-PET, MRI, EEG, blood tests, neuropsychological evaluation, clinical evolution, etc.). When FDG-PET2 was considered useful (43/59 cases), raters judged that the most useful test was FDG-PET itself in 91% of the cases. When FDG-PET allowed diagnostic change (14/59 cases) FDG-PET was not considered the most determinant test in only one case (positive EMG with normal FDG-PET). In cases where FDG-PET was not considered the most useful (16 cases), raters mentioned being more influenced by clinical evolution (3 patients), neuropsychological evaluation (3 patients), MRI (2 patients), EMG (2 patients), or ‘no useful test’ (6 patients; when prePET2 and postPET2 differential diagnosis remains the same).
Illustrative cases
Case 1: Tracking MCI progression to AD
A 58-year-old HIV-seropositive patient with psychiatric comortibities was evaluated for cognitive impairment throughout her hospitalization. A normal FDG-PET (Fig. 2A, B) drew serious doubts on the organic nature of her cognitive symptoms. Psychiatrists rather suggested a diagnosis of conversion disorder. However, deterioration in objective cognitive testing despite improvement of affective state was considered as an atypical evolution, and led the behavioral neurologist to order a second FDG-PET 18 months later. FDG-PET2 (Fig. 2C, D) showed a typical AD hypometabolic pattern, allowing an early diagnosis and ChEI initiation.
Case 2: Confirmation of frontotemporal dementia
A 50-year-old entrepreneur presented with progressive cognitive decline. He scored 17/30 on the Montreal Cognitive Assessment. His wife mentioned recent personality changes, in the form of apathy and global disinterest. The patient had recently had a violent car accident, but apparently the symptomatology was not temporally related to the event. The clinician initially suspected frontotemporal dementia, but also considered AD and traumatic encephalopathy. A first FDG-PET was equivocal on a possible bilateral frontal hypometabolism (see Fig. 3A). A second FDG-PET repeated 13 months later showed clear progression of frontal hypometabolism (see Fig. 3B) and allowed to diagnose frontotemporal dementia with high diagnostic confidence.
Case 3: Regressing hypometabolism
A 43-year-old mechanic was referred to our Memory Clinic by a psychiatrist to evaluate the possibility of frontotemporal dementia. The patient had suffered for 4 years of behavioral changes and had a compatible FDG-PET profile (Fig. 4A, B), which supported the diagnosis. However, by looking further at the history, he and his wife mentioned that he somewhat got slightly better after a few days off work. The patient mentioned a mild dull headache while at work, that would resolve at home. A diagnosis of solvent-induced chronic toxic encephalopathy (by paints used to repaint antique cars) was considered and the patient was temporarily withdrawn from his work place. A control FDG-PET, repeated 9 months later (Fig. 4C, D) was perfectly normal, with a regression of previously noted hypometabolic deficits. The patient was permanently withdrawn from his workplace and he progressively remitted from his cognitive symptoms. This case underscores that cerebral hypometabolism can be reversible and caused by unsuspected non-degenerative conditions.
DISCUSSION
This study retrospectively explored the clinical impact of a repeat FDG-PET scan in a sample of 59 atypical dementia patients in a tertiary care memory clinic. Our results show that repeating an FDG-PET can reveal a progression of regional hypometabolic patterns that can greatly reduce the number of unclear diagnoses (from 80% to 34%), while leading to diagnostic changes and altering the clinical management in a substantial proportions of cases (24% and 22% , respectively). We also identified factors associated with a greater impact of FDG-PET repetition: 1) interPET delay ≥12 months; 2) low pre-test diagnostic confidence; and 3) high MMSE score. In these conditions (n = 26), as much as 88% of repeated PET were considered useful (73% in overall sample).
It was previously reported that FDG-PET could help the diagnosis of atypical/unclear dementia cases [6–14], where it was shown to increase diagnostic accuracy [7, 10–14], reduce the number of unclear diagnoses, and refine patients’ treatment plan [6–9]. Notwithstanding the diagnostic clarification that FDG-PET allows, a significant proportion of scans (16–18%) in these studies were inconclusive or incongruent and left the clinician in a diagnostic dilemma [6–9]. For these difficult cases, it remained unclear what would be the next diagnostic step. Our results suggest that repeating FDG-PET 12 months later can improve the diagnostic accuracy and allow for a more appropriate treatment plan to be made.
Optimal conditions for a second FDG-PET
An important clinical insight from our study draws is the recommended time a clinician should wait before ordering a second FDG-PET. The retrospective nature of the study, with a wide range of interPET delays (3 months to 5 years), and the relatively large sample size for such a precise situation allowed us to draw clinically relevant conclusions regarding this question. It seems, according to our results, that below 12 months of delay the metabolic deficits do not have the time to progress enough to give additional information compared to the first. Indeed, most of the scans repeated in a short interval were read as roughly identical to the first. These results are concordant with Torosyan et al. study [16], which showed that FDG-PET repeated 12 months after the first could bring important additional information in predicting progression from MCI to AD. Torosyan and Silverman algorithm [25] also recommended a delay of at least one year before repeating FDG-PET in clinical setting, but this seemed more based on clinical insight than actual scientific evidence. Altogether, we provide enlightening evidence suggesting that clinicians should wait at least 12 months before repeating an FDG-PET in clinical practice.
Previous studies [6, 27] demonstrated that PET clinical impact was maximal when prior diagnostic confidence was low. Our results corroborate these findings. We also showed that using FDG-PET in milder stages (high MMSE) was more likely to impact patient care (diagnosis, treatment plan) than in more advanced stages, although most differences did not reach statistical significance. In other words, repeating FDG-PET in clinically advanced patients or when diagnostic uncertainty is low seems more likely to confirm clinician’s feeling and increase his confidence than to concretely affect the patient’s diagnosis and treatment. Therefore, given its cost and the exposition to radiation it implies, we should reserve the use of FDG-PET, and a fortiori the FDG-PET repetition, in selected atypical/unclear cases where diagnosis is uncertain and the disease is at an early stage.
AD in atypical/unclear dementias
As reported in previous studies [3, 29], a significant proportion of the so-called atypical/unclear dementias further reveal underlying AD pathology. In our sample of highly atypical/unclear dementia cases, as much as 44% (26/59) patients were further diagnosed with AD, often with one of its newly described focal variants (language AD, frontal AD, posterior cortical atrophy). This corroborates previous findings and underlines the importance of diagnosing early and efficiently these atypical dementias in order to prevent delay in treatment initiation, which deprives patients from enhanced cognition and delayed deterioration [4].
Multimodal use of imaging technologies
Whereas this study focuses on FDG-PET repetition, we acknowledge that many other imaging and complementary investigations should be considered in complex/atypical dementia cases: lumbar puncture and cerebrospinal fluid analysis, amyloid imaging, structural MRI, etc. Each of these have their pros, cons, and associated cost, hence underlining the importance of using them in a logical and cost-effective manner. Recently, Torosyan and Silverman proposed an algorithm for the optimal use of available neuroimaging modalities [25]. This algorithm notably states that amyloid imaging may assist in the differential diagnosis of patients whose FDG-PET scans cannot be clearly interpreted. We agree with most of this algorithm, but suggest on the basis of our results that repeating FDG-PET may also be a reliable alternative when a first FDG-PET is inconclusive, especially in resource-limited settings (where FDG-PET cost-effectiveness has been demonstrated: [30, 31]), in countries where amyloid imaging is not approved, and in clinical situations where amyloid imaging is less useful, such as in older patients or in differentiating lesional dementia from non-AD neurodegenerative disorders.
Limitations
Our study has some limitations. First, our modest sample size somehow limits the strength of our conclusions, many differences observed in our different subgroups (less/more than 12 months, low/high pre-test confidence, and MMSE) not reaching statistical significance. However, for such a specific clinical situation, we believe that our sample of 59 patients having received two or more FDG-PET in their diagnostic work-up is sufficient to draw clinically relevant findings. Second, the delay between pre- and postPET2 evaluations (mean: 7 months) may have led to an overestimation of FDG-PET2 impact, since clinical deterioration may have influenced the clinician’s differential diagnosis. Nevertheless, since MMSE pre- and postPET2 remained globally stable (24.4 to 23.8), we believe that clinical deterioration can hardly be accountable for most of the progression of clinician’s differential diagnosis. Furthermore, sensitivity analyses showed that the delay between pre- and post-PET2 evaluations was neither associated with FDG-PET2 clinical utility (Chi-square test, p = 0.27), nor with diagnostic (Chi-square test, p = 0.46) or treatment changes (Chi-square test, p = 0.41). Third, evaluating FDG-PET value by its influence on the clinician’s differential diagnosis, without histopathologic material or longitudinal clinical follow-up to compare, is an important limitation. In FDG-PET studies, pathologic confirmation is the gold standard. When FDG-PET disagrees with the clinical diagnosis, the correct pathological diagnosis is in fact more likely to be congruent with FDG-PET than with clinical diagnosis [13, 14], whereas incongruent FDG-PET findings were often rated as confusing or useless by the clinicians in our study. However, in FDG-PET studies using clinical diagnosis at longitudinal assessment [32–34] or postmortem diagnosis [13, 35], FDG-PET result showed overall high sensitivity and specificity for the different dementing disorders. Moreover, we should emphasize that the goal of the study was not to assess FDG-PET accuracy, which has already been studied (see [7] for a review), but rather evaluating its clinical impact in real-life practice. Fourth, the retrospective design of the study raises potential bias related to variability in report quality. Indeed, it is worth noting that FDG-PET readings were performed by multiple nuclear medicine specialists with variable experience in cerebral FDG-PET reading. Nevertheless, all nuclear medicine specialists were trained through the same residency program, and therefore share a similar visual reading approach. Moreover, the availability of semi-quantitative statistical analysis to corroborate visual findings helped reduce variability in report quality between readers and centers, being particularly helpful to moderately-skilled readers due to its high specificity [36, 37]. Evolution of FDG-PET criteria from 2008 to 2015 also induced variability in report quality, new criteria allowing increased specificity and refined differential diagnosis. Future efforts may therefore require a prospective setting, or standardized blinded reading of the scans to limit variability in report quality. Despite these shortcomings, we believe this study realistically reflects the day-to-day collaboration between dementia and nuclear medicine experts in complex cases of degenerative diseases.
Conclusion
Taken together, the results of this retrospective study suggest that when diagnostic uncertainty persists after extensive evaluation and a first equivocal FDG-PET, repeating FDG-PET imaging can greatly clarify differential diagnosis, allowing a better-tailored treatment and improved planning of care, particularly when repeated at least 12 months later, when prior diagnostic confidence is low and when the disease is still at an early stage. Such clarification contributes to a better understanding of how intelligent and evidence-based use of multimodal imaging technologies helps solving difficult diagnostic dilemmas in tertiary care memory clinics.
