The recent focus on biomarkers in the diagnosis of Alzheimer's disease (AD) and its prodromal stage have created a need to translate research findings into tools for use in everyday clinical practice. Although AD and mild cognitive impairment (MCI) are commonly diagnosed using sets of clinical criteria [1, 2], MRI findings may aid the clinical diagnosis and may predict clinical progression [3]. New research criteria have recently been proposed for AD [1, 4] and MCI [2] that incorporate (disproportionate) medial temporal lobe or hippocampal atrophy on MRI as one of the supportive features.

Since its development in 1992, the medial temporal lobe atrophy scale (MTA) [5] has been used in numerous studies to score MTA qualitatively. The MTA score has been validated in multiple ways (for a review see: [6]), such as correlation with pathology and volumetric MRI methods, as well as in differentiating between AD and controls and prediction of progression to AD in MCI. As this scale is very easy to learn and can be quickly scored, it is suitable to be used in routine clinical practice. However, operationalization and standardization of MTA scoring to provide reliable and practical cut-off values were lacking, until now!

In this issue, Pereira et al. [7] make an effort to tackle this question by exploring the influence of different cut-off scores of qualitative medial temporal lobe atrophy scores in the classification of patients with AD and control subjects, and in prediction of progression of MCI subjects to AD in a 1-year follow-up period. In addition, they evaluated the influence of demographic and genetic factors on this classification.

For this purpose, they used a large number of subjects (n = 1143) of two prospectively recruited cohorts from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database ( and AddNeuroMed project [8, 9]. For all subjects, atrophy of the medial temporal lobe was scored for both sides of the brain by a single rater, blinded for diagnosis and age, using the Scheltens scale for MTA (0–4) [5]. The authors used two different cut-offs. An age-dependent cut-off for subjects <75 years, with a MTA score of ≥2 (in at least one hemisphere), was considered abnormal, whilst for subjects >75 years, a score of ≥3 (in at least one hemisphere) or more was considered abnormal. Secondly, a cut-off was calculated based on the average of the MTA scores of both hemispheres, with a resulting score ≥1.5 being considered abnormal, independent of age.

Using these cut-offs, the authors found a diagnostic accuracy (AD vs. controls) for both cut-off scores of about 76%, with a higher sensitivity for the ≥1.5 cut-off (84.5% vs 69.9%) and a higher specificity for the age-dependent cut-off (83.2% vs 68.1%), implicating that the ≥1.5 cut-off score classifies more control subjects incorrectly as AD patients. Most likely, these incorrectly classified subjects were healthy elderly subjects with age-related hippocampal atrophy. We have demonstrated in an earlier study that increasing age is associated with hippocampal atrophy in controls as well as in AD patients to the same extent [10]. In addition, we showed that early- as well as late-onset AD patients have the same amount of atrophy after correction for age, suggesting that the effect of AD on hippocampal volume equals the effect of roughly 17 years of ageing [10]. This is confirmed by this study showing that adding an alternative cut-off score of ≥ 2 (based on the average score of both hemispheres) for older AD subjects (>75 years) increases the sensitivity to 81.5% for this group. This is in contrast to a lower sensitivity of 63% in the older age group when applying the age-adjusted cut-off score of ≥3 based on at least one hemisphere, wrongly classifying AD patients as controls, suggesting that this cut-off is too strict. Another reason in favour of using the average score of both hemispheres instead of the mostly severe side is the typically symmetrical distribution of underlying pathology in AD. Using the average will classify strongly asymmetrical scores (such as a score of 0 and 2 in all age groups or 0 and 3 in the older group) as non-AD cases. For the above reasons, one could conclude that the optimal cut-off value for differentiating AD from controls is ≥1.5 in younger (<75 years) and ≥2 in older (>75 years) patients, based on the average score of both hemispheres.

Secondly, the authors implemented the different cut-off scores in the classification of MCI subjects remaining stable and MCI subjects converting to AD at 1-year follow-up. Using the ≥1.5 cut-off 75.8% of converters was correctly classified; however, in stable MCI, more than 50% (56%) was misclassified as AD-like. This was mainly due to higher age and a higher number of subjects with the ApoE e4 allele, both associated with more severe medial temporal lobe atrophy [10,11 ]. As follow-up was restricted to only 1 year, it is possible that some of these subjects would have converted to AD in the next years [11]. Using the age-dependent cut-off, only 60% of converters were correctly classified and 65.5% of stable MCI was recognized. Unfortunately, no data were presented on the prediction of conversion in MCI classification of the additional cut-off of ≥2 based on the average of both hemispheres, as was performed in the first analysis; classifying AD versus controls. This might have decreased the number of controls that were classified as AD-like. However, it remains evident that whichever cut-off score is chosen, MTA scoring is helpful but never sufficient by itself to predict AD in MCI for the individual patient.

This study not only assessed age, but also ApoE e4 status in relation to MTA scores. From previous studies [12], it is clear that ApoE e4 is the most important genetic risk factor for sporadic AD. The ApoE e4 genotype drives the pathology to the medial temporal lobe, evoking a typical amnestic presentation, even in early-onset patients [12]. The atypical presentation of early-onset AD presenting with focal neurological deficits such as language and visuospatial deficits remains less common and is less often associated with ApoE e4 [13]. This study confirms this by showing that early-onset APOE non-carriers had less severe MTA in comparison with APOE carriers.

Overall, this is the first study evaluating MTA cut-off scores and measuring performance of MTA in diagnosing and predicting AD in clinical practice, in two large cohorts. It shows the effects on diagnostic accuracy of using different cut-offs, stressing that in certain cases, one should be careful with interpretation of MTA scores. Clearly, older age is associated with more atrophy, suggesting that an age-based cut-off to correct for this effect would be preferable. This study affirms that young patients with a non-memory presentation (ApoE E4 non-carriers) have relatively spared medial temporal lobes, which is an important clinical caveat. With this knowledge in mind, MTA scores and cut-offs can be safely and easily implemented.

This study suggests that the optimal cut-off is a score of ≥1.5 in younger (<75 years) and ≥2 in older (>75 years) patients, based on the average score of both hemispheres. The authors are to be commended to provide values that can now be used and validated in clinical settings, which will further improve clinical care for these patients. Following the French example [14], we hope that other countries will adopt this tool in clinical practice, now helped with age-related cut-off scores.

Conflict of interest statement

  1. Top of page
  2. Conflict of interest statement
  3. References

No conflict of interest to declare.


  1. Top of page
  2. Conflict of interest statement
  3. References