. The prevalence of cortical gray matter atrophy may be overestimated in the healthy aging brain. Neuropsychology. 2009 Sep;23(5):541-50. PubMed.


Please login to recommend the paper.


  1. In this paper, Burgmans et al., from Maastricht University, suggest that “The Prevalence of Cortical Gray Matter Atrophy May Be Overestimated in the Healthy Aging Brain.” Their key result is that subjects “who remained cognitively stable over a 3-year period…did not exhibit an age effect on the gray matter volume.”

    This intriguing study should be of interest to researchers working on the borderlands between healthy aging and neurodegeneration. It questions the received wisdom, which I subscribe to, that “normal aging” is associated with a small but definite year-on-year loss of cerebral volume. Although the decline into symptomatic Alzheimer disease is presaged by an increased (and accelerating) rate of atrophy (Chan et al, 2003), numerous papers have reported that even aging itself does involve a discernible and substantial rate of regional (e.g., hippocampal) and global cerebral loss (e.g., Fox and Schott, 2004; Driscoll et al., 2009).

    It can, of course, always be argued that that depends on how you define normal or healthy aging. In this paper the authors argue that “healthy” control groups used to study aging, or used in comparisons with patient groups such as those afflicted with AD, typically include some subjects who have latent problems. These subjects later show cognitive decline, and also exhibit atrophy prior to this clinically measurable decline—as is also found in MCI (McDonald et al, 2009) and in preclinical familial AD (Ridha et al., 2006). If, as is implied in Burgmans et al., you exclude subjects who are not 100 percent healthy, then you may no longer see age-related loss. Perhaps. But it might equally be that even healthy aging is associated with some losses, and that measuring loss reliably is challenging, with some study designs better suited to it than others.

    As statisticians are quick to point out, absence of evidence (i.e., a statistically significant effect, for the decline in volume with age here) is not evidence of absence. In other words, there might be an age effect in cognitively stable individuals, but the study did not find sufficient evidence for it. There is some suggestion that this could be the case in Figure 3 of this study, in which all regions show a negative slope to some (non-significant) degree.

    Lack of significance can result from a lack of statistical power, which in turn could arise from any of these reasons: small sample sizes, unaccounted confounds, or imprecise measurements. This study arguably suffers from all these to some extent.

    Only 28 cognitively healthy individuals who remained healthy contribute to the conclusion that the healthy brain does not shrink. There are less than 10 subjects per decade of life, and they are not spread evenly across the decades—for example, there are only three subjects under 65. It can also be seen from the figures that the 50-year-old person might have a big effect on the regression line. The authors were concerned about this: "Excluding the youngest participants did not change the results in the cognitively healthy group." However, removing the youngest participants reduces to just 20 years the range of ages over which one is trying to see an effect of age. It is perhaps not surprising, therefore, that the lack of evidence for a decline is not changed by removing subjects, since this further reduces the power to declare any apparent decline as significant.

    Regarding confounds, one needs young and old subjects to be well matched in order to discern the effects of age. Adjusting regional volumes through division by head size (TIV) assumes that all those regions scale with TIV in a linear y = x manner (without an intercept). I don't think that is naturally the case with human biology—does someone who is two metres tall have a spine that is twice as long as someone one metre tall?—the difference may be mostly in the legs! Hence, the method of adjustment for TIV may not have been ideal. Gender can also affect brain volume, so unless one believes the TIV adjustment to have removed any effect of gender, this should also have been considered.

    Regarding precision of the measurements, the cross-sectional nature of the study means it is weakened by the huge natural variation across subjects. Imagine if you were trying to measure changes in height based on a cross-sectional study; if you were unlucky enough to include some individuals who happened to be both shorter and younger than the mean, you might conclude we grew with age! The problem of controlling for between-subject differences is the reason why clinical trials have to randomize and include large numbers of subjects (and still need to adjust for baseline differences). Measuring within-subject change in a longitudinal study would be more powerful, and the authors indeed identify this as necessary further work.

    Precision is also often lower in automated methods compared to expert manual segmentations. It would be interesting to see if the paper’s findings are replicated in regional measurements, for example, of the hippocampus. It is also arguably easier and more reliable to measure whole-brain atrophy than gray-matter atrophy (especially since gray-white tissue contrast itself declines with age). While this would not provide such specific conclusions, a finding of significant whole-brain atrophy accompanying non-significant gray-matter atrophy (in carefully characterized subjects such as these) would tend to support the view that even (relatively) healthy aging does involve brain volume loss even if the extent of this could have been exaggerated in the literature, as argued by this interesting, thought-provoking paper.


    . Change in rates of cerebral atrophy over time in early-onset Alzheimer's disease: longitudinal MRI study. Lancet. 2003 Oct 4;362(9390):1121-2. PubMed.

    . Imaging cerebral atrophy: normal ageing to Alzheimer's disease. Lancet. 2004 Jan 31;363(9406):392-4. PubMed.

    . Longitudinal pattern of regional brain volume change differentiates normal aging from MCI. Neurology. 2009 Jun 2;72(22):1906-13. PubMed.

    . Regional rates of neocortical atrophy from normal aging to early Alzheimer disease. Neurology. 2009 Aug 11;73(6):457-65. PubMed.

    . Tracking atrophy progression in familial Alzheimer's disease: a serial MRI study. Lancet Neurol. 2006 Oct;5(10):828-34. PubMed.

  2. Age and Atrophy: Nothing to Worry About?
    In this paper, Burgmans et al. argue that previous studies suggesting that age causes cortical atrophy on MRI may be flawed because they included subjects who, in fact, had preclinical dementia. If true, this would seriously undermine all “corrections for age” that we usually make in imaging studies.

    But is it true? The authors included only individuals who did not decline over a period of three years after the initial scan. They started with 44 healthy individuals who underwent a scan and ultimately analyzed 35 who were still cognitively intact three years after the scan and compared them to 30 decliners. In fact, there were only 28, because there were another seven who were not cognitively stable in the three years after scanning. Cognitive functions were tested with MMSE and six tests used in the MAAS study from which the sample was drawn. So far so good, although the exact numbers remain difficult to interpret, and comparisons between the 30 decliners are made with both the 35 and the 28.

    The MRI measurements included seven critical regions, a.o. hippocampus, cingulated cortices, orbital and prefrontal cortices. The mean age of the groups was 69 at baseline (three years prior to scanning) but with a range between 50 and 80 roughly. The main message is that there was no age effect on the gray matter volume of the parahippocampal, cingulated, and frontal areas in cognitively healthy participants who remained cognitively stable over three years, whereas a significant age effect in these brain areas was found in participants with substantial cognitive decline. The hippocampus was the only structure in which an age effect was present in both groups, when comparing the 35 to 30, but that disappeared when they looked at the 28 super healthy, excluding the participants who were no longer cognitively stable three years after the scanning.

    Although interesting at first sight, there are problems with the study that hamper immediate acceptance of the results. First, MRI was done cross-sectionally, so the age effect was studied when comparing individuals of different ages, and not looking within individuals using serial MRI. A recent study using serial MRI in the ADNI data set showed an average annual hippocampal volume loss of 0.8 ± 0.5 percent in normal elderly (Schuff et al., 2009). A recent study from our own group also showed values between 0.3 to 0.6 percent annual change in different brain regions, including most of the ones studied by Burgmans et al., using serial MRI with a mean interval of 1.9 years (Sluimer, 2009).

    The other problem is power. To conclude that there is no age effect, power should be sufficient, especially when one considers 35 (or 28) subjects, seven brain regions, and 20 years age span, leaving a mere seven to eight subjects per five-year age stratum. These small numbers suggest a type 2 error (false negative) until proven otherwise.

    Finally, there is, of course, the issue of normality. How normal is normal? Selecting the happy few who remain normal over such a period, in neuropsychological terms, may be interesting, but it remains a question if that represents real normality. I am ready to accept that these individuals have a brain resistant to change over time, but only if shown in a large enough sample with proper serial MRI methodology.


    . MRI of hippocampal volume loss in early Alzheimer's disease in relation to ApoE genotype and biomarkers. Brain. 2009 Apr;132(Pt 4):1067-77. PubMed.

    . Accelerating regional atrophy rates in the progression from normal aging to Alzheimer's disease. Eur Radiol. 2009 Dec;19(12):2826-33. PubMed.

Make a Comment

To make a comment you must login or register.