Normalcy, like pathology, consists of a summation of various parts that, when viewed as a whole, indicate that all is as it should be with no exceptions. Normal, therefore, is often defined as the complete absence of abnormal. In dealing with pathologic processes on cross-sectional imaging, we often encounter cases in which the anatomic imaging appears normal, but the underlying function or metabolism is indeed abnormal. Thus “normal” becomes far more difficult not only to define, but to recognize as well. Is it not more appropriate to assume that for a tissue or organ to be truly normal, all of its parts must be normal in form, metabolism, and function?
To complicate matters even more, we are often asked in pediatric imaging research to compare a pathologic population to one that is proven to be “normal.” This normal population is often referred to as the “normal control population” and is used to assess just how far the pathologic group has strayed from the fold. The problem in pediatric imaging research, as in all imaging research, is that we are caught on a double-edged sword. Our research is often considered incomplete without a comparison to a normal control population, but we are also told that to sedate or expose a normal, healthy child to clinically nonindicated imaging for the sole purpose of securing normative data is unethical and often refused by the local Institutional Review Board. The use of normative data from a population of children undergoing a clinically indicated examination, but who are otherwise “normal,” is one logical way around this dilemma. Unfortunately, this practical solution is often snubbed by academic purists who argue, “How can you ever be sure these children are truly normal if they should happen to have a clinical problem that requires imaging?” One way around this would be to recruit “normal” subjects from among the families of these academic purists, or more practically, we could better define acceptable inclusion criteria for a normal control population that makes it easy to gather such data.
In the current issue of the AJNR, Choi et al (page 1354) attempt to deal with one part of this difficult issue by defining a range of values (peak area ratios) using magnetic resonance spectroscopy (MRS) for the allocortex and isocortex in the normal developing human brain (that is the hippocampal formation and the peripheral cortex, respectively). Their intent was to define the range of normal using MRS images from different regions of the developing brain at different ages. Their study consisted of 30 subjects in different age groups, who were defined as normal based on an appropriate developmental history, the absence of identifiable disease, a normal neurologic examination, and normal cross-sectional imaging. Single-voxel MRS using stimulated-echo acquisition mode was used to assess portions of the limbic cortex, often defined as the allocortex, versus parietal or frontal peripheral cortical regions, often referred to as the isocortex. This division was sound, as it is well known that these two regions perform differently both structurally and functionally. They are also regions that are often affected by disease states in the pediatric age group.
Their results revealed a trend of metabolic ratio values, which allowed differentiation of the two regions, and that is in agreement with previous work in this area. The presence of N-acetylaspartate/total Creatine (tCr) was found to be significantly lower in the allocortex compared to the isocortex; amounts of choline/tCr and myo-inositol/tCr were found to be significantly higher in the allocortex compared to the isocortex. These trends give us further insight into the differences between these two distinct regions, and provide normative data, which can be used to characterize a pathologic process when anatomic changes may be absent. Obvious weaknesses do exist in their work. First, the number of subjects enrolled in the study is insufficient, especially when different age groups are taken into consideration. The second weakness is the rather loose documentation of normalcy in their study group, which is a topic we should explore further.
Let me begin by saying the work by Choi et al, in my opinion, is a valid contribution to neuroscience. Despite the two issues I have raised, which are common to many similar studies, this work represents a well-performed study with valid preliminary information toward creating a normative database for MRS in specific regions of the developing brain. How many subjects are sufficient is always a critical issue. Clearly, 30 subjects are not enough for a normative database that may require hundreds, if not thousands, of observations to reach statistical significance. The real issue is when are subjects really normal? There is, of course, no definitive answer to this question, but some generalizations of practical importance can be made.
Normal for an imaging study must take into consideration both minimal clinical as well as imaging criteria. Clinical issues must be resolved for each subject to assure they fit a profile of clinical normalcy that is acceptable based on observation and examination, perhaps by more than one observer. In the case of children, this must include an adequate assessment of childhood development and achievement. For example, does the child function at an appropriate school level? Have appropriate developmental milestones been on time? While the methods to assess childhood development are complex, and acceptable standards for clinical normalcy remain controversial, no study to create a normative database using brain imaging should fail to provide adequate documentation of normal development. At the very least, such data should always be included for each subject even if it is not used to include or exclude subjects. For the same reason, documentation of a normal general and a normal neurologic examination is also essential, while recognizing that it is difficult to standardize such an examination or overcome interobserver variability in performing the examination.
Finally, what are minimal criteria for normalcy with respect to the imaging examination? Two approaches have some validity as well as pitfalls. The first is to assume that whatever we recognize by imaging is to be considered normal if the child is clinically determined to be normal. The second is to accept minimal criteria for normalcy based on imaging corroborated by a normal clinical assessment. While these two approaches at first sound similar, their outcome and the way subjects are recruited may be quite different. In the first approach, minor abnormalities revealed by imaging often may be found even if the child is considered clinically normal. These may take the form of congenital malformations that are often clinically silent. Such an approach always raises questions with respect to whether such normal control populations are indeed valid. This approach also fails to deal with the dilemma of recruiting normal children to have examinations, which may not be clinically indicated, and therefore cannot be sedated even if they are too young to hold still for the examination. In the second approach, such abnormalities would be excluded because they would fail to meet the criteria for normal imaging as well as a normal clinical examination. Recruitment using this approach allows one to include children undergoing clinically indicated examinations and use them as normal controls. As long as the imaging is normal and the clinical examination is likewise normal, these subjects may well fulfill the needs of a normal control population.
But how are we to deal with cases of structural normalcy with metabolic or functional compromise? This brings me back to my original point. Is it not more appropriate to assume that for the brain to be considered truly normal, it must be normal in form, metabolism, and function? The answer is yes, as one might expect. We cannot assume that the brain is normal based on anatomic definition alone, but the presence or absence of normal metabolism, and perhaps even function, must also be taken into consideration. Normal function may be assumed if the clinical examination is normal, which leaves us only to resolve metabolic issues. The work by Choi et al thus takes on an even more important role as we attempt to define normal control populations. We should begin to look, based on multi-spectral imaging, at what are or are not acceptable criteria for recruitment into a normal control population that does not tie our hands or limit our options. One could argue validly that as long as the clinical examination is normal, and the anatomic and MRS images are normal, a subject might be an acceptable normal control despite any clinical indication for the examination itself. By doing so, we may untie our hands with respect to identifying acceptable control populations that meet minimal standards, but whose data are easily collected in statistically valid numbers, while maintaining our ethical and legal obligations.