Clinical Perspective in Normal Pressure Hydrocephalus

In the commentary on our paper concerning shunt-responsive normal-pressure hydrocephalus (NPH) in the August issue of the American Journal of Neuroradiology, Dr. Bradley stressed the importance of clinical perspective in medical studies (1). We agree with his intention, but we do not accept his clinical perspective. Although it is very important to view a study from a clinical perspective, the issue is—what perspective is important for the management of NPH? In this context, his comments need to be examined closely.

Dr. Bradley advised that “anyone publishing a new sign should ensure the sign has not been previously published.” He cited an article of George (2), and maintained that the finding of large Sylvian cisterns with NPH is not a new observation but a rehash of a previously reported finding, obtained this time with a “high-tech” tool. That article is actually a commentary on a report by Dr. Bradley and colleagues, published in the same issue of the AJNR (3). We carefully read it. George's commentary did not contain anything about large Sylvian cisterns with NPH, but did include criticisms of the value of the CSF flow void in determining shunt-responsive NPH. In Bradley's original paper, there was no statement about large Sylvian cisterns. Although a patient with moderate NPH (illustrated in Figure 1 of their paper) has large Sylvian cisterns, this was not mentioned, and they seemed unaware of the significance of the finding. We have found no report indicating that large Sylvian cisterns are supportive of the diagnosis of NPH. This is probably because the finding is seemingly counter-intuitive, although “it is not surprising given that NPH is a form of communicating hydrocephalus” (1). In our article (4), we have stressed the significance of this finding on clinical and imaging grounds.

The diagnosis of idiopathic NPH in the elderly remains a substantial issue in the field of dementia, memory disorders, neurosurgery, and neuroradiology (5, 6). Many patients with possibly curable NPH often are misdiagnosed as having Alzheimer's disease or vascular dementia. In fact, most of our subjects were referred to us with such diagnoses, even with cranial CTs showing prominent Sylvian cisterns. Large Sylvian cisterns, with focally dilated sulci, often are misinterpreted as atrophy. Recognition of NPH in those patients is of great importance. Determination of patients likely to respond to shunts and those who will not is important, and disregarding the possibility of NPH before further diagnostic workup would be unfortunate. Dr. Bradley, citing studies with colleagues (3, 7), claimed that their finding of an increased CSF flow void reliably determines those likely to respond to shunts from those unlikely to do so. Other investigators have not replicated their view (8, 9), and the CSF flow-void sign generally has not been accepted as a good indicator of shunt responsiveness (2, 5). In our 11 subjects, the CSF flow void sign was noted in only six patients on either proton density–weighted or cardiac-gated gradient-echo images or both. A comparison between shunt responders and nonresponders is obviously important to identify predictive signs. A prospective trial in which all patients fulfilling some criteria, including those thought not to have NPH who also would undergo a shunting procedure, would be difficult to justify ethically, as perioperative mortality and risk of shunt-related complications are considerably high in the elderly. In addition, a retrospective study would suffer from the small number of patients with negative imaging sign or with negative shunt result, as Bradley discussed in his own article (7). In this circumstance, we believe that identifying MR features of idiopathic NPH in confirmed cases (ie, those who responded to shunts) is worthwhile. We specifically have selected candidates for shunts, and a positive response to shunt surgery is confirmatory for NPH (6).

Dr. Bradley claimed that 5-mm images with a 2.5-mm interslice gap were not appropriate for volumetry. As he pointed out, the 3D gradient-echo thin-slice images are now applied for hippocampal volumetry. The ability to resolve spatially two adjacent areas of brain structure is a function of the inherent spatial resolution of the images, and the imaging voxels must be of a size that is appropriately matched to the object being measured. We agree with Bradley's view that bad data subjected to sophisticated processing are still bad data. In fact, we are using 3D images and processing them mainly for research purposes (Mori et al Am J Psychiatry 1997; Mori E et al J Neurol Neurosurg Psychiatry 1997; Kitagaki et al Radiology 1998; Hashimoto et al Neurology 1998; Yasuda et al Am J Psychiatry 1998); however, unlike small structures such as the hippocampus, high spatial resolution images are not required for volumetry of large objects such as CSF spaces, which have a high contrast difference compared to neighboring structures. We chose this imaging technique for this reason and also for the practical availability of images and processing. One must allocate enormous time and labor to processing 3D images in the absence of an automated software system.

In summary, it remains necessary to maintain broad and objective clinical perspectives when evaluating new findings.