Authors’ response

We thank Dr McCleery & Dr Quinn, from the Cochrane Dementia and Cognitive Improvement Group, for their interest in our recent paper on the accuracy of dopaminergic imaging as a biomarker for mild cognitive impairment with Lewy bodies (MCI-LB).¹ However, we disagree with their argument that we have not demonstrated [¹²³I]N-ω-fluoropropyl-2β-carbomethoxy-3β-(4-iodophenyl)nortropane (¹²³I-FP-CIT) to be useful in MCI-LB, because of biases arising from an imperfect reference standard and exclusion of patients with uncertain cases.

We agree that it is challenging to design a perfect diagnostic accuracy study in the field of neurodegenerative diseases, and that many studies, including our own, have limitations. However, we discuss these limitations in the paper and they do not detract significantly from our findings. A recent systematic review used Bayesian methods to adjust for the lack of neuropathological reference standard in most studies, finding similar accuracy results for the diagnosis of clinical dementia with Lewy bodies (DLB) with ¹²³I-FP-CIT before and after adjustment.² Our results of moderate sensitivity and high specificity are valid and demonstrate that a positive (abnormal) dopaminergic imaging result in suspected MCI-LB is indeed useful in supporting the diagnosis. We agree that a negative result does not exclude MCI-LB, but it is now well-known that a normal scan does not completely exclude DLB in patients with established dementia. Clinicians recognise that a test with sensitivity estimated at 66% is not going to identify all cases.

Drs McCleery & Quinn are, of course, correct that a clinical diagnosis of MCI-LB is an imperfect reference standard, but what is their alternative? The gold standard of post-mortem pathology diagnosis is not feasible for early-stage neurodegenerative studies. In our study we sought to make our clinical diagnoses as reliable as we possibly could by incorporating cardiac metaiodobenzylguanidine (mIBG) findings, where available, and using a panel of three old-age psychiatrists all specialising in Lewy body disease. Our participants have all been invited to donate to our tissue resource on death and in time we intend to update the results of this study with pathological diagnoses as the reference standard, where available.

Optimising our reference standard was our rationale for excluding patients with an uncertain follow-up diagnosis of possible MCI-LB. We are aware that excluding patients with the most uncertain cases risks partial verification bias,³ but in this case we strongly feel that the improvement to the quality of our clinical diagnoses and thus our reference standard outweighs the risk of overestimating sensitivity and specificity. On balance, our results are likely to be a worse estimate of the true diagnostic accuracy with the uncertain cases included than with them excluded. We continue to follow these participants up, but feel it is inappropriate to group them as having MCI-LB at present for the purposes of diagnostic accuracy assessment. This prospective review is the key to making our reference clinical diagnoses as accurate as possible.

We did not aim to image only patients with uncertain diagnoses at baseline, although we agree this would be a very valuable future study. Rather, we aimed to end up with two reasonably definitive groups of patients at follow-up – those who we could be as confident as possible do have MCI-LB and those who we are confident do not. In the absence of neuropathological verification, it is only by selecting patients with probable diagnoses that we can usefully test the accuracy of baseline dopaminergic imaging in MCI-LB.