The MoCA
Well-suited screen for cognitive impairment in Parkinson disease

We read the article by Dalrymple-Alford et al. who compared the diagnostic accuracy of the MoCa with the S-MMSE and the SCOPA-COG in healthy controls, PD patients without dementia (PD-N), PD patients with mild cognitive impairment (PD-MCI), and PD patients with dementia (PD-D). [1]

As developers of the SCOPA-COG, [2] we were interested in the comparison between the SCOPA-COG and the MoCa. The authors concluded that "all 3 mental status tests produced excellent discrimination of PD-D from patients without dementia, and PD-MCI from PD- N patients but [that] the MoCA was generally better suited across both assessments." The S-MMSE and MoCa were administered to all patients allowing head-to-head comparisons between both scales. In addition, it permitted the transformation of the data to other prevalence rates (see footnote c in Table 3).

However, the SCOPA-COG was not administered in 38 out of the 114 PD patients (37 PD-N and 1 PD-D). We were surprised that the diagnostic accuracy between the MoCa and SCOPA-COG was directly compared. A direct comparison of these parameters is only justified if the populations are identical; the comparison should therefore have been restricted to subjects in whom both instruments were administered.

Even at a constant specificity, a higher number of non-cases within the sample -as in the case of the MoCa- would affect the NPV and PPV. At higher specificity values, as in the current situation, this would have led to an increase of the NPV and a decrease of the PPV. The effect of these errors is likely to pertain to the comparison between PD-MCI and PD-N patients in particular, since here the SCOPA-COG was administered in less than half the number of PD-N patients compared to the MoCa. We ignored that a different specificity value could have been obtained for the SCOPA-COG, had all PD patients been assessed with this scale. This is very possible and would make the data even less comparable.

The authors can not draw valid conclusions regarding the difference in diagnostic accuracy between the MoCa and SCOPA -COG on the basis of the presented data. We encourage them to re-analyze the data while restricting the analysis to the population in whom both the MoCa and the SCOPA-COG was administered.

References

1. Dalrymple-Alford JC, MacAskill MR, Nakas CT, et al. The MoCA. Neurology 2010 Nov 9;75:1717-1725.

2. Marinus J, Visser M, Verwey NA, et al. Assessment of cognition in Parkinson's disease. Neurology 2003;61:1222-1228.

Disclosure: Dr. van Hilten served on the scientific advisory board and as a consultant for Novartis and GSK; received funding from the Netherlands Ministry of Economic Affairs (TREND [Trauma RElated Neuronal Dysfunction]; and is an editorial board member of Movement Disorders. Drs. Marinus and Verbaan report no disclosures.

On reading our study of screening instruments for cognitive impairment in PD, [1] Marinus et al. [2] suggested that we made an inappropriate comparison between the validity of the MoCA and their instrument, the SCOPA-COG. They correctly noted that these instruments were not administered to all patients.

The critical issue is their relative value in discriminating PD patients with relatively normal cognition (PD-N) from those meeting our criteria for Mild Cognitive Impairment (PD-MCI). Although not explicitly stated in our article, comparisons between the area under the curve (AUC) of any two instruments requires restriction to those individuals tested on both measures and this comparison was reported. The two instruments were equivalent when assessing dementia (PD-D) relative to no dementia, but a difference emerged in favor of the MoCA when examining the separation between PD-N and PD-MCI.

As we stated, the AUC difference in favor of the MoCA was 12% (95% confidence interval [CI], 0.3-24.0%, p=0.045, p.1720). [1] The AUC for the MoCA itself for those PD-MCI and PD-N patients tested on both MoCA and SCOPA-COG was 93% (CI, 83-98%). Therefore, the AUC was at least marginally better than when examining all patients tested on the MoCA (90%; CI, 82-95%). It is unlikely that the specificity of the SCOPA-COG for the PD-N versus PD-MCI comparison would improve with a larger sample for a screening cut-off because this instrument is sensitive to minor impairments evident in the PD-N group (Table 2). [1]

Relative to the values given for the whole sample in Table 3, [1] no diagnostic performance values worsened and some improved when the MoCA analysis was restricted to those PD-N and PD-MCI patients who were tested on both the MoCA and the SCOPA-COG. The cut-offs suggested for optimal screen, diagnostics, and maximum accuracy remained the same. However, specificity for the optimal screen cut-off increased from 75% to 86% while positive predictive value (PPV) increased from 61% to 73%; the PPV for the optimal diagnostic cut-off increased from 79% to 90%; and both specificity (75% to 86%) and PPV (61% to 73%) increased for the maximum accuracy cut- off. The base-rates used to estimate PPV and NPV were population base- rates, not sample base-rates, which will vary across different criteria especially in PD-MCI. [3]

This evidence suggests that our original conclusion of the value of MoCA regarding cognitive screening for PD is sound.

Reference

3. Dalrymple-Alford JC, Livingston L, MacAskill MR, Graham C et al. Characterizing mild cognitive impairment in Parkinson's disease. Movement Disorders. doi: 10.1002/mds.23592. published online: 1 FEB 2011.

Disclosures: See original article for full disclosure list.