Limits...
The reliability of repeated TMS measures in older adults and in patients with subacute and chronic stroke.

Schambra HM, Ogden RT, Martínez-Hernández IE, Lin X, Chang YB, Rahman A, Edwards DJ, Krakauer JW - Front Cell Neurosci (2015)

Bottom Line: Using variances estimated from a random effects model, we calculated the SDC and ICC for each TMS measure.TMS measures should be used based on their reliability in particular contexts.More work establishing their validity, responsiveness, and clinical relevance is still needed.

View Article: PubMed Central - PubMed

Affiliation: Motor Performance Laboratory, Department of Rehabilitation and Regenerative Medicine, Columbia University New York, NY, USA.

ABSTRACT
The reliability of transcranial magnetic stimulation (TMS) measures in healthy older adults and stroke patients has been insufficiently characterized. We determined whether common TMS measures could reliably evaluate change in individuals and in groups using the smallest detectable change (SDC), or could tell subjects apart using the intraclass correlation coefficient (ICC). We used a single-rater test-retest design in older healthy, subacute stroke, and chronic stroke subjects. At twice daily sessions on two consecutive days, we recorded resting motor threshold, test stimulus intensity, recruitment curves, short-interval intracortical inhibition, and facilitation, and long-interval intracortical inhibition. Using variances estimated from a random effects model, we calculated the SDC and ICC for each TMS measure. For all TMS measures in all groups, SDCs for single subjects were large; only with modest group sizes did the SDCs become low. Thus, while these TMS measures cannot be reliably used as a biomarker to detect individual change, they can reliably detect change exceeding measurement noise in moderate-sized groups. For several of the TMS measures, ICCs were universally high, suggesting that they can reliably discriminate between subjects. TMS measures should be used based on their reliability in particular contexts. More work establishing their validity, responsiveness, and clinical relevance is still needed.

No MeSH data available.


Related in: MedlinePlus

SDCgroupdecreases dramatically with increasing n. This example, using SICI80 in lesioned hemisphere of subacute stroke, demonstrates for that for an individual, the SDC is quite large, requiring a change greater than 0.45 in SICI80 in the lesioned hemisphere to be deemed real. By calculating SDCgroup (= SDCindiv/), it is apparent that even with modest sample sizes of 10–20, changes exceeding measurement error could be conceivably detected, and the measure would be considered reliable.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4555014&req=5

Figure 1: SDCgroupdecreases dramatically with increasing n. This example, using SICI80 in lesioned hemisphere of subacute stroke, demonstrates for that for an individual, the SDC is quite large, requiring a change greater than 0.45 in SICI80 in the lesioned hemisphere to be deemed real. By calculating SDCgroup (= SDCindiv/), it is apparent that even with modest sample sizes of 10–20, changes exceeding measurement error could be conceivably detected, and the measure would be considered reliable.

Mentions: Importantly, SDCindiv is provided to enable future investigators to generate SDCgroup for their samples of size n. Given that measurement error changes with context, it is imperative that investigators calculate the SDCgroup pertaining to each subject type, hemisphere, and sample size. As would be expected, SDCgroup shrinks dramatically as group size increases. A graphical example is given for SICI80 in the lesioned hemisphere of subacute stroke subjects (Figure 1). For SICI80, changes greater than SDCindiv = 0.46 would be required to exceed measurement error for an individual, but as n increases beyond 19, SDCgroup ≤ 0.10.


The reliability of repeated TMS measures in older adults and in patients with subacute and chronic stroke.

Schambra HM, Ogden RT, Martínez-Hernández IE, Lin X, Chang YB, Rahman A, Edwards DJ, Krakauer JW - Front Cell Neurosci (2015)

SDCgroupdecreases dramatically with increasing n. This example, using SICI80 in lesioned hemisphere of subacute stroke, demonstrates for that for an individual, the SDC is quite large, requiring a change greater than 0.45 in SICI80 in the lesioned hemisphere to be deemed real. By calculating SDCgroup (= SDCindiv/), it is apparent that even with modest sample sizes of 10–20, changes exceeding measurement error could be conceivably detected, and the measure would be considered reliable.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4555014&req=5

Figure 1: SDCgroupdecreases dramatically with increasing n. This example, using SICI80 in lesioned hemisphere of subacute stroke, demonstrates for that for an individual, the SDC is quite large, requiring a change greater than 0.45 in SICI80 in the lesioned hemisphere to be deemed real. By calculating SDCgroup (= SDCindiv/), it is apparent that even with modest sample sizes of 10–20, changes exceeding measurement error could be conceivably detected, and the measure would be considered reliable.
Mentions: Importantly, SDCindiv is provided to enable future investigators to generate SDCgroup for their samples of size n. Given that measurement error changes with context, it is imperative that investigators calculate the SDCgroup pertaining to each subject type, hemisphere, and sample size. As would be expected, SDCgroup shrinks dramatically as group size increases. A graphical example is given for SICI80 in the lesioned hemisphere of subacute stroke subjects (Figure 1). For SICI80, changes greater than SDCindiv = 0.46 would be required to exceed measurement error for an individual, but as n increases beyond 19, SDCgroup ≤ 0.10.

Bottom Line: Using variances estimated from a random effects model, we calculated the SDC and ICC for each TMS measure.TMS measures should be used based on their reliability in particular contexts.More work establishing their validity, responsiveness, and clinical relevance is still needed.

View Article: PubMed Central - PubMed

Affiliation: Motor Performance Laboratory, Department of Rehabilitation and Regenerative Medicine, Columbia University New York, NY, USA.

ABSTRACT
The reliability of transcranial magnetic stimulation (TMS) measures in healthy older adults and stroke patients has been insufficiently characterized. We determined whether common TMS measures could reliably evaluate change in individuals and in groups using the smallest detectable change (SDC), or could tell subjects apart using the intraclass correlation coefficient (ICC). We used a single-rater test-retest design in older healthy, subacute stroke, and chronic stroke subjects. At twice daily sessions on two consecutive days, we recorded resting motor threshold, test stimulus intensity, recruitment curves, short-interval intracortical inhibition, and facilitation, and long-interval intracortical inhibition. Using variances estimated from a random effects model, we calculated the SDC and ICC for each TMS measure. For all TMS measures in all groups, SDCs for single subjects were large; only with modest group sizes did the SDCs become low. Thus, while these TMS measures cannot be reliably used as a biomarker to detect individual change, they can reliably detect change exceeding measurement noise in moderate-sized groups. For several of the TMS measures, ICCs were universally high, suggesting that they can reliably discriminate between subjects. TMS measures should be used based on their reliability in particular contexts. More work establishing their validity, responsiveness, and clinical relevance is still needed.

No MeSH data available.


Related in: MedlinePlus