Author Response

We thank Giordano and colleagues¹ for their comments on our work.² Their letter raises several concerns that oblige us to respond. To begin with, we agree with the notion that our results must not be generalized to populations other than the one studied, and, at the same time, we do not recognize that we have suggested otherwise. Giordano et al do not provide an example in support of their critique, nor are we able to find one when scrutinizing the text. On the contrary, we frequently refer to the population under study throughout the article, including in the title. In the Discussion section, we also explicitly state that “the fact that our sample was homogeneous limits the generalizability of our results. In a strict sense, the results apply only to people with mild to moderate Parkinson disease (PD), and additional studies are needed to determine the validity of the test in people with more severe PD or other populations with balance problems.” In our opinion, the article is unambiguous in this regard.

Giordano et al move on to argue that our study did not cover the actual general performance of the Mini-Balance Evaluation Systems Test (Mini-BESTest), as severe or very severe balance deficits were not represented in our sample. However, we contend that the exact opposite may be true; our study indicated that the Mini-BESTest does not perform optimally in a sample with mild to moderate PD and that the general performance of the test cannot be taken for granted. Put another way, the results of a study on one population may not be uncritically generalized to another. In its aim to cover a wide spectrum of individuals with balance disorders, the Mini-BESTest might not be optimal for assessing balance in specific subpopulations.

As pertains to the Hoehn & Yahr scale, we would like to accentuate that a score of less than 3 does not exclude the possibility of balance impairments. The scale consists of one singular balance item (the pull test [ie, postural responses]) and has been used for almost half a century to help clinicians assess PD severity. As such, it is neither sensitive nor up-to-date for detecting balance problems in a broader sense. We used the scale merely for the purpose of selecting an appropriate sample for an intervention study—to help us identify and exclude individuals who would be unfit (with a score of >3) to partake in an advanced balance training program. Today, we understand that balance is multifaceted^3,4 and that balance problems emerge early in PD,⁵ being present even at the time of diagnosis.⁶ The importance of early interventions in PD has been emphasized in a recent Cochrane-report by Tomlinson et al⁷ and is acknowledged in the European Physiotherapy Guidelines for Parkinson's Disease.⁸ Indeed, the majority of balance training interventions for people with PD primarily include participants with a score of 2 or 3 on the Hoehn & Yahr scale.^7,8 Thus, there is a need for sensitive evaluation tools that may identify even subtle balance impairments.

Giordano et al also raise concerns about our sample size. Although we agree that a limited sample size could pose a threat to the external validity of the study, we would not go as far as to completely disqualify our results based on a “not large” sample size alone. The potential drawbacks of a small sample were addressed in our discussion; however, it deserves to be added that according to the COnsensus-based Standards for the selection of health Measurement Instruments (COSMIN) checklist,⁹ a sample size of 100 participants or more is considered excellent (the highest score on a 4-graded scale) for studies on structural validity. Regardless of whether one accepts the consensus-based guidelines or not, it appears to us somewhat inconsistent to consider our sample of 112 individuals insufficiently large and at the same time emphasize the importance of a validation study by King et al,¹⁰ who used an even smaller sample (N=97).

Additionally, King et al,¹⁰ in their study, did not investigate structural validity but rather convergent validity—a different aspect of validity entirely. We do not see a conflict between our study and theirs. On the contrary, the findings of King et al were taken into consideration in our discussion, along with several other related references, and we gave it the same credit as Giordano et al do when recommending the continued use of the Mini-BESTest despite its limitations.

The request by Giordano et al for additional information about our analyses is justified. Starting with reliability indexes, we have now calculated that the Person Separation Index (PSI)=1.78 and the Item Separation Index (ISI)=4.27. We interpret the relatively low PSI as indicative of the Mini-BESTest lacking in sensitivity to satisfactorily discriminate between low and high performers in the present sample, whereas the relatively high ISI might be taken to indicate that our sample is large enough to confirm the hierarchy of item difficulties.

We also ran, as suggested by Giordano et al, a parallel analysis using the psych R package and the following code: fa.parallel(data, n.iter=50, fa=“pc,” cor=“poly”). The analysis had difficulties finding a stable solution, probably due to the very low variance on some items, but over several runs, the recommended number of components varied between 3 and 6. This finding further indicates the presence of more than one dimension.

For the analysis of dimensionality (principal component analysis of the Rasch residuals), we present the following R-code (package eRm), as requested:

• pcmres <- PCM(data)

• ppar <- person.parameter(pcmres)

• res <- residuals(ppar)

• ev <- eigen(cor(res))

• ev[1]

Also upon request, the following MPlus syntax was used for the exploratory factor analysis (EFA):

• variable:

  • names are id best_1-best_14;

  • usevariables are best_1-best_6 best_8-best_14;

  • categorical are best_1-best_6 best_8-best_14;

  • missing are all (999);

• analysis:

  • type is efa 1 6;

  • rotation=varimax(orthogonal);

  • iterations=1,000;

  • convergence=0.00005;

  • coverage=0.10;

• output:

  • standardized tech4 tech8;

In addition to the statistical concerns addressed above, we want to clarify that, as much as we appreciate the reference to Wright and Linacre,¹¹ it was not our intention to remove any of the items; however, due to convergence issues, the EFA would not run unless item 7 was left out.² Moreover, the factor loadings for a 2-factor-solution that are asked for are provided in our Table 3, showing that the 2 factors comprise 8 and 5 items, respectively. The main division lies between items reflecting postural responses and the rest, which, in our opinion, does not seem completely random but perhaps even reasonable. Within this context, Giordano and colleagues' view of “reasonable contrasts” as something different from dimensions leaves us confused. If it were to be taken literally, no multifactor solution would be considered valid, even if it corresponded perfectly with the 4 respective dimensions of the BESTest.

Taken together, we still believe that our results demonstrate that the Mini-BESTest does not behave optimally, or at least not as originally intended, in a sample of people with mild to moderate PD. As to Giordano and colleagues' critique that our results do not concur with the literature, we do not think that earlier research findings should dictate the reporting of future research findings.¹² Pluralism in research is a good thing, and there are numerous examples in the history of science where earlier “truths” have been questioned by later findings—sometimes even rightfully so. With that being said, we do not claim that our study has disproven the unidimensionality and near-perfect performance of the Mini-BESTest. What we do claim, however, is that the unidimensionality and near-perfect performance of the test has not been proven once and for all. After all, as much as the actual items of the Mini-BESTest derive from a theoretically defensible, multidimensional model of balance control (ie, the BESTest),⁴ the Mini-BESTest itself is more loosely anchored in theory, illustrated not least by the fact that it remains unclear just what exactly the construct “dynamic balance” refers to.

• © 2016 American Physical Therapy Association