The Influence of Cochlear Volume on Temporal Changes of Impedance among Cochlear Implant Patients

Round 1

Reviewer 1 Report

The influence of cochlear volume in temporal changes of impedance among cochlear implant patients

Major comments:

The manuscript presents a study with an important clinical value, investigating 1) cochlear volume, 2) impedance of basal, middle, and apical sets of electrodes in a single type of array, all implanted by the same surgeon and using the same processing strategy, and 3) audiometric thresholds. This was conducted in a population of 34 CI recipients (arguably, very heterogeneous sample) between 2008 and 2017, at a single site in Brazil.

The paper is well written and concise. The goal is clearly defined (the rationale could be expanded in some respects – see comments below) and the methodology appropriately addressing it. The analysis, on the other hand, could be strengthened. Typically, the statistics reported for Figure 3 and 4 should be revised. The findings would be presumably more solid than those currently acknowledged, provided that the authors considered a full regression model with place, time, and mode. I advise the authors to treat time as a continuous variable, rather than categorical, although not necessarily linear. But if the authors wish to continue with a categorical variable (e.g. to make a point about the peak of impedance at 3 months), then a repeated-measures ANOVA with 3 within-sub factors would provide better statistical power than what the authors have at the moment. In the same vein (i.e. trying to strengthen the measures reported), many correlation attempts could be condensed to make the story simpler. The discussion is bringing relevant literature and direct comparisons with similar studies. I have noted a few comments throughout the text below. But overall, I believe the conclusion is safely supported by the current data and this paper would be of interest to the readership of the Journal of Otorhinolaryngology, Hearing and Balance Medicine.

Comments throughout the manuscript:

Line 18: assessment of … audiometric thresholds and electrode impedance?

Line 34: typo “well-established device”

Line 56: Please unfold the rationale implied by study [15] or [16]. Is it that a large diameter (and correspondingly a larger volume) give more room for the electrode array to be inserted with relatively little damage? This might depend on the skills of the surgeon, presumably?

Line 57: typo “length”

Line 63: This sounds very relevant. Can you expand: what is seen in children then?
Line 64-67: same; it would help to know more about the findings in these studies. Are impedance values increasing, decreasing, unchanged, and what caused possible discrepancies across these studies?

Line 70: I like that the introduction goes straight to the point, but I do think the rationale could be substantiated somewhat. Do the authors think that the studies aforementioned are not sufficient, or limited in their interpretation somehow? Why adding a measure of cochlear volume will help us better understand the changes in electrode impedance? Why is it critical to track these measures over >1 year? Why is one year enough?

Also, clarify that cochlear volume is measured only once, prior to CI surgery. (I believe that several manufactures are working towards making some devices MR-compatible; so this kind of work could perhaps one day be feasible longitudinally.. not that cochlear volume would be expected to change dramatically, but best to clarify in any case.)

Figure 1: I appreciate the authors’ desire for transparency, but a zoom-in view of each cochlea without showing all the software functionalities would be more interesting / revealing.

Line 99: This is key. Could you add how many surgeries this surgeon had already conducted by the time they were involved in this project?

Line 100: How many?

Line 102: Why excluding patients with bilateral implants? It would be nice to have a larger sample… (in fact, it would be nice to have two measures of thresholds and impedance, within the same participants).

Table 1: There’s a division between participant 1-13 and 14-34. I cannot quite figure out why?

Table 2: Fix the formatting of this table; some of the decimals are going to the next line and make it confusing.  Is Figure 2 an average of left and right thresholds? Maybe this could be condensed, avoiding Table 2 and 3 for simplicity. (I don’t think anyone will doubt that the surgery was effective.)

Line 141: The statement “cochlear volume of the implanted ear” reinforces the doubt I had earlier about MR-compatible device. I strongly suspect that this measurement was done pre-implantation, simply on the ear that was later going to receive the electrode array. Can you confirm for sure? (With a MR-compatible CI, there should be so much artifact that I doubt one could accurately conduct the work described here. So, I urge the authors to repeat and make very explicit that this is not done with the implant sitting inside the cochlea).

Line 142: A plot of cochlear volume estimations across the 34 subjects, perhaps simply as a function of their age at surgery (or anything showing individual data points) would be welcome.

Line 144-153: This paragraph is not great, e.g. making a key statement (that was in the abstract) before saying “although it was not significant”. But there are certainly ways to improve this analysis. For a start, you have 4 time points so you would gain statistical power by considering them as 4 levels in a continuous variable, rather than making individual comparisons (which also suffer from Bonferroni adjustments that were omitted). It seems to me that you have a typical dataset for a regression model with 2 fixed factors: place (basal, middle, apical) and time (t0, t3, t6, t12). The effect of time seems roughly consistent for middle and apical, but certainly different for basal electrodes, so you should see a neat interaction between place and time.  Also, note that the effect of time doesn’t need to be defined as linear. For example, it seems to me that a curvature might show up for the basal electrodes around the 3-month mark, so this would be interesting to prove (and perhaps with a quadratic function, you would be able to pinpoint a slightly more refined peak, perhaps around 4 or 4.5 month, even though nothing was measured at this timepoint – it would be a prediction of your model considering the evolution of impedance over the full year).

Why is Figure 4 small compared to Figure 3? Also, note that mode of testing, CG vs MP1+2 could itself be added as an additional within-sub factor. This would let you test whether any aspect of the data differs between the two modes. It seems roughly similar to me, so here again you would gain power.

Line 161: technically, the correlation coef is r; r² is the variance explained. But you are now showing 24 correlation attempts for CG, and 24 others for MP, all without correcting for the inflation of type1 error. I recommend you to state that impedance estimations were strongly correlated between CG and MP, and thus averaged together to limit the number of correlation attempts.  Then, because the t3, t6, and t12 are likely also correlated with one another, you could condense them as well. For example, if you follow a regression-based approach and extract the main effect of time (either linear or quadratic) at each location, then you have only 1 (or perhaps 2) value per participant which reflects the progressive change from t0. Thus, you might be left with much fewer correlation attempts, still highlighting the distinction between basal and middle/apical that was emphasized in the text.

Line 185: this was a surprising finding. What do you think is causing it?

Line 206-207: I understand but why would the same explanation not hold for middle or apical regions of the cochlea? Presumably, they have also been forced/damaged to some degree (especially considering it gets narrower and narrower). So, how is this interpretation explaining the distinct patter for basal vs apical regions?

Line 239-241: indeed; thresholds tended to be even better for the left ear, which had the smaller volume. But a plot of thresholds as a function of volume would address this, along with a correlation attempt. Is there really no relationship between cochlear volume and averaged audiometric thresholds across participants?

Line 251: It wasn’t made explicit how many participants were excluded.

Line 253: it’s a shame.

Line 254-256: not a limitation; that’s a good strength of the study.

Overall, congratulations on an interesting research project and an enjoyable reading.

------------------------------------------------------

Author Response

Thank you very much for the opportunity and for your time reviewing our manuscript.

We have addressed each of the topics that concerned you in order to improve reading and comprehension of our study.

Line 18: assessment of … audiometric thresholds and electrode impedance?

We deleted the part of audiometric thresholds.

Line 34: typo “well-established device”

We changed for "helpful tools".

Line 56: Please unfold the rationale implied by study [15] or [16]. Is it that a large diameter (and correspondingly a larger volume) give more room for the electrode array to be inserted with relatively little damage? This might depend on the skills of the surgeon, presumably?

That's right. Previous authors have found that larger volume was associated to residual hearing preservation.

Line 57: typo “length”

We've corrected

Line 63: This sounds very relevant. Can you expand: what is seen in children then?

We added the sentence: "It is seen elevation of thresholds, comfortable levels, and dynamic range during the first six months, thereafter levels stabilized.

Line 64-67: same; it would help to know more about the findings in these studies. Are impedance values increasing, decreasing, unchanged, and what caused possible discrepancies across these studies?

We added the sentences: "Electrode impedance values decreased from implantation to the first month, then also stabilized. It is also observed that fibrosis and new bone formation, that might occur up to one year after surgery (leading to a reduced cochlear volume), is negatively correlated to the postoperative word recognition."

Line 70: I like that the introduction goes straight to the point, but I do think the rationale could be substantiated somewhat. Do the authors think that the studies aforementioned are not sufficient, or limited in their interpretation somehow? Why adding a measure of cochlear volume will help us better understand the changes in electrode impedance? Why is it critical to track these measures over >1 year? Why is one year enough?

Previous studies have used different approaches. Our study have addressed the impedance values in a group of patients with soft surgery insertion technique. We understand that up to one year after the surgery, the fibrosis process is finally stable, therefore, it was not necessary to evaluate patients after this endpoint.   

Also, clarify that cochlear volume is measured only once, prior to CI surgery. (I believe that several manufactures are working towards making some devices MR-compatible; so this kind of work could perhaps one day be feasible longitudinally.. not that cochlear volume would be expected to change dramatically, but best to clarify in any case.)

The MRI was obtained only once, prior to the surgery.

Figure 1: I appreciate the authors’ desire for transparency, but a zoom-in view of each cochlea without showing all the software functionalities would be more interesting / revealing.

We edited the figure

Line 99: This is key. Could you add how many surgeries this surgeon had already conducted by the time they were involved in this project?

We also added this information (435 surgeries)

Line 102: Why excluding patients with bilateral implants? It would be nice to have a larger sample… (in fact, it would be nice to have two measures of thresholds and impedance, within the same participants).

Patients were included into 2 different studies, and in the second side it was not performed soft surgery technique.

Table 1: There’s a division between participant 1-13 and 14-34. I cannot quite figure out why?

We combined both tables.

Table 2: Fix the formatting of this table; some of the decimals are going to the next line and make it confusing.  Is Figure 2 an average of left and right thresholds? Maybe this could be condensed, avoiding Table 2 and 3 for simplicity. (I don’t think anyone will doubt that the surgery was effective.)

We excluded tables and edited figure.

Line 141: The statement “cochlear volume of the implanted ear” reinforces the doubt I had earlier about MR-compatible device. I strongly suspect that this measurement was done pre-implantation, simply on the ear that was later going to receive the electrode array. Can you confirm for sure? (With a MR-compatible CI, there should be so much artifact that I doubt one could accurately conduct the work described here. So, I urge the authors to repeat and make very explicit that this is not done with the implant sitting inside the cochlea).

MRI was performed only once before the surgery.

Line 142: A plot of cochlear volume estimations across the 34 subjects, perhaps simply as a function of their age at surgery (or anything showing individual data points) would be welcome.

We believe the manuscript is clearer in its current presentation.

Line 144-153: This paragraph is not great, e.g. making a key statement (that was in the abstract) before saying “although it was not significant”. But there are certainly ways to improve this analysis. For a start, you have 4 time points so you would gain statistical power by considering them as 4 levels in a continuous variable, rather than making individual comparisons (which also suffer from Bonferroni adjustments that were omitted). It seems to me that you have a typical dataset for a regression model with 2 fixed factors: place (basal, middle, apical) and time (t0, t3, t6, t12). The effect of time seems roughly consistent for middle and apical, but certainly different for basal electrodes, so you should see a neat interaction between place and time.  Also, note that the effect of time doesn’t need to be defined as linear. For example, it seems to me that a curvature might show up for the basal electrodes around the 3-month mark, so this would be interesting to prove (and perhaps with a quadratic function, you would be able to pinpoint a slightly more refined peak, perhaps around 4 or 4.5 month, even though nothing was measured at this timepoint – it would be a prediction of your model considering the evolution of impedance over the full year).

We performed ANOVA and we believe it is more concise now.

Why is Figure 4 small compared to Figure 3? Also, note that mode of testing, CG vs MP1+2 could itself be added as an additional within-sub factor. This would let you test whether any aspect of the data differs between the two modes. It seems roughly similar to me, so here again you would gain power.

We edited the figures to be at the same size.

Line 161: technically, the correlation coef is r; r² is the variance explained. But you are now showing 24 correlation attempts for CG, and 24 others for MP, all without correcting for the inflation of type1 error. I recommend you to state that impedance estimations were strongly correlated between CG and MP, and thus averaged together to limit the number of correlation attempts.  Then, because the t3, t6, and t12 are likely also correlated with one another, you could condense them as well. For example, if you follow a regression-based approach and extract the main effect of time (either linear or quadratic) at each location, then you have only 1 (or perhaps 2) value per participant which reflects the progressive change from t0. Thus, you might be left with much fewer correlation attempts, still highlighting the distinction between basal and middle/apical that was emphasized in the text.

We corrected the data presentation.

Line 185: this was a surprising finding. What do you think is causing it?

Probably because of the fibrosis. 

Line 206-207: I understand but why would the same explanation not hold for middle or apical regions of the cochlea? Presumably, they have also been forced/damaged to some degree (especially considering it gets narrower and narrower). So, how is this interpretation explaining the distinct patter for basal vs apical regions?

Because surgical trauma is mainly found in basal turn, according to histopathological studies.

Line 239-241: indeed; thresholds tended to be even better for the left ear, which had the smaller volume. But a plot of thresholds as a function of volume would address this, along with a correlation attempt. Is there really no relationship between cochlear volume and averaged audiometric thresholds across participants?

We couldn't find this correlation. 

Line 251: It wasn’t made explicit how many participants were excluded.

Almost 400.

Line 253: it’s a shame.

Line 254-256: not a limitation; that’s a good strength of the study.

Overall, congratulations on an interesting research project and an enjoyable reading.

Reviewer 2 Report

The authors present the results of a retrospective study examining the relationship between the electrode impedance measurements and Cochlear Volume deducted post-operatively using an MRI scan in Cochlear Implant recipients. The results are not novel, the impedance changes in time have been evaluated by many groups, the relationship with cochlear volume is interesting and provides a novel aspect. Some improvements are necessary, 

English revision is necessary - Thirty-four not third-four etc.

Detailed comments below:

Methodology:

1. The authors state this is a retrospective study in the ethical section at the end of the manuscript, yet the residual cochlear volume was obtained from an MRI scan after cochlear implantation. MRI scans after CI surgery cannot be considered routine, the CI produces an artifact which distorts the inner ear image. Please explain: 1. was the MRI obtained before or after implantation? 2. If MRI was conducted after implantation how did you cope with the magnetic distortion in the images? 3. Cochlear volume was calculated using software and the inner ear spaces were outlined by the software oparator, potential error in measurement exists, did the authors perform one or more volume calculations? Is the value an average of these measurements? 

2. Statistical methodology: It is stated that correlations were performed using Pearson correlation, yet paragraph 3.4 uses correlation coefficient for goodness of fit r-squared and a linear regression model. 

Is this a linear regression model or Pearson correlation providing r-squared values? Please clarify statistical methodology.

Results:

146-147 if the increase in impedance is statistically insignificant it is interpreted as “no difference” not an increase 

Figure 3. Brackets indicate significant difference for apical electrodes in left and right ear (decrease in time); for left ear only: decrease for middle turn electrodes to to t12 and decrease for basal t3 to t12; no significant differences between other electrodes in time. 

“ We also observed an increased middle electrodes impedance from t0 147 to t3, except for MP1+2 mode on the left ear which we observed a decreased impedance 148 value. ” Fig 3 - this is not what can be read from the figure - middle electrodes decrease from t0 to t12? 

Line 260-261: Our results are consistent with the hypothesis that newly formed fibrous tissue results in a progressive increasing of impedance. - this is contradictory to results in figure 3 and 4. In general only significant difference was noted for apical electrodes and it is a decrease in impedance.

Could the authors maybe provide mean data for impedance measurements and p values in table as supplementary materials? Proper statistical methodology would involve ANOVA test and post-hoc analysis to assess differences in 4 time points. 

Author Response

Thank you very much for your time and comments on our manuscript. 

We hope our modifications could improve the reading and comprehension towards our study.

English revision is necessary - Thirty-four not third-four etc.

We corrected this mistake.

Detailed comments below:

Methodology:

1. The authors state this is a retrospective study in the ethical section at the end of the manuscript, yet the residual cochlear volume was obtained from an MRI scan after cochlear implantation. MRI scans after CI surgery cannot be considered routine, the CI produces an artifact which distorts the inner ear image. Please explain: 1. was the MRI obtained before or after implantation? Only once before surgery.
2. If MRI was conducted after implantation how did you cope with the magnetic distortion in the images? No MRI study was performed after CI surgery in order to make 3-D reconstruction images.
3. Cochlear volume was calculated using software and the inner ear spaces were outlined by the software oparator, potential error in measurement exists, did the authors perform one or more volume calculations? Is the value an average of these measurements? We performed the calculations with the help of two experienced radiologists, in order to avoid potential errors.

2. Statistical methodology: It is stated that correlations were performed using Pearson correlation, yet paragraph 3.4 uses correlation coefficient for goodness of fit r-squared and a linear regression model. 

Is this a linear regression model or Pearson correlation providing r-squared values? Please clarify statistical methodology.

We corrected this information. Indeed, it was Pearson correlation method.

Results:

146-147 if the increase in impedance is statistically insignificant it is interpreted as “no difference” not an increase 

We corrected this sentence.

Figure 3. Brackets indicate significant difference for apical electrodes in left and right ear (decrease in time); for left ear only: decrease for middle turn electrodes to to t12 and decrease for basal t3 to t12; no significant differences between other electrodes in time. 

“ We also observed an increased middle electrodes impedance from t0 147 to t3, except for MP1+2 mode on the left ear which we observed a decreased impedance 148 value. ” Fig 3 - this is not what can be read from the figure - middle electrodes decrease from t0 to t12? 

They decrease over time

Line 260-261: Our results are consistent with the hypothesis that newly formed fibrous tissue results in a progressive increasing of impedance. - this is contradictory to results in figure 3 and 4. In general only significant difference was noted for apical electrodes and it is a decrease in impedance.

Could the authors maybe provide mean data for impedance measurements and p values in table as supplementary materials? Proper statistical methodology would involve ANOVA test and post-hoc analysis to assess differences in 4 time points. 

Round 2

Reviewer 2 Report

Thank you for improving your work, some of my comments were not adressed in the cover letter but were implemented in the manuscript; a minor english check of the new content is required, other than that I have no furter recommendations.