Diagnostic Considerations for Neurolymphomatosis: A Natural History Analysis

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The reviewed manuscript addresses a challenging problem in hematology very effectively and appears to be a strong fit for publication with minimal editing. The authors use an already established sample, namely, an analysis of 559 cases of neurolymphomatosis (NL) collected from 231 publications across many years (1951-2022). The fact that the statistics involve mutually exclusive categorical sub-analyses along with the Bonferroni correction makes them reliable. The clinical relevance of this paper is extremely high, as it addresses the main challenge in diagnosing a rare and aggressive condition, for which prompt diagnosis can be crucial. The study provides practically applicable conclusions: the only statistically significant variable for improving survival rate was biopsy, whereas imaging and neurophysiological techniques should be used solely to define biopsy sites. I suggest Minor Revisions Before Publishing.

Indeed, the manuscript presents minor data contradictions and minor discrepancies. Specifically, when stating that "Five times more secondary NL patients had a biopsy compared to those with primary NL," the authors do not take into account the number differences in non-biopsied (134 vs. 26) and biopsied (124 vs. 91) groups, meaning that the five-fold difference concerns the former group and not the latter one, in which the difference equals a 1.4-fold difference. Thus, the phrasing should be changed slightly.

There are small typos in the figures' captions (Fig 1).

To summarize, the manuscript is highly scientific, well-structured, and offers useful information about a hard-to-treat malignancy. After polishing the text and figure captions, it will be a wonderful addition to medical literature.

Author Response

Reviewer 1:

The reviewed manuscript addresses a challenging problem in hematology very effectively and appears to be a strong fit for publication with minimal editing. The authors use an already established sample, namely, an analysis of 559 cases of neurolymphomatosis (NL) collected from 231 publications across many years (1951-2022). The fact that the statistics involve mutually exclusive categorical sub-analyses along with the Bonferroni correction makes them reliable. The clinical relevance of this paper is extremely high, as it addresses the main challenge in diagnosing a rare and aggressive condition, for which prompt diagnosis can be crucial. The study provides practically applicable conclusions: the only statistically significant variable for improving survival rate was biopsy, whereas imaging and neurophysiological techniques should be used solely to define biopsy sites. I suggest Minor Revisions Before Publishing.

Response: We thank Reviewer 1 for his/her positive comments.

Indeed, the manuscript presents minor data contradictions and minor discrepancies. Specifically, when stating that "Five times more secondary NL patients had a biopsy compared to those with primary NL," the authors do not take into account the number differences in non-biopsied (134 vs. 26) and biopsied(124 vs. 91) groups, meaning that the five-fold difference concerns the former group and not the latter one, in which the difference equals a 1.4-fold difference. Thus, the phrasing should be changed slightly.

Response: We thank Reviewer 1 for pointing this out.  We agree that the wording in the two sentences is incorrect when we stated, “Five times more secondary NL patients (n=134) had a biopsy compared to those with primary NL (n=26). This difference is only 1.4 times between secondary (n=124) and primary (n=94) NL patients.”  We changed the wording to “Five times more secondary NL patients (n=134) did not have a biopsy compared to those with primary NL (n=29). Among those with biopsy, this difference is only 1.4 times between secondary (n=124) and primary (n=94) NL patients.”

There are small typos in the figures' captions (Fig 1).

Response: Figure 1 has been updated.

To summarize, the manuscript is highly scientific, well-structured, and offers useful information about a hard-to-treat malignancy. After polishing the text and figure captions, it will be a wonderful addition to medical literature.

Response: We thank Reviewer 1 for his/her positive comments.

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript presents a retrospective natural history analysis of neurolymphomatosis (NL), aiming to identify optimal diagnostic approaches using pooled data from 559 cases extracted from 231 publications. The topic is clinically relevant because NL remains diagnostically challenging and delayed diagnosis contributes to poor outcomes. The effort to aggregate rare disease data is commendable; however, important methodological limitations and interpretation issues require clarification.

Major Comments:

1. The association between biopsy and improved survival may be influenced by selection bias and unadjusted confounders.
2. Lack of multivariable analysis limits the validity of biopsy-associated survival conclusions.
3. Aggregate-to-individual data conversion requires stronger methodological justification.
4. Temporal heterogeneity across studies may affect diagnostic and survival outcomes.
5. Small subgroup sizes reduce statistical power and reliability.
6. Biopsy may reflect earlier diagnosis rather than a direct survival benefit.
7. Missing data reporting requires greater transparency.
8. Imbalance between primary and secondary NL cases may introduce confounding bias.
9. Evidence supporting PET + MRI + EMG-NCS as the optimal strategy remains limited.
10. Sensitivity analyses are needed to confirm the robustness of findings.

Minor Comments:

1. Figure quality is suboptimal; several boxplots and Kaplan–Meier curves are difficult to interpret and require improved resolution.
2. Figure 6 legend appears confusing:

“Five times more secondary NL patients had a biopsy compared to primary NL”
The wording should be revised for clarity.

3. The manuscript contains multiple grammatical issues and typographical inconsistencies (e.g., “beĴer”, “seĴing”), suggesting formatting or encoding problems requiring correction.
4. The term “treatment 1” should be defined more clearly at first use.
5. Abbreviations should be consistently defined upon first mention (e.g., EMG-NCS).
6. Supplementary Figure numbering appears duplicated (“Figure S2” repeated). This should be corrected.
7. The discussion comparing NL with intravascular lymphomatosis is interesting but somewhat speculative and could be shortened.
8. References are very few and very old. Author recommended must be cite recent citation.
9. keywords must be different from the title of the manuscript.

Author Response

Reviewer 2:

Major Comments:

1. The association between biopsy and improved survival may be influenced by selection bias and unadjusted confounders.

Response: We thank Reviewer 2 for pointing this out. It is difficult to adjust for bias and confounders because very little is known about prognostic factors in this population of NL.  Nevertheless, we did our best to minimize the confounding variables, specifically with regards to age and primary vs. secondary NL, which we found in our prior publication having prognostic value (please see below).

2. Lack of multivariable analysis limits the validity of biopsy-associated survival conclusions.

Response: We performed a Cox proportional hazard analysis to validate biopsy-associated survival, by adjusting for age and primary vs. secondary NL.

3. Aggregate-to-individual data conversion requires stronger methodological justification.

Response: We thank Reviewer 2 for pointing this out. Our use of aggregate vs. individual data conversion was debated during peer review of our prior publication (Xu E, et al. BJC Rep. 2024;2:34). The decision to deconvolute aggregate data into single points was to provide greater granularity of various accompanying data in the population. But, as Reviewer 2 correctly pointed out, this is done at the expense of generalizing and biasing some of the findings in the larger cohorts. Fortunately, we only have to do this for 3 publications with a total n=115, which is a minority or 20% of the population (Grisarius, et al. Blood. 2010;115:5005-5011 [n=50]; Khurana A, et al. Blood Adv. 2021;5:1379-1387 [n=40]; and DeVries AH, et al. Skeletal Radiol. 2019;48:1043-1050 [n=25]).  Therefore, we do not believe that this aggregate-to-individual data conversion will affect significantly our overall analysis and subsequent conclusions.

4. Temporal heterogeneity across studies may affect diagnostic and survival outcomes.

Response: We acknowledge Reviewer 2’s concern about temporal heterogeneity across studies. This is an inherent problem with rare diseases and the lack of uniform reporting criteria, particularly with the older studies before year 2000. However, this retrospective analysis from the literature is the only means of gaining a global view for this rare malignancy. We have acknowledged this issue in the limitations section of Discussion.

5. Small subgroup sizes reduce statistical power and reliability.

Response: We agree with Reviewer 2 on this point. Even with 559 cases, our analyses are limited by smaller sample size when the population was broken down into smaller subgroups. We added acknowledgement in the limitations section of Discussion.

6. Biopsy may reflect earlier diagnosis rather than a direct survival benefit.

Response: We agree with Reviewer 2 that biopsy reflects an earlier diagnosis but we would not discount that biopsy per se does not have a direct survival benefit. Malignant lymphoma like NL would be invariably fatal if not treated. Therefore, timely diagnosis has a direct survival benefit and this is exactly the point of our work here. Importantly, when adjusted for covariates, biopsy was associated with decreased hazard ratio as shown in Fig. 5E.

7. Missing data reporting requires greater transparency.

Response: We thank Reviewer 2 for pointing this out. In Methods, we stated the missing data in various domains of the database.

8. Imbalance between primary and secondary NL cases may introduce confounding bias.

Response: We acknowledge the concern from Reviewer 2 on the imbalance between primary (n=92) and secondary (n=225) NL. Although the survival difference of 5 months between the two is statistically significant (15 vs. 10 months from Xu E, et al. BJC Rep. 2024;2:34), it is probably not that different clinically, considering that lymphomas have a high rate of response to treatment and patients may live for years and perhaps a decade if placed in remission.

9. Evidence supporting PET + MRI + EMG-NCS as the optimal strategy remains limited.

Response: We appreciate Reviewer 2 for pointing this out. Unfortunately, NL is not a common malignancy and its rarity precludes prospective data collection and retrospective analysis from the literature is the only means of comprehensive analysis for this disorder. This is pointed out in the limitations section of Discussion.

10. Sensitivity analyses are needed to confirm the robustness of findings.

Response: We respectfully disagree with Reviewer 2 because performing a sensitivity analysis will incorporate a lot of assumptions. The way to go forward is to find a validation dataset to confirm biopsy and/or PET + MRI + EMG-NCS make a survival difference. We are stating that in the limitation section of Discussion.

Minor Comments:

1. Figure quality is suboptimal; several boxplots and Kaplan–Meier curves are difficult to interpret and require improved resolution.

Response: We understood Reviewer 2’s concern and we have made some improvement. We hope that this is satisfactory.

2. Figure 6 legend appears confusing:

“Five times more secondary NL patients had a biopsy compared to primary NL”

The wording should be revised for clarity.

Response: We thank Reviewer 2 for this comment. The wording has been revised.  We changed the wording to “Five times more secondary NL patients (n=134) did not have a biopsy compared to those with primary NL (n=29). Among those with biopsy, this difference is only 1.4 times between secondary (n=124) and primary (n=94) NL patients.”

3. The manuscript contains multiple grammatical issues and typographical inconsistencies (e.g., “beĴer”, “seĴing”), suggesting formatting or encoding problems requiring correction.

Response: They are corrected.

4. The term “treatment 1” should be defined more clearly at first use.

Response: The statement in Introduction on “treatment 1” is removed, while treatment 1 is defined in Methods in line 73. In the abstract, “treatment 1” is now replaced by ‘first treatment.”

5. Abbreviations should be consistently defined upon first mention (e.g., EMG-NCS).

Response: Thank you. The abbreviations are now first defined in Methods in lines 73 to 75. 

6. Supplementary Figure numbering appears duplicated (“Figure S2” repeated). This should be corrected.

Response: Yes, and thank you. This is corrected.

7. The discussion comparing NL with intravascular lymphomatosis is interesting but somewhat speculative and could be shortened.

Response: The authors respectfully would like to keep our discussion on intravascular lymphomatosis. We believe that it is important to give another perspective on another nervous system manifestation of lymphoma and the challenges in achieving timely diagnosis.

8. References are very few and very old. Author recommended must be cite recent citation.

Response: We thank reviewer 2 for this point. We incorporated Table S16 as supplementary data on the publications that were used in this study. The more recent larger scale studies, references 4 (year 2010), 5 (year 2021) and 13 (year 2019), are included in the references.

9. Keywords must be different from the title of the manuscript.

Response: We thank Reviewer 2 for this point. The keywords have been changed to: “Lymphoma; Peripheral Nerves; Imaging; Biopsy.”  None of these words are in the title of the manuscript.

Reviewer 3 Report

Comments and Suggestions for Authors

• The association between biopsy and improved survival appears to be substantially affected by selection bias and confounding factors, limiting causal interpretation. Patients who did not undergo biopsy were disproportionately affected by secondary neurolymphomatosis and, according to the authors' own discussion, often had pre-existing lymphoma diagnoses, inaccessible lesions, or clinical conditions that made biopsy unsafe. These characteristics are independently associated with prognosis and may explain the observed differences in survival. Furthermore, the lack of significant differences in the time from symptom onset to diagnosis and treatment outcomes between the biopsied and non-biopsied groups weakens the proposed mechanism by which biopsy could improve survival. Therefore, the authors should temper their conclusions and emphasize that biopsy is associated with, rather than responsible for, improved outcomes. To strengthen the analysis, additional multivariable adjustments, stratified analyses according to primary versus secondary neurolymphomatosis, sensitivity analyses excluding high-risk non-biopsied patients, and propensity score-based approaches should be considered to better account for baseline differences between groups.

 

• A significant concern is the apparent contradiction between these findings. Although biopsy was associated with longer survival, it did not result in an earlier diagnosis or improved treatment outcomes. Therefore, the mechanism underlying the observed survival advantage remains unknown. A more plausible explanation may be selection bias, whereby patients eligible for biopsy had more favorable baseline characteristics and prognoses. The authors should discuss this possibility in greater depth and avoid implying a causal relationship between biopsies and survival.

 

• Another concern is the apparent inconsistency between the study conclusions and clinical recommendations. While the manuscript emphasizes biopsy as a key factor associated with improved survival, the authors also recommend empirical systemic chemotherapy for clinically unstable patients or those with lesions inaccessible to biopsy. This suggests that treatment decisions can be made without histological confirmation in selected cases, thereby weakening the argument that biopsy is the critical determinant of outcome. The authors should reconcile these statements and clarify whether biopsy is a causal contributor to improved survival or simply a marker of patients with favorable clinical characteristics.

 

• Several methodological limitations that could influence the interpretation of the results are insufficiently discussed in this study. First, the possibility of immortal time bias should be considered, as patients who eventually undergo biopsy must survive long enough to receive the procedure, potentially inflating the survival estimates. Second, the definition of disease progression is not clearly standardized across the heterogeneous case reports included in the analysis, which may affect the reliability of the progression-related endpoints. Third, missing data were not imputed, and the effective sample size for each analysis was not always evident, raising concerns about potential bias from non-random missingness. Finally, the study did not adequately address lead-time bias, whereby an earlier diagnosis may artificially prolong measured survival without necessarily improving the natural course of the disease. These limitations should be acknowledged and discussed in more detail.

Author Response

Reviewer 3:

The association between biopsy and improved survival appears to be substantially affected by selection bias and confounding factors, limiting causal interpretation. Patients who did not undergo biopsy were disproportionately affected by secondary neurolymphomatosis and, according to the authors' own discussion, often had pre-existing lymphoma diagnoses, inaccessible lesions, or clinical conditions that made biopsy unsafe. These characteristics are independently associated with prognosis and may explain the observed differences in survival. Furthermore, the lack of significant differences in the time from symptom onset to diagnosis and treatment outcomes between the biopsied and non-biopsied groups weakens the proposed mechanism by which biopsy could improve survival. Therefore, the authors should temper their conclusions and emphasize that biopsy is associated with, rather than responsible for, improved outcomes. To strengthen the analysis, additional multivariable adjustments, stratified analyses according to primary versus secondary neurolymphomatosis, sensitivity analyses excluding high-risk non-biopsied patients, and propensity score-based approaches should be considered to better account for baseline differences between groups.

Response: We thank Reviewer 3 for his/her comment. We take heart to the comments on selection bias, confounding factors and limited causal interpretation. There are multiple points that Reviewer 3 raised and we are addressing them one by one:

1. The issue on patients who did not undergo biopsy were disproportionately affected by secondary neurolymphomatosis should be addressed by Cox proportional hazard analysis when survival was adjusted for primary and secondary NL in Fig. 5E and 5F.

2. In our Kruskal-Wallis analyses, biopsy appears to be associated with a longer survival time from treatment 1 to progression, from diagnosis and from symptom onset (non-mutually exclusive in Fig. 3 and mutually exclusive in Fig. 4). Reviewer 3 correctly pointed out that there is no difference in the Kaplan-Meier survival between the biopsied and non-biopsied groups from treatment 1 to progression in Fig. 5. But in our prior publication (Xu et al. BJC Rep. 2024;2:34), there was a difference in time from treatment 1 to progression when stratified by age in our prior publication. Therefore, our Cox proportional hazard adjusted for primary vs. secondary NL and age should account for this difference.

3. We agree with Reviewer 3 to temper our conclusion and emphasize that biopsy is associated with, rather than responsible for, improved outcomes. We also performed a Cox hazard model for primary NL and secondary NL, and we found that biopsy was associated with a significant survival benefit for secondary but not primary NL. This is probably due to a balanced number of secondary NL patients without (n=134) and with biopsy (n=124) as shown in Fig. 6A. But primary NL patients without (n=26) and with (n=91) biopsy are not balanced. This is addressed in Discussion.

A significant concern is the apparent contradiction between these findings. Although biopsy was associated with longer survival, it did not result in an earlier diagnosis or improved treatment outcomes. Therefore, the mechanism underlying the observed survival advantage remains unknown. A more plausible explanation may be selection bias, whereby patients eligible for biopsy had more favorable baseline characteristics and prognoses. The authors should discuss this possibility in greater depth and avoid implying a causal relationship between biopsies and survival.

Response: We thank Reviewer 3 for this comment. We think that the reason behind biopsy was associated with longer survival but did not result in an earlier diagnosis or improved treatment outcomes may be due to the fact that more patients with secondary NL did not undergo biopsy before treatment and these patients may have more advanced lymphomas. Our prior publication found that primary NL had a better prognosis than secondary NL (Xu et al. BJC Rep. 2024;2:34). We therefore incorporated primary and secondary NL as an adjustment in our multivariate Cox proportional hazard model.

Another concern is the apparent inconsistency between the study conclusions and clinical recommendations. While the manuscript emphasizes biopsy as a key factor associated with improved survival, the authors also recommend empirical systemic chemotherapy for clinically unstable patients or those with lesions inaccessible to biopsy. This suggests that treatment decisions can be made without histological confirmation in selected cases, thereby weakening the argument that biopsy is the critical determinant of outcome. The authors should reconcile these statements and clarify whether biopsy is a causal contributor to improved survival or simply a marker of patients with favorable clinical characteristics.

Response: We thank Reviewer 3 for this valuable comment in our Discussion. We agree that it is confusing and therefore we removed our comments on empirical systemic chemotherapy in the fifth paragraph of Discussion.

Several methodological limitations that could influence the interpretation of the results are insufficiently discussed in this study. First, the possibility of immortal time bias should be considered, as patients who eventually undergo biopsy must survive long enough to receive the procedure, potentially inflating the survival estimates. Second, the definition of disease progression is not clearly standardized across the heterogeneous case reports included in the analysis, which may affect the reliability of the progression-related endpoints. Third, missing data were not imputed, and the effective sample size for each analysis was not always evident, raising concerns about potential bias from non-random missingness. Finally, the study did not adequately address lead-time bias, whereby an earlier diagnosis may artificially prolong measured survival without necessarily improving the natural course of the disease. These limitations should be acknowledged and discussed in more detail.

Response: We will address Reviewer 3’s comments one by one below:

1. We do not think immortal time bias is a major issue in our population. This is because malignant lymphoma is a relentless disease and patients die quickly without treatment.

2. We agree with Reviewer 3 that disease progression in NL is not clearly standardized across the heterogeneous case reports included in the analysis, and therefore this issue may affect the reliability of the progression-related endpoint. This is an inherent problem with this rare disease. However, our approach is the only means of gaining a global view for this rare malignancy. This limitation is addressed in Discussion.

3. We did not impute missing data because (i) the reported outcome of NL patients are heterogeneous in the literature, and (ii) we will be making too many assumptions by imputation. Accounting for reporting heterogeneity and imputing assumptions will become a subjective exercise.

4. We respectfully disagree with Reviewer 3 on this point. The major problem with this NL malignancy from our previous work (Xu et al. BJC Rep. 2024;2:34) is that timely diagnosis remains a challenge despite advances in diagnostic modalities and lymphoma treatments. Since lymphomas in general is very responsive to treatment, we believe that the difference in survival is huge between diagnosed and undiagnosed patients. Therefore, whatever lead time bias caused by earlier diagnosis will be small compared to the survival derived from treatment against the lymphoma.

Reviewer 4 Report

Comments and Suggestions for Authors

It should be interesting to see histological and radiological images of this disease.

Author Response

Reviewer 4:

It should be interesting to see histological and radiological images of this disease.

Response: We thank Reviewer 4’s comments. We respectfully disagree that we should show histological and radiological images. Our work is an aggregate analysis of published literature, not case series. Therefore, we regrettably will not include these images.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Author improved manuscript substantially, present form of manuscript is recommended for the publication.

Reviewer 3 Report

Comments and Suggestions for Authors

Accept in present form