Alternate Pembrolizumab Dosing Interval in Advanced NSCLC with PD-L1 TPS ≥ 50%: 3 Weekly Compared to 6 Weekly Dosing

Round 1

Reviewer 1 Report (Previous Reviewer 1)

Interessing and relevant topic

Major comments

Abstract : conclusion should be less definitive. The study had several limits (ie retrospective, small sample of patients in W6 group,…) and I don’t think that we can stated that “This study validates the use of Q6W pembrolizumab dosing….” No significant difference don’t mean equivalence…..

Can the authors confirm that there is no fixe dose of Pembrolizumab in any groups?

It’s not clear how the assignment to Q3W or Q6W cohorts was based? The authors stated “ on the schedule used for the majority (greater than 50%) of the patient’s treatment” can the authors precise it?

What is the median (95%CI) follow up of the 2 groups?

Table 2: usually if no significant in univariate analysis, factors are not tested in multivariate analysis. Can the authors correct or explain why they test non-significant factors?

Duration of pembrolizumab was tested but results were not reported

In case control analysis, only 37/41 W6 group why? Data’s missing?

Why don’t the authors chose to do a 2/1 case control model?

Minor comments

Probability of before “survival” is missed

The number of events below the figure is missed

Author Response

Major comments

1. Abstract: conclusion should be less definitive. The study had several limits (ie retrospective, small sample of patients in W6 group,…) and I don’t think that we can stated that “This study validates the use of Q6W pembrolizumab dosing….” No significant difference don’t mean equivalence…..

I agree with the suggestion and have changed the wording to “support” instead of “validate”.

2. Can the authors confirm that there is no fixe dose of Pembrolizumab in any groups?

Yes, per institution standard at BC Cancer all patients were treated with weight-based dosing. This was noted in the manuscript by including “All patients were treated with weight-based dosing of pembrolizumab…”

3. It’s not clear how the assignment to Q3W or Q6W cohorts was based? The authors stated “ on the schedule used for the majority (greater than 50%) of the patient’s treatment” can the authors precise it?

The decision for dosing schedule was based on the treating oncologists’ recommendation. Some patients would have been started on Q3W dosing given that the beginning of the interval for data collection was prior to institution change of adjusting to include Q6W dosing option. For patients that had treatment on both intervals they were classified as Q3W or Q6W in our dataset based on the dosing interval that they received for the majority of their treatment. We included the following in the methods “The dosing interval prescribed for patients was at the discretion of the treating medical oncologist”.

4. What is the median (95%CI) follow up of the 2 groups?

The following was added to the manuscript “The median duration of follow up was Q3W 16.0 months [interquartile range (IQR) 6.5-29.1] and Q6W 17.2 months (IQR 9.1-26.3).”

5. Table 2: usually if no significant in univariate analysis, factors are not tested in multivariate analysis. Can the authors correct or explain why they test non-significant factors?

Non-significant variables in univariate analysis were included in the multivariate analysis because it is recognized that age and sex are strong prognostic variables. The authors felt that it was reasonable to include them in the MVA because of the prognostic implications.

The MVA was updated per reviewers recommendation to only include the PS and the variable of interest, dosing schedule and there was no new signal.

6. Duration of pembrolizumab was tested but results were not reported

The following was included in the manuscript “The median duration of therapy was Q3W 4.2 months [interquartile range (IQR) 1.4-12.6] and Q6W 6.2 months (IQR 1.7-20.6).”

7. In case control analysis, only 37/41 W6 group why? Data’s missing?

The case control analysis matched on key variables based on the smaller Q6W cohort as outlined. Some of the patients in the Q6W cohort were unable to be matched to the Q3W because of the 4 specific parameters. As a result, the case cohort match was slightly smaller then the full data set.

8. Why don’t the authors chose to do a 2/1 case control model?

We updated the case match analysis to a 2:1 model. The results do not change but we have revised the paper accordingly.

Minor comments:

9. Probability of before “survival” is missed

We recognize that outlining the expected survival of patients with stage IV NSCLC and PD-L1 TPS >50% is important to comment on prior to improvement with immunotherapy. In background text the description of KEYNOTE-024 findings the OS benefit is outlined that pembrolizumab improved OS from 13.4 months (with chemo alone) to 26.3 months

10. The number of events below the figure is missed

This data has been added.

Reviewer 2 Report (New Reviewer)

To investigate whether the overall survival differs between dose level Q3W vs Q6W of pembrolizumab, authors applied statistical models on the data from BC Cancer patients with stage IV NSCLC and PDL1 ³50% treated with pembrolizumab. With the analysis results, the authors concluded that there was no OS difference demonstrated with pembrolizumab dosing Q3W compared to Q6W in a multivariate analysis that controlled covariates.  I read this paper with great interest.  I agree that the motivating research question is relevant and essential, but I have a few comments and questions, that I hope will help improve the manuscript, about the methodology employed in this paper.

1.     On page 2, “OS was defined as the Stage IV NSCLC date of diagnosis to date of death”.  Usually, the survival analysis time origin for such kind of analysis starts from the date of the first dose (initial treatment).  Can you explain why do you choose the diagnosis date?  

2.     Your sensitivity analysis includes a case-matched analysis with the covariates.  How did you match the case to the control? In other words, what kind of algorithms have you used? And why use such algorithm? I would like to see some discussion in the manuscript.

3.     Based on my experience, clinical trial data have missing date issues. For example, the patients may loss of follow-up or take prohibited/rescue medication.  How did you handle the missing data when you perform the analysis? Loss of follow-up sometimes is not that critical; it is more like missing at random. But if the patient took prohibited/rescue medication, the efficacy could be impacted, which could bias the analysis results. 

4.     “There was no difference in age, sex, histology or smoking status between the Q3W and Q6W groups (Table 1).”  It is better to express in this way, “No statistical difference was detected…”.  The reason is, based on your dataset, the statistical analyses, e.g. t-test or chi-square test, could not reject the null hypothesis (no difference). But it does not mean it is true.  

5.     In addition to the log-rank test, I would suggest reporting more metrics, like Restricted Mean Survival Time (RMST) for the KM analysis.  Because in Figure 1, it seems the non-proportional hazard issue exists. Log-rank test in this case is not the best choice. 

Author Response

1.     On page 2, “OS was defined as the Stage IV NSCLC date of diagnosis to date of death”.  Usually, the survival analysis time origin for such kind of analysis starts from the date of the first dose (initial treatment).  Can you explain why do you choose the diagnosis date?  

In the initial iteration of this research we had intended to compare BSC vs first line pembrolizumab and had chosen metastatic diagnosis date for comparison purposes. The dosing question became of greater interested and so we focused on that group for this manuscript. We agree that using the date of first treatment helps eliminate lead time bias and the results have been updated accordingly.

2.     Your sensitivity analysis includes a case-matched analysis with the covariates.  How did you match the case to the control? In other words, what kind of algorithms have you used? And why use such algorithm? I would like to see some discussion in the manuscript.

The methods were edited to include more details as requested “A case control matching was conducted with an exact match for sex (male or female), tolerance for age was +/- 5 years, exact match PS (0-1 and 2 or greater) and tolerance for duration of pembrolizumab therapy +/- 2 months for Q3W matched to Q6W. Sampling without replacement was performed with the matching to be conducted to maximize execution performance.”

3.     Based on my experience, clinical trial data have missing date issues. For example, the patients may loss of follow-up or take prohibited/rescue medication.  How did you handle the missing data when you perform the analysis? Loss of follow-up sometimes is not that critical; it is more like missing at random. But if the patient took prohibited/rescue medication, the efficacy could be impacted, which could bias the analysis results. 

For variables for which we were unable to collect data we included the category “unknown” in the tables to acknowledge missing variables. We did not collect data on medications and have included this information into the limitation of our analysis. For patients who were lost to follow up they were censored at the time of last follow up, lab test or investigation that confirmed that they were still alive at that time point.

4.     “There was no difference in age, sex, histology or smoking status between the Q3W and Q6W groups (Table 1).”  It is better to express in this way, “No statistical difference was detected…”.  The reason is, based on your dataset, the statistical analyses, e.g. t-test or chi-square test, could not reject the null hypothesis (no difference). But it does not mean it is true.  

Agree with this suggestion, I have made this change in the manuscript.

5.     In addition to the log-rank test, I would suggest reporting more metrics, like Restricted Mean Survival Time (RMST) for the KM analysis.  Because in Figure 1, it seems the non-proportional hazard issue exists. Log-rank test in this case is not the best choice. 

Thank for this comment. The following was added to the manuscript

The mean survival time was Q3W 21.0 (95% CI 19.7-22.8) and Q6W 20.6 months (95% CI 16.1-25.1).

Round 2

Reviewer 1 Report (Previous Reviewer 1)

ok for me

Reviewer 2 Report (New Reviewer)

Thanks for the revision.  no further comment from me.  Good luck!

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Interessing and relevant topic

Major comments

Abstract : conclusion should be less definitive. The study had several limits (ie retrospective, small sample of patients in W6 group,…) and I don’t think that we can stated that “This study validates the use of Q6W pembrolizumab dosing….” No significant difference don’t mean equivalence…..

Can the authors confirm that there is no fixe dose of Pembrolizumab in any groups?

It’s not clear how the assignment to Q3W or Q6W cohorts was based? The authors stated “ on the schedule used for the majority (greater than 50%) of the patient’s treatment” can the authors precise it?

What is the median (95%CI) follow up of the 2 groups?

Table 2: usually if no significant in univariate analysis, factors are not tested in multivariate analysis. Can the authors correct or explain why they test non-significant factors?

Duration of pembrolizumab was tested but results were not reported

In case control analysis, only 37/41 W6 group why? Data’s missing?

Why don’t the authors chose to do a 2/1 case control model?

Minor comments

Probability of before “survival” is missed

The number of events below the figure is missed