Wait Times and Survival in Lung Cancer Patients across the Province of Quebec, Canada

Round 1

Reviewer 1 Report

This manuscript entitled “Wait Times and Survival in Lung Cancer Patients across the Province of Quebec, Canada” by Marie-Hélène Denault et al. comprehensive to evaluate the impact of wait times on survivals in lung cancer patients. The authors have largely improved the quality of this manuscript. I have some comments for this manuscript.

1. Although the authors updated the data of calculable patients, I don’t consider this is the adequate resolution. The patients with losing follow-up still can be counted and they should be censored when do survival analysis. I agree with authors’ reply that the follow-up time of 10% patients is too close to the diagnosis. The authors should mention this point in the discussion or limitation.

 

2. Patients who got SBRT were included in the “definitive radiation” category, and those patients have early-stage disease. This can explain why patients in the “definitive radiation” category did better than those in the “definitive chemoradiation” category, combined treatment being the treatment of choice for unresectable stage III disease. This should be addressed in the discussion.

 

3. It is very interesting about the negative intervals for certain condition. Could authors address this in somewhere of manuscript.
4. It is unusual to test for biomarkers more than once. In those rare occasions, the interval would represent the time between pathology result and the first biomarkers result. In this manuscript, the biomarkers are EGFR, ALK, PDL1 which should be detected by different methods. Therefore, I don’t consider you can have the results at the same time. In addition, the first biomarker result may be not enough to determine the treatment. Usually, IHC with ALK and PD-L1 results can be obtained earlier and EGFR testing takes longer time. The treatment can not be suggested if the EGFR result is unknown.

 

 

Author Response

Please see attachment.

Author Response File: Author Response.docx

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

This manuscript entitled “Wait Times and Survival in Lung Cancer Patients across the Province of Quebec, Canada” by Marie-Hélène Denault et al. comprehensive to evaluate the impact of wait times on survivals in lung cancer patients.  I have some comments for this manuscript.

1. Overall survival was calculable for 901 of 1309 patients (69 %). I have no idea why the survival of 31% patients was not calculable?
2. Table 3. why is the case number for each wait time different ?

 

3. Hospital network was significant in univariate analyses, but did not remain significant when the other variables were added in.

 

4. Definitive radiation seems be better than definitive chemoradiation for survival in table 4. Why ?

 

 

5. The treatment is based on stage so why is the correlation between treatment and stage ?

 

6. The main limitation comes from losses to follow-up and missing data, which cost us 31% of our sample for the survival analysis. Usually, they are censored rather than deleted. If the authors exclude such data would result in selection bias and possible overestimated the survivals.

 

7. How could the diagnosis to treatment intervals and/or supportive treatment be negative ? If the data is not reliable, how can authors include them in other analysis ?
8. It is better that the authors can provide an figure which illustrate all the time interval they investigated.

 

9. Pathology result to biomarkers result. If the patients had several times of biomarker testing, how did the authors count?

 

 

10. The authors can not clearly define the patients enrolled in different interval so the case number for different intervals are not consistent. For example,

1st appointment with specialist to surgery (n=304) and Diagnosis to surgery (n=180) are different ?

Radiation referral to first radiation treatment (n=188) and Diagnosis to definitive radiation (n=92) plus Diagnosis to definitive chemoradiation (n=107) are different.

 

 

11. The shorter wait times observed for advanced NSCLC and SCLC might indicate a tendency from clinicians to act quicker on sicker patients. It is unfair for early NSCLC patients. In addition, different specialists are responsible for early and advanced lung cancer and different procedures should be done before the treatment should be the main reasons.

 

12. The median survival was only 7 months even this study enrolled early stage lung cancer. Why ?

Author Response

This manuscript entitled “Wait Times and Survival in Lung Cancer Patients across the Province of Quebec, Canada” by Marie-Hélène Denault et al. comprehensive to evaluate the impact of wait times on survivals in lung cancer patients.  I have some comments for this manuscript.

1. Overall survival was calculable for 901 of 1309 patients (69 %). I have no idea why the survival of 31% patients was not calculable?

 

We were able to obtain information on survival for 271 more patients, making overall survival now calculable for 1172/1309 (90 %) patients. The remaining 10 % were lost to follow-up.

 

2. Table 3. why is the case number for each wait time different ?

The number of patients for whom both the specific time interval and survival were available was variable for each analysis, explaining the different case numbers.

3. Hospital network was significant in univariate analyses, but did not remain significant when the other variables were added in.

With the updated survival analysis, hospital network is no longer significant in univariate analyses.

4. Definitive radiation seems be better than definitive chemoradiation for survival in table 4. Why ?

Patients who got SBRT were included in the “definitive radiation” category, and those patients have early-stage disease. This can explain why patients in the “definitive radiation” category did better than those in the “definitive chemoradiation” category, combined treatment being the treatment of choice for unresectable stage III disease.

5. The treatment is based on stage so why is the correlation between treatment and stage ?

There was no interaction between treatment and stage in multivariate analyses. Collinearity remains a possibility but was not formally tested.

6. The main limitation comes from losses to follow-up and missing data, which cost us 31% of our sample for the survival analysis. Usually, they are censored rather than deleted. If the authors exclude such data would result in selection bias and possible overestimated the survivals.

As explained in point 1, we have obtained survival data for 90 % of the cohort now, which significantly reduces bias. For the remaining 10 %, the only possible time point for censoring would be the date of first treatment, too close to diagnosis (median of 31 days for this interval in table 2). This would certainly result in an underestimation of survival.

7. How could the diagnosis to treatment intervals and/or supportive treatment be negative ? If the data is not reliable, how can authors include them in other analysis ?

 

Some patients went straight to surgery without a pathological diagnosis, so the date of diagnosis was the date of the pathology report from the surgical specimen, available after surgery. This would result in a negative time interval between diagnosis and treatment.

 

Moreover, some patients needed urgent radiation for symptom relief, and the result of the biopsy was known later on.

 

8. It is better that the authors can provide an figure which illustrate all the time interval they investigated.

A figure was added in the methods section.

9. Pathology result to biomarkers result. If the patients had several times of biomarker testing, how did the authors count?

It is unusual to test for biomarkers more than once. In those rare occasions, the interval would represent the time between pathology result and the first biomarkers result.

10. The authors can not clearly define the patients enrolled in different interval so the case number for different intervals are not consistent. For example,

1st appointment with specialist to surgery (n=304) and Diagnosis to surgery (n=180) are different ?

Radiation referral to first radiation treatment (n=188) and Diagnosis to definitive radiation (n=92) plus Diagnosis to definitive chemoradiation (n=107) are different.

There were missing data making some intervals impossible to calculate for some patients. Concerning the second comment, “radiaion referral to first radiation” interval includes patients from both “definitive radation” and “definitive chemoradiation” categories. “Diagnosis to definitive radiation” and “diagnosis to definitive chemoradiation” are intervals between diagnosis and two different treatment modalities, hence the different numbers of patients.

11. The shorter wait times observed for advanced NSCLC and SCLC might indicate a tendency from clinicians to act quicker on sicker patients. It is unfair for early NSCLC patients. In addition, different specialists are responsible for early and advanced lung cancer and different procedures should be done before the treatment should be the main reasons.

Thank you for your comment. This is a hypothesis that we have formulated based on the results we obtained, to try to explain those results. We agree that different specialists and different procedures can affect wait times, and that patients in a curative situation (early-stage NSCLC, for instance) should be managed in a timely manner (see discussion and conclusion in the main text).

12. The median survival was only 7 months even this study enrolled early stage lung cancer. Why ?

 

There were a lot of losses to follow-up, and the majority of those had early-stage disease, resulting in an underestimation of overall survival. Fortunately, we were able to obtain survival data for supplemental patients and median survival is now 12.9 months, aligning with the latest data from the Canadian Cancer Society.

Reviewer 2 Report

This manuscript describes lung cancer patients from Quebec, their treatment wait times, and survival.  The authors utilize data from four hospital networks between February 1, 2017 through April 30, 2017 and followed patients for 3 years. The authors discuss wait times in relation to target benchmarks.  The topic will be of interest to readers of Current Oncology.  However, several issues should be addressed, especially around data quality and missing data.

Introduction

A little more detail on the efficacy of treatment would be useful.  Which treatments are thought to extend life and among whom?

Materials and Methods

It would be helpful if the authors provided some information about the quality of data available in the Quebec Cancer Registry.  The QCR is not certified by the North American Association of Central Cancer Registries.  Indeed, only 64% of cases identified through registries were found to qualify for inclusion in the study.  It seems the authors have aimed to circumvent some of these issues by conducting chart reviews in four hospital centers, but more information is necessary in order to permit the reader to judge the quality of the results.

The QCR data have also not been deemed fit for survival analysis (https://www.naaccr.org/certified-registries/).  More discussion is warranted to explain why the results should be interpreted as unbiased.

In the discussion section the authors state that 30% of mortality data were missing from the provincial death registry.  This is substantial. A more detailed discussion of the quality of the mortality data (how it is collected and verified) is necessary.

Why was overall survival selected as the outcome instead of cancer-specific survival?  If this is related to poor data quality in the mortality data file, this should be discussed.

Results

Line 122 – What would cause major delays in investigations? A bit more detail on these circumstances is necessary as they themselves are likely related to survival.

Figure 1. The top box of this figure is a bit confusing.  Are the cases ascertained from the Quebec Cancer Registry as described in the Methods section?  If so, that can be stated here.

Table 4. The title and footnote formatting should be edited

Discussion

The conclusions should be tempered in light of the concerns over data quality and the limited data available for analyses due to missingness.

The authors should speak in more detail about the effect the biases could have on the results.  Are point estimates over or under-estimated?  How sensitive are the current results to data issues?

Author Response

This manuscript describes lung cancer patients from Quebec, their treatment wait times, and survival.  The authors utilize data from four hospital networks between February 1, 2017 through April 30, 2017 and followed patients for 3 years. The authors discuss wait times in relation to target benchmarks.  The topic will be of interest to readers of Current Oncology.  However, several issues should be addressed, especially around data quality and missing data.

Introduction

A little more detail on the efficacy of treatment would be useful.  Which treatments are thought to extend life and among whom?

We did not include information about survival according to treatment type, as this is not the focus of this paper and we wanted to keep the introduction straightforward. Moreover, lung cancer survival in registries and databases is reported by stage, regardless of treatment.

 

 

 

Materials and Methods

It would be helpful if the authors provided some information about the quality of data available in the Quebec Cancer Registry.  The QCR is not certified by the North American Association of Central Cancer Registries.  Indeed, only 64% of cases identified through registries were found to qualify for inclusion in the study.  It seems the authors have aimed to circumvent some of these issues by conducting chart reviews in four hospital centers, but more information is necessary in order to permit the reader to judge the quality of the results.

The Quebec Cancer Registry provides reliable data. It does include cases of suspected, not histologically proven cancer, and this is why 121 patients were excluded. Another 101 patients were excluded because they were not residents of Quebec, because the date of diagnosis did not fit our inclusion criterion or because they presented with cancer recurrence. The main reason for exclusion remains excessive delay in investigation/management caused by an external factor such as patient non compliance/refusal, comorbid issues or prior follow-up for a lesion that was thought to be benign (Figure 2). Those reasons refer to our inclusion criteria, not to the quality of the data available in the Registry. Finally, the Registry was used to identify eligible patients, but data were abstracted from patients’ charts in individual hospitals.

The QCR data have also not been deemed fit for survival analysis (https://www.naaccr.org/certified-registries/).  More discussion is warranted to explain why the results should be interpreted as unbiased.

The QCR data have not been deemed fit for survival analysis because they are not up to date for survival and dates of death, which is why we obtained this information through chart review.

In the discussion section the authors state that 30% of mortality data were missing from the provincial death registry.  This is substantial. A more detailed discussion of the quality of the mortality data (how it is collected and verified) is necessary.

We went back into the charts to obtain the missing survival data, which is now available for 90 % of our cohort.

Why was overall survival selected as the outcome instead of cancer-specific survival?  If this is related to poor data quality in the mortality data file, this should be discussed.

Unfortunately, the cause of death was impossible to identify for most cases. This is why we selected overall survival instead of cancer-specific survival.

 

 

Results

Line 122 – What would cause major delays in investigations? A bit more detail on these circumstances is necessary as they themselves are likely related to survival.

As outlined in Figure 2, causes of major delays in investigations were :

-Initial follow-up for a lesion judged benign (n=332)

-Initial refusal or non compliance of patient (n=40)

-Treatment of comorbidity needed (n=85)

 

Figure 1. The top box of this figure is a bit confusing.  Are the cases ascertained from the Quebec Cancer Registry as described in the Methods section?  If so, that can be stated here.

Because of the addition of a figure in the methods section, this figure is now labeled Figure 2. Yes, the cases were identified through the Quebec Cancer Registry. The top box of this figure has been modified for clarity.

Table 4. The title and footnote formatting should be edited

We are not sure what should be edited. Could you be more specific?

Discussion

The conclusions should be tempered in light of the concerns over data quality and the limited data available for analyses due to missingness.

We believe the data quality is adequate. We have commented on the losses to follow-up and the missing data in the discussion. The conclusion has been modified in light of this comment.

The authors should speak in more detail about the effect the biases could have on the results.  Are point estimates over or under-estimated?  How sensitive are the current results to data issues?

Thank you for this comment. With only 10 % of missing data for survival, we do not think there is a risk for significant bias. Concerning the time intervals, it is hard to speculate on the exact effect of missing data. We do not think there is a systematic bias direction for those analyses, as both short and long intervals could be missing. Overall, missing data for both time intervals and survival could have resulted in a type II error, failing to detect a significant relationship between the two. The fact that we excluded negative time intervals and time intervals of patients who got supportive care mitigates potential dilution of the association, however. The point estimates were quite precise, as shown by their narrow confidence intervals (Table 3). A more detailed analysis of potential biases was added to the discussion.

 

 

Reviewer 3 Report

Dear Authors, 

We were a little bit surprised with a low median survival of only 7 months. Do you have any additional comment on that? 

Kind regards. 

Author Response

We were able to obtain information on survival for 271 more patients, making overall survival now calculable for 1172/1309 (90 %) patients. The remaining 10 % were lost to follow-up.