Phosphodiesterase-5 inhibitors use and risk for mortality and metastases among male patients with colorectal cancer

Phosphodiesterase-5 (PDE5) inhibitors are suggested to have anti-tumor effects and to inhibit surgery-induced immunosuppression. We aimed to explore whether post-diagnostic use of PDE5 inhibitors was associated with a better prognosis among male patients with colorectal cancer (CRC) and the role of open surgery in the association. Here we show that post-diagnostic use of PDE5 inhibitors is associated with a decreased risk of CRC-specific mortality (adjusted HR = 0.82, 95% CI 0.67-0.99) as well as a decreased risk of metastasis (adjusted HR = 0.85, 95% CI 0.74-0.98). Specifically, post-operative use of PDE5 inhibitors has a strong anti-cancer effect. The reduced risk of metastasis is mainly due to distant metastasis but not regional lymphatic metastasis. PDE5 inhibitors have the potential to be an adjuvant drug for patients with CRC to improve prognosis, especially those who have undergone open surgery.

6. Was date of dispensing or date of prescription used? The Swedish prescribed drug register only contains prescription and dispensing dates for those drugs that were actually dispensed. So, if prescription date was used, then there are likely some missing prescriptions i.e. people that were prescribed but did not dispense their PDE5 inhibitors. This should be discussed.
7. When defining exposure by pre-and post-operative use, only those patients whose last PDE5 inhibitor prescription was before the surgery were defined as being exposed prior to surgery. However, patients that were prescribed both before and after surgery were defined as postoperative users. Please can you explain the motivation behind this? Because the post-operative group includes a mix of individuals who were both pre-and post-operative users, and those who were solely post-operative users.
8. Why is does follow up for death end at March 2017, but follow up for metastases end at December 2015?
9. What happened to men that died of other causes during follow up? These deaths by other causes (such as CVD) could be competing events. Especially in the cases where PDE5 inhibitors were prescribed for hypertension.
10. The first line of the statistical analysis states that time-dependent Cox regression was used to reduce immortal time bias. This is not possible. The only way to stop immortal time bias is through carefully ensuring it does not exist in the study design. Time-updating exposure will not immediately stop immortal time bias. 11. As touched on above in point 7, authors time-updated the exposure. However, it is not clear how confounders were defined. Were they only adjusted for at baseline (time of colorectal cancer diagnosis)? Please make this clear. If time-updating exposure, it is also important to think if there is any treatment-confounder feedback i.e. do time-varying confounders change the likelihood of changing exposure? If so, then merely time-updating confounder is not enough. More advanced methods such as inverse probability weighting or the g-formula need to be used.
12. On page 13, line 306, authors state that sensitivity analyses were performed to control potential biases. It is not possible to control for biases through sensitivity analyses. You can only explore if potential biases will impact our results.
14. What was the average follow-up in those exposed and not exposed to PDE5 inhibitors? Hazard ratios give the average difference between the hazards of the two groups over all of follow up. If the follow-up differed between the two groups, then the difference in outcome could be a consequence of this difference. Providing simple Kaplan-Meier plots would also help to explore this.
Reviewer #2 (Remarks to the Author): The authors examined the associations between post-diagnostic use of phosphodiesterase-5 (PDE5) inhibitors and CRC-specific mortality as well as metastasis among CRC patients from the Swedish Cancer Registry. The authors reported that post-diagnostic use of PDE5 inhibitors was associated with a decreased risk of CRC-specific mortality and metastasis. Some data presentations and analyses need to be shaped up substantially as suggested below.
1. Since the CRC patients and drug information are from two big pools, the Swedish Cancer Registry and the Swedish Prescribed Drug Register, and also the authors applied several inclusion and exclusion criteria to select the patients for final analysis (also sensitivity analyses), a clean flow-chart describing the process of the patients' selection (including the numbers occurred in each inclusion and exclusion criteria) would be helpful to the readers to better understand the study design. 4. Considering that clinical stage at diagnosis is the most important factor that affects the CRC prognosis, drug use may have an inconsistent effect on the prognosis and metastasis for each stage of cancer. Did the authors get a chance to look at the associations stratified by the stage? Also, the authors excluded patients with CRC at stage IV for metastasis-related analysis; does the drug use affect the mortality for the patients with CRC at stage IV? 5. In the metastasis-related analyses, a sensitivity analysis after excluding the patients who died during follow-up would be worth to try in order to further examine the independent association with drug use.
6. English proofreading would be helpful to improve the quality of the manuscript.

Reviewer #1:
Huang et al. use Swedish register data to show that PDE5 inhibitors (otherwise known as Viagra) reduce the risk of colorectal cancer specific mortality and metastases in men diagnosed with colorectal cancer. This is a question that had not yet been addressed, and Swedish registers present a unique opportunity to provide an answer. I, however, have several concerns regarding the methodology and analysis: 1. The first results reported are that of the association between PDE5A expression and survival in patients with colorectal cancer. However, there is no extensive description of the methods used for this analysis. All that is mentioned in the methods is that the COADREAD cohort was used. The results then state that this only includes data from 430 patients. Who are these patients? Were they all also included in the main analysis? You show a Kaplan Meier plot, but don't explain when follow up started. Was it directly after colorectal cancer diagnosis? To understand these results, the reader needs further description of the methods. >>> We apologized for the unclear description. We have added detail information related to the study cohort, as well as the statistical analyses in the method section #Page 14. Patients in this analysis were retrieved from public data resources TCGA, who were not included in the main analyses using Swedish registers.  >>> Thank the reviewer for pointing out this. We agree that the National quality register for colon and rectal cancer will add more information about severity of colorectal cancer and cancer treatments. Unfortunately, we did not get the permission to use it even we have tried to contact the owner of this register before.
4. The authors state that patients were required to have at least two PDE5 inhibitor prescriptions after colorectal cancer diagnosis before exposure. Did exposure to PDE5 inhibitors then start at the date of second prescription, or was it predated to first prescription once they received their second prescription? This is not clear. >>> We apologized for being unclear of the definition. To overcome non-adherence of PDE5 inhibitors use, we identified patients with two or more dispensations of PDE5 inhibitors after diagnosis defined as users. Exposure to PDE5 inhibitors was defined as the date of first dispensation.
5. Was there any restriction on the time allowed between the two prescriptions? For example, could a patient have one PDE5 inhibitor prescription directly after colorectal cancer diagnosis, then another 18 months later and still be defined as being exposed? >>> There is no restriction on the time interval between two prescriptions. The median value of time interval between first and second prescription in this study was 4 months.
6. Was date of dispensing or date of prescription used? The Swedish prescribed drug register only contains prescription and dispensing dates for those drugs that were actually dispensed.
So, if prescription date was used, then there are likely some missing prescriptions i.e. people that were prescribed but did not dispense their PDE5 inhibitors. This should be discussed.
>>> Thank the reviewer pointing out this important question. Date of dispensing was used in the analysis, and we revised the description in the manuscript. 7. When defining exposure by pre-and post-operative use, only those patients whose last PDE5 inhibitor prescription was before the surgery were defined as being exposed prior to surgery. However, patients that were prescribed both before and after surgery were defined as post-operative users. Please can you explain the motivation behind this? Because the postoperative group includes a mix of individuals who were both pre-and post-operative users, and those who were solely post-operative users.  Table 4.  Table 4. >>> Thank the reviewer for this suggestion. In our original definition, we defined confounding factors as status before the end of follow up. We agree that it could be better to define confounders as status at baseline (time of colorectal cancer diagnosis). We re-defined these confounders in the revised version. All results related to multivariate model have been re-analyzed and the description has been revised in method section. Regarding for the potential treatment-confounder feedback caused by the time-updating exposure, we added a sensitivity analysis (Sensitivity analysis 3) in matched cohort design, where the treatmentconfounder feedback was not supposed to exist.
12. On page 13, line 306, authors state that sensitivity analyses were performed to control potential biases. It is not possible to control for biases through sensitivity analyses. You can only explore if potential biases will impact our results.
>>> We totally agree with the reviewer and revised the sentence accordingly.
13. There is no need to include p-values in Table 1. >>> It has been deleted.
14. What was the average follow-up in those exposed and not exposed to PDE5 inhibitors?
Hazard ratios give the average difference between the hazards of the two groups over all of follow up. If the follow-up differed between the two groups, then the difference in outcome could be a consequence of this difference. Providing simple Kaplan-Meier plots would also help to explore this.
>>> The median follow up in users and non-users have been described in the results #Page 5, and Kaplan-Meier plots is present in Figure 2.

Reviewer #2:
The authors examined the associations between post-diagnostic use of phosphodiesterase-5 (PDE5) inhibitors and CRC-specific mortality as well as metastasis among CRC patients from the Swedish Cancer Registry. The authors reported that post-diagnostic use of PDE5 inhibitors was associated with a decreased risk of CRC-specific mortality and metastasis.
Some data presentations and analyses need to be shaped up substantially as suggested below. 1. Since the CRC patients and drug information are from two big pools, the Swedish Cancer Registry and the Swedish Prescribed Drug Register, and also the authors applied several inclusion and exclusion criteria to select the patients for final analysis (also sensitivity analyses), a clean flow-chart describing the process of the patients' selection (including the numbers occurred in each inclusion and exclusion criteria) would be helpful to the readers to better understand the study design. >>> Thank you for the suggestion. We added a flowchart as shown in Supplementary Figure 1. 2. For the evaluation of the use of PDE5 inhibitors, the authors included three inhibitors (sildenafil, tadalafil, and vardenafil). Actually, there are many other types of PDE5 inhibitors (e.g., Thiosildenafil, rac Xanthoanthrafil… etc.). Can the authors validate the reason that why they only did check three types (except the reason that these are commonly used drugs in Sweden). Some appropriate references might be helpful.
>>> We agree with the reviewer that there are many other types of PDE5 inhibitors.
However, only sildenafil, tadalafil, and vardenafil were used in Sweden and can be retrieved from the Swedish Prescribed Drug Register.
3. In addition to the frequency of drug use, the dose and duration of drug use are also important indices in terms of evaluation for the relationship between drug use and disease outcomes. However the duration of PDE5 inhibitors use has not been discussed (due to lack of information?). Also, the analysis and description regarding dose of drug use are not clearly described.
>>> We agree with the reviewer that the dose and duration of drug use are also important indices. We added the description about cumulative dose of drug use in method section #Page 13. Regarding for the duration of drug use, unlike some drugs, such as anti-hypertension which should be used continuously, the use of PDE5 inhibitors was randomly (which is usually taken only when needed, 30 minutes to 1 hour before sexual activity), so we could not determine the duration based on the Swedish Prescription Drug Register. 4. Considering that clinical stage at diagnosis is the most important factor that affects the CRC prognosis, drug use may have an inconsistent effect on the prognosis and metastasis for each stage of cancer. Did the authors get a chance to look at the associations stratified by the stage? Also, the authors excluded patients with CRC at stage IV for metastasis-related analysis; does the drug use affect the mortality for the patients with CRC at stage IV? >>>We agree that clinical stage at diagnosis is the most important factor, thus a stratified analysis by earlier or advance stage was added to explore the difference of associations between PDE5 inhibitors use and CRC-specific death. As patients with stage IV have already experienced metastasis, we thus excluded patients diagnosed at stage IV in the main analysis.
However, we added a sensitivity analysis (sensitivity analysis 5) in Table 4 in which patients with stage IV and unknown stage were included. The protective effect of PDE5 inhibitors remained significant.
5. In the metastasis-related analyses, a sensitivity analysis after excluding the patients who died during follow-up would be worth to try in order to further examine the independent association with drug use. >>> Death has the potential to compete the risk of metastasis, we thus add a sensitivity analysis (sensitivity analysis 4) using competing risk model. The results were shown in Table 4. 6. English proofreading would be helpful to improve the quality of the manuscript. >>> Our scientific editor has gone through the paper.