HealingWell
Search Close Search
☰

Large Swedish study on Radiation vs Surgery

Support Forums
>
Prostate Cancer
✚ New Topic ✚ Reply
❬ ❬ Previous Thread |Next Thread ❭ ❭
Posted 7/14/2018 3:20 AM (GMT 0)
Here is a just published large Swedish study on surgery vs radiation. It notes that while there is a difference in mortaility between radiation and surgery in low to intermediate risk PC, the risk of PC death is so tiny in low to intermediate PC, that the differences are irrelevant. For higher risk PC, where risk of PC death is higher, there was no difference between radiation and surgery.

/www.europeanurology.com/article/S0302-2838(17)31031-X/fulltext

Conclusion: In comparison with previous studies, the difference in Pca mortality after RT and RP was much smaller in this study than in previous studies and very small in absolute terms. Thus, the choice between these two treatments should be guided by the risk of side effects and patient preference rather than by the risk of Pca death.
Posted 7/14/2018 5:22 AM (GMT 0)
Thanks for the link!
Posted 7/14/2018 8:42 AM (GMT 0)
good info
Posted 7/14/2018 11:41 AM (GMT 0)
Thanks for the link!

(Hey, and IMHO that's a clearly formatted post. Well done.)
Posted 7/14/2018 4:30 PM (GMT 0)
Thank you

Glen
Posted 7/14/2018 6:14 PM (GMT 0)
These studies are almost useless because there is no accounting for the wide disparity in age, comorbidities, Gleason scores, and PSAs between the groups. There is a statistical technique called "propensity score matching" that can help compensate for for confounding variables (as was done in the Kishan study), but that was not done here. Also, there are wide variations in the RT treatments they got. "Confounding remained due to nonrandom allocation to treatment." As you can see, the HRs included 1.0, which rules out any statistically significant difference. 7 years of f/u is not long enough to detect a difference in survival among men with localized cancer.

A much better study was ProtecT, which randomized patients with localized to surgery, radiation, or active surveillance, finding no difference after 10 years. It did not include many high risk patients, however.
Posted 7/14/2018 6:19 PM (GMT 0)
I have had both and I am still alive!! Yea. LOL
Posted 7/15/2018 1:12 PM (GMT 0)

Tall Allen said...
These studies are almost useless because there is no accounting for the wide disparity in age, comorbidities, Gleason scores, and PSAs between the groups. There is a statistical technique called "propensity score matching" that can help compensate for for confounding variables (as was done in the Kishan study), but that was not done here. Also, there are wide variations in the RT treatments they got. "Confounding remained due to nonrandom allocation to treatment." As you can see, the HRs included 1.0, which rules out any statistically significant difference. 7 years of f/u is not long enough to detect a difference in survival among men with localized cancer.

A much better study was ProtecT, which randomized patients with localized to surgery, radiation, or active surveillance, finding no difference after 10 years. It did not include many high risk patients, however.

TA, you are pointing out the difference between observational studies and randomized clinical trials. Observational studies are less statistically refined than randomized trials but can be done on a larger population and still have good relevance. For example, ProtecT looked at about 1,643 men where the Swedish trial looked at 41,503. It does appear that the Swedish study attempted some "propensity score matching" because they mention that they adjusted the covariates to reduce confounding. Only randomized clinical trials like Kishan's can eliminate confounding. In the Swedish trial they had a lot of good data for a whole country population, whereas the Kishan and ProtecT had far fewer participants and more specific ones. So, overall, I think these types of large observational studies are relevant and should be encouraged.
Posted 7/15/2018 3:17 PM (GMT 0)
I'm thinking it's not so much the treatment you choose, it's more about how aggressive your cancer is to begin with...The Gleason 9 and 10 guys are fighting grade 5 prostate cancer which is a completely different animal than the much more common and survivable grade 3 and 4 variety...
Posted 7/15/2018 3:33 PM (GMT 0)
IMHO all data is important. That said we must as in all things in life be observant and decide for ourselves what is significant and what is not.

Roger
Posted 7/15/2018 3:34 PM (GMT 0)
Weightloss-

No amount of sample size in any retrospective study can make up for the fact that it is retrospective. The Swedish study was actually a registry study - it describes what happened to the whole population of treated patients regardless of how they were treated. It is therefore useless for the patient trying to decide between one specific therapy or another. They did not use "propensity score matching" which is a very specific statistical technique - where the sample is limited to those who in all important ways matches the test group. All they did was include some factors in their multivariable analysis. This is of very limited value when when there is not significant overlap between the groups (as with age) or when significant variables (like radiation dose or ADT extent) have not been captured.

Ronald Chen discusses the problems with drawing conclusions from registry data in this accompanying editorial to the Ennis et al NCDB study. As I wrote, "He points out that patient selection criteria are not completely reflected in comorbidity data. He believes that those who are selected for EBRT are just less healthy than those who can undergo anesthesia for surgery or brachytherapy. Other unmeasured confounders include burden of disease, and patient and physician preferences."

ascopubs.org/doi/10.1200/JCO.2017.77.5833

The Kishan study was not a randomized clinical trial. It was retrospective and multi-institutional. It is a lower level of evidence (level 2a) than, say, the ProtecT or ASCENDE-RT randomized clinical trials. It does reflect the best treatment given at some of the top institutions in the 21st century (i.e., since dose escalation and with 2 years of ADT). However, it does not account for such variables as the extent of pelvic lymph node treatment, and there were very low levels of salvage therapy given after radiation.
Posted 7/15/2018 6:15 PM (GMT 0)

Tall Allen said...
Weightloss-

No amount of sample size in any retrospective study can make up for the fact that it is retrospective. The Swedish study was actually a registry study - it describes what happened to the whole population of treated patients regardless of how they were treated. It is therefore useless for the patient trying to decide between one specific therapy or another. They did not use "propensity score matching" which is a very specific statistical technique - where the sample is limited to those who in all important ways matches the test group. All they did was include some factors in their multivariable analysis. This is of very limited value when when there is not significant overlap between the groups (as with age) or when significant variables (like radiation dose or ADT extent) have not been captured.

TA, if you look at the footnotes, the author of the Swedish study did the statistical analysis using R; and used the ROBBINS-I tool by Sterne. Propensity Score Matching is part of the tool.

Here is a reference /www.bmj.com/content/bmj/suppl/2016/10/12/bmj.i4919.DC1/stej032658.w1_default.pdf
Posted 7/15/2018 6:38 PM (GMT 0)

Tall Allen said...

Ronald Chen discusses the problems with drawing conclusions from registry data in this accompanying editorial to the Ennis et al NCDB study. As I wrote, "He points out that patient selection criteria are not completely reflected in comorbidity data. He believes that those who are selected for EBRT are just less healthy than those who can undergo anesthesia for surgery or brachytherapy. Other unmeasured confounders include burden of disease, and patient and physician preferences."

ascopubs.org/doi/10.1200/JCO.2017.77.5833

TA, I read through Ronald Chen's article on the limitations of the NCDB study. It was interesting in pointing out that the NCDB study had limited data to reduce confounders. If you look at the Swedish study, it had considerably more data than the NCDB study, such as the number of positive cores, which allows adjustment for seriousness of the PC. The Swedes had captured more data than was used for the NCDB study. It appears to be as robust as such oberservational studies can be and I don't see anywhere where the author was tardy. As both of us have pointed out these types of retrospective studies cannot eliminate confounding completely.
Posted 7/15/2018 6:40 PM (GMT 0)
ROBINS-I is a tool for calculating risk of bias, and has nothing to do with propensity score matching.

/www.ncbi.nlm.nih.gov/pmc/articles/PMC5062054/

R is a suite of statistical analyses (I used to use SAS which is similar). Just because a certain kind of statistical analysis is included as an option in the suite, doesn't mean the researchers chose to use it. If they had, they would make a point of it. Because they could not capture so many confounding variables, they could not use it. Not every statistical technique is valid for every data set.
Posted 7/15/2018 7:07 PM (GMT 0)

WeightLoss said...
The Swedes had captured more data than was used for the NCDB study. It appears to be as robust as such observational studies can be and I don't see anywhere where the author was tardy.

I'm not saying they were tardy. I think they did the best with what they had. But I'm saying that the study provides no direction for the interested patient, and like similar studies is useless for the patient.

Answer these questions and you will quickly see why:
What radiation techniques were used (e.g., EBRT, BT, SBRT, proton, 3D-CRT, 2D-RT - See Table 1)?
What radiation doses were used?
Was ADT given with radiation, and if so, for how long? Did it vary by risk group?
What percent of those originally having surgery had adjuvant/salvage radiation?
Did those having surgery have any systemic adjuvant therapy?
What percent of those having RT had adjuvant/salvage therapy?
Posted 7/16/2018 1:54 AM (GMT 0)

Tall Allen said...
ROBINS-I is a tool for calculating risk of bias, and has nothing to do with propensity score matching.

/www.ncbi.nlm.nih.gov/pmc/articles/PMC5062054/

R is a suite of statistical analyses (I used to use SAS which is similar). Just because a certain kind of statistical analysis is included as an option in the suite, doesn't mean the researchers chose to use it. If they had, they would make a point of it. Because they could not capture so many confounding variables, they could not use it. Not every statistical technique is valid for every data set.

TA, you are right. It doesn't look like it was used.
Posted 7/16/2018 1:59 AM (GMT 0)

Tall Allen said...

WeightLoss said...
The Swedes had captured more data than was used for the NCDB study. It appears to be as robust as such observational studies can be and I don't see anywhere where the author was tardy.

I'm not saying they were tardy. I think they did the best with what they had. But I'm saying that the study provides no direction for the interested patient, and like similar studies is useless for the patient.

Answer these questions and you will quickly see why:
What radiation techniques were used (e.g., EBRT, BT, SBRT, proton, 3D-CRT, 2D-RT - See Table 1)?
What radiation doses were used?
Was ADT given with radiation, and if so, for how long? Did it vary by risk group?
What percent of those originally having surgery had adjuvant/salvage radiation?
Did those having surgery have any systemic adjuvant therapy?
What percent of those having RT had adjuvant/salvage therapy?

I think it did provide some guidance and I don't think you disagree with the authors conclusion

In comparison with previous studies, the difference in Pca mortality after RT and RP was much smaller in this study than in previous studies and very small in absolute terms. Thus, the choice between these two treatments should be guided by the risk of side effects and patient preference rather than by the risk of Pca death.
✚ New Topic ✚ Reply