Dear StataList
I'm looking for advice re. creating "matched" treatment vs control cohorts for a KM survival analysis, possibly using a propensity score matching method (or similar) prior the survival analyses.
I've run a large ophthalmic analysis, comparing the outcomes of tube surgery (referred to as MIGS) for those with vs without prior laser surgery (referred to as SLT).
Despite our treatment + control groups having relatively similar demographics at baseline, reviewers suggested the effect of the SLT on the MIGS outcomes cannot be concluded in our fully adjusted stcox models and our KM survival curves, as there are only ~10,000 treated vs ~250,000 control patients (with slight variation in baseline demographics).
My PI suggested we create a "matched" cohort of ~10,000 controls, matching on age, sex, ethnicity, glaucoma type, and glaucoma severity (creating 10,000 "matched" controls rather than using all 250,000 of them), and then re-run the survival analyses to see if we get similar results. She's suggested using propensity score matching to do this.
I've looked at Stata videos on Youtube/read blogs, but it looks (to me) like propensity score matching is its "own" analysis method - rather than a technique you can use to create two "matched" cohorts with "similar" baseline risk profiles for subsequent survival analyses; i.e. looking at ATET, and similar outputs - rather than creating a "similar" baseline control cohort that can be used for later survival analyses.
Any advice re. how to best create a "matched" control cohort (of those without prior SLT, matched on the above 5 baseline criteria) would be appreciated.
Below is a sample of my data.
There are 4 cohorts - those undergoing MIGS only (with vs without prior SLT), and those undergoing MIGS + phacoemulsification (essentially cataract surgery, which is sometimes done in addition to the MIGS operation) (with vs without prior SLT).
Thanks for your consideration,
William
I'm looking for advice re. creating "matched" treatment vs control cohorts for a KM survival analysis, possibly using a propensity score matching method (or similar) prior the survival analyses.
I've run a large ophthalmic analysis, comparing the outcomes of tube surgery (referred to as MIGS) for those with vs without prior laser surgery (referred to as SLT).
Despite our treatment + control groups having relatively similar demographics at baseline, reviewers suggested the effect of the SLT on the MIGS outcomes cannot be concluded in our fully adjusted stcox models and our KM survival curves, as there are only ~10,000 treated vs ~250,000 control patients (with slight variation in baseline demographics).
My PI suggested we create a "matched" cohort of ~10,000 controls, matching on age, sex, ethnicity, glaucoma type, and glaucoma severity (creating 10,000 "matched" controls rather than using all 250,000 of them), and then re-run the survival analyses to see if we get similar results. She's suggested using propensity score matching to do this.
I've looked at Stata videos on Youtube/read blogs, but it looks (to me) like propensity score matching is its "own" analysis method - rather than a technique you can use to create two "matched" cohorts with "similar" baseline risk profiles for subsequent survival analyses; i.e. looking at ATET, and similar outputs - rather than creating a "similar" baseline control cohort that can be used for later survival analyses.
Any advice re. how to best create a "matched" control cohort (of those without prior SLT, matched on the above 5 baseline criteria) would be appreciated.
Below is a sample of my data.
There are 4 cohorts - those undergoing MIGS only (with vs without prior SLT), and those undergoing MIGS + phacoemulsification (essentially cataract surgery, which is sometimes done in addition to the MIGS operation) (with vs without prior SLT).
Thanks for your consideration,
William
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input str43 id_no float(glaucoma_type glauc_severity) long _sex float(race_ethnic _age migsphaco_withslt_2yr migsnophaco_withslt_2yr) "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "00000fce9b964b7bbba0e45ca5acf6c7Left" . . 2 1 51 . . "000049ec671d4da491f12738eb517ff6Right" . . 1 1 71 0 . "000049ec671d4da491f12738eb517ff6Right" . . 1 1 71 0 . "000049ec671d4da491f12738eb517ff6Right" . . 1 1 71 0 . "000049ec671d4da491f12738eb517ff6Right" . . 1 1 71 0 . "000049ec671d4da491f12738eb517ff6Right" . . 1 1 71 0 . "000049ec671d4da491f12738eb517ff6Right" . . 1 1 71 0 . "000049ec671d4da491f12738eb517ff6Right" . . 1 1 71 0 . "000049ec671d4da491f12738eb517ff6Right" . . 1 1 71 0 . "00005f63385542f78f87378a11a10870Right" . . 2 1 73 0 . "00005f63385542f78f87378a11a10870Right" . . 2 1 73 0 . "00005f63385542f78f87378a11a10870Right" . . 2 1 73 0 . "00005f63385542f78f87378a11a10870Right" . . 2 1 73 0 . "00005f63385542f78f87378a11a10870Right" . . 2 1 73 0 . "00005f63385542f78f87378a11a10870Right" . . 2 1 73 0 . "00005f63385542f78f87378a11a10870Right" . . 2 1 73 0 . "00005f63385542f78f87378a11a10870Right" . . 2 1 73 0 . "00005f63385542f78f87378a11a10870Right" . . 2 1 73 0 . "00005f63385542f78f87378a11a10870Right" . . 2 1 73 0 . "0000b8b8d04046408e96445ac5694055Right" . . 1 1 81 0 . "0000b8b8d04046408e96445ac5694055Right" . . 1 1 81 0 . "0000b8b8d04046408e96445ac5694055Right" . . 1 1 81 0 . "0000b8b8d04046408e96445ac5694055Right" . . 1 1 81 0 . "0000b8b8d04046408e96445ac5694055Right" . . 1 1 81 0 . "0000c6d5083244a0a1028555fccb8966Left" . . 1 3 64 . . "0000c6d5083244a0a1028555fccb8966Left" . . 1 3 64 . . "0000c6d5083244a0a1028555fccb8966Left" . . 1 3 64 . . "0000c6d5083244a0a1028555fccb8966Left" . . 1 3 64 . . "0000c6d5083244a0a1028555fccb8966Left" . . 1 3 64 . . "0000c6d5083244a0a1028555fccb8966Left" . . 1 3 64 . . "0000c6d5083244a0a1028555fccb8966Left" . . 1 3 64 . . "0001173042a24cb28834472bd087e9ccRight" . . 1 1 65 0 . "0001173042a24cb28834472bd087e9ccRight" . . 1 1 65 0 . "0001173042a24cb28834472bd087e9ccRight" . . 1 1 65 0 . "0001173042a24cb28834472bd087e9ccRight" . . 1 1 65 0 . "0001173042a24cb28834472bd087e9ccRight" . . 1 1 65 0 . "0001173042a24cb28834472bd087e9ccRight" . . 1 1 65 0 . "0001173042a24cb28834472bd087e9ccRight" . . 1 1 65 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 1 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 1 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 1 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 1 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0001df54663c4ed093efc3c5e2feddebRight" 2 . 2 1 80 0 . "0002584e0f8841fa89b1104e788db1aeLeft" . . 2 1 80 0 . "0002584e0f8841fa89b1104e788db1aeRight" . . 2 1 80 0 . "0002584e0f8841fa89b1104e788db1aeRight" . . 2 1 80 0 . "0002584e0f8841fa89b1104e788db1aeRight" . . 2 1 80 0 . "0002584e0f8841fa89b1104e788db1aeRight" . . 2 1 80 0 . end label values glaucoma_type _glaucoma_diagosis label def _glaucoma_diagosis 1 "suspect", modify label def _glaucoma_diagosis 2 "poag", modify label values glauc_severity _severity label def _severity 1 "mild", modify label values _sex _sex label def _sex 1 "Female", modify label def _sex 2 "Male", modify label values race_ethnic _race_ethnic label def _race_ethnic 1 "NH White", modify label def _race_ethnic 3 "Hispanic/Latino", modify
Comment