Optimal screening age scenarios

[darcyrao]

Historical (up to 2020):

• baseline screening with cytology age age 35
• bivalent vaccination at 86% coverage

Intervention scenarios:

• run out to 2120 to take a look, but most likely will cut at 2070 -- this will allow for CC to have developed in all women who might have been infected in 2020, but doesn't go out so far that start seeing huge effects of cases averted due to vaccination. We are interested in near-term screening policies, since in the long term will differ as vaccinated cohorts age through
• screening ages: 20, 25, 30, 35, 40, 45, 50
• screening algorithms:
  • cytology: sensitivity 57% for CIN2+
  • HPV testing: sensitivity 88% for CIN2+
  • triage and treatment: FOR NOW assume the same follow-up/receipt of treatment and treatment effectiveness for the two algorithms. Treatment with cryotherapy (in reality some women get LEEP if not eligible for cryo, so could try to find data on what proportion that is and adjust efficacy/persistent HPV params accordingly)
  • Differential treatment effectiveness: 77% HIV+, 91% HIV- (read Debeaudrap et al. 2019)
  • for post-treatment persistence, use 30%. From Hoffman et al. (2017), persistence is 48% for cryo and 28% for LEEP, but this seems like it includes treatment failure. Estimate of 30% accounts for the fact that some women would get LEEP and the fact that we already represent treatment failure...

Outcomes:

• Percent of cases averted and percent of deaths averted
  • By HIV status
  • Overall in a scenario where both groups screen at their optimal age?
• Incidence and mortality rates in 2020 and 2070 under the baseline scenario
  • Overall and by HIV status

[darcyrao]

Changes to the scripts

• mainCalibrated.m

  • Change filepath for directory to save results to "OptimalAge/filename"
  • set vaxRate to 0.86*(0.7/0.9)
  • change baseline screening to "baseline.screenCover = [0.0, 0.18, 0.48, 0.70, 0.70, 0.70, 0.70]" to have coverage for screening at 70% throughout the simulation period
  • for now keep the colpo retention and treatment coverage params the same (lower than set for analyses from 2020 on)

• slurm_mainCalibrated.sbatch

  • Change directory to my gscratch folder instead of Cara's, change email to mine
  • submit file from command line in login node as "sbatch -p csde -A csde slurm_mainCalibrated.sbatch"

• In simVaxCEA2.m:

  • define function at the top to loop through different ages and screening algorithms (pap or HPV testing): "function [] = simVaxCEA2 (AGESCRN, SCRNALG);"
  • change directory for historical run input file to "/HHCoM_Results/filename"
  • change directory to save results to "pathModifier = ['090319_OPTAGEPAP' , num2str(AGESCRN), num2str(SCRNALG)] ;" and change the if ~exist([.... part to create a folder in 'HHCoM_Results/OptimalAge'
  • set screenAlgorithm = SCRNALG
  • set school-based vaccination to be the same as the baseline (i.e. 9-year olds with bivalent vaccination at 86% coverage)
  • set baseline.screenAge = AGESCRN;
  • For cytology screening (baseline) and HPV testing (CISNET), set coverage to 70% from 2020 onwards
  • Set colpoRetain for both to 80%
  • Set cinTreatRetain and ccTreatRetain for both to 100%
  • Set cinTreatHpvPersist to 30%
  • set cisnet.screenAge = AGESCRN;

• slurm_simVaxCEA2.sbatch

  • Change directory to my gscratch folder instead of Cara's, change email to mine
  • change echo to describe the script (i.e. "Running MATLAB simVAxCEA2 for optimal age analyses."
  • under this line write: echo "${AGESCRN}" export AGESCRN echo "${SCRNALG}" export SCRNALG
  • change the line that starts with "matlab -nodisplay" to "matlab -nodisplay -nosplash -r "simVaxCEA2(${AGESCRN}, ${SCRNALG})"
  • submit file from command line in login node as "sbatch -p csde -A csde slurm_mainCalibrated.sbatch"

• Make a new script called OptAge.sh (type vi OptAge.sh)

  • This function will loop through the ages and screening algorithms to call sbatch for each one echo "Running simulations" for AGESCRN in $(seq 21 5 51); do # to loop through ages 20 - 50 in 5yr increments for SCRNALG in $(seq 1 1 2); do #to loop through pap and HPV testing export AGESCRN export SCRNALG sbatch -p csde -A scde slurm_simVaxCEA2.sbatch done done

• Then in command line just type sh OptAge.sh and it will run all scenarios

To export results:

• Transfer .mat files from Hyak to CSDE sim cluster, logout of Hyak and navigate to where I want to transfer files to on the sim cluster, then type: scp dpwhite@mox.hyak.uw.edu:/gscratch/csde/dpwhite/HHCoM/HHCoM_Results/Folder/*.mat . But replace "Folder" with the relevant folder.
• To export the results to Excel, use the vaxCEA_whoP2_082219.m script.
• In the "load saved results" section, save your historical file as "curr" (change directory to HHCoMResults\OptimalAge....) and scenario runs as "dirName[]." Then list all these "dirName_[]" values in the "simVec" vector, and matching file titles in the "fileTits" vector.
• Then hit Run and it should save everything for you in separate excel files. One note is that right now this is set up for ages 9-74 (indices 10:75). If you want a different age range, change all the instances of 10:75 to 1:80 in the non-age-standardized sections of the script (control find should do the trick). If you need age-standardized results let me know and I can help you change those sections of the script for different ages.

[darcyrao]

REVISED CHANGES WITHOUT THE FUNCTION (IT WAS CAUSING SOME ISSUES)

The changes to mainCalibrated are as indicated above. For SimVaxCEA2:

• change directory for historical run input file to "/HHCoM_Results/filename"
• change directory to save results to pathModifier = ['090519__scrnscenario' ; and change the if ~exist([.... part to create a folder in '/HHCoM_Results/OptimalAge'
• set screenAlgorithm = 1 for cytology or 2 for HPV DNA testing
• Create a new variable called AgeScrn = [age], where age = whichever age to target + 1
• set school-based vaccination to be the same as the baseline (i.e. 9-year olds with bivalent vaccination at 86% coverage)
• set baseline.screenAge = AgeScrn;
• For cytology screening (baseline) and HPV testing (CISNET), set coverage to 70% from 2020 onwards (in the baseline no screening scenario, set to 0.0 from current year on)
• Set colpoRetain for both to 80%
• Set cinTreatRetain and ccTreatRetain for both to 100%
• Set cinTreatHpvPersist to 30%
• set cisnet.screenAge = AgeScrn;

Then, each time I run it for the different scenario, change the pathModifier, the screenAlgorithm, and/or the AgeScrn and submit the job (wait until it starts running to make changes and submit the next one)

Skip the step about creating .sh file.

Follow the same instructions to export the results.

[darcyrao]

REVISED SCENARIOS AND SCRIPT CHANGE INSTRUCTIONS

---

Historical (up to 2020):

• baseline screening with cytology age age 35
• bivalent vaccination at 57% coverage
• No ART or VMMC scale-up from latest observed values

Intervention scenarios:

• run out to 2070 -- this will allow for CC to have developed in all women who might have been infected in 2020, but doesn't go out so far that start seeing huge effects of cases averted due to vaccination. We are interested in near-term screening policies, since in the long term will differ as vaccinated cohorts age through
• screening ages: none (implement by putting screening at age 0-4), 20-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64, 65-69
• screening algorithms:
  • cytology: sensitivity 57% for CIN2+
  • HPV testing: sensitivity 88% for CIN2+
  • triage and treatment:
  • with HPV screening, 81% compliance and 85% of CIN2+ accurately detected with same-day colpo, for 68.85% colpo 'retention'. With cytology, 72% colpo retention.
  • with HPV screening, 51% of diagnosed CIN2+ cases treated, 40% of diagnosed CC cases. With HPV screening, assume 100% tested and treated.
  • 28% HPV persistence
  • Differential treatment effectiveness: 77% HIV+, 91% HIV- (read Debeaudrap et al. 2019)
• No ART or VMMC scale-up from latest observed values

Outcomes:

• Percent of cases averted and percent of deaths averted relative to a scenario with no screening
  • By HIV status
  • Overall in a scenario where both groups screen at their optimal age?
• Incidence and mortality rates in 2020 and 2070 under the baseline scenario
  • Overall and by HIV status

Changes to the code, if editing a master branch. Otherwise, see below to make scenario-specific modifications, run analyses, and process outcomes using the OptScrnAge branch

• Clone the appropriate master branch to a new branch called OptScnAge (first switch to the appropriate master branch with git checkout [master] then run git checkout -b OptScrnAge
• Make the changes to loadUp2.m and historicalSim.m as indicated for the PAF_HIV analysis (issue #55). Set ART and VMMC not to scale-up, and set the vaccination coverage to 57% (adjusted for bi-valent). Differences from the PAF_HIV analysis for loadUp2.m: is that we will remove screenAge and screenAgeMults parameters from the cytgen and hpvgen scenarios. These are the screening algorithms we will use, and we will se the ages and age multipliers in the futureSim.m file. For both of these scenarios set coverage to 70% from currYear onwards.
• Make changes to the call to loadUp2.m in other files as indicated in issue #55. Set ART are VMMC not to scale-up in the future. Difference from the PAF_HIV analysis: revise the cyt0 screening algorithm to have coverage of 0 from curryear onwards.
• Make changes to futureSim.m to define two new variables around line 38 (under screenAlgorithm = 3':ScrnAges = [ #]andScrnAgeMults = [0.2];. Then, around line 137, above where the code defines% HIV-positive screening ages`, past the below code:

  
  % Screening ages
  cyt0.screenAge = ScrnAges;
  cytgen.screenAge = ScrnAges;
  hpvgen.screenAge = ScrnAges;

cyt0.screenAgeMults = ScrnAgeMults; cytgen.screenAgeMults = ScrnAgeMults; hpvgen.screenAgeMults = ScrnAgeMults;

- Make changes to the slurm scripts to use my username and email address.

**To run the scenarios**
- Run historicalSim.m and change the filepath to where and what I want to save the file as.
- Run the futureSim.m file with the following changes:
  - Update the pathmodifier for the historical sim
  - Define a new file name for the relevant scenario
  - In the screening section, set the screenAlgorithm to 3 if running cytology or 5 for HPV screening. Set the ScrnAges to a number 5-14 for age groups 20-24 through 65-69. Make sure hivPosScreen = 0.
  - If running the no screening scenario, set the screenAlgorithm to 2 (cyt0) and ScrnAges to [1]  (this will screen 0-4 year-olds, as a safeguard in case the coverage set in the cyt0 algorithm doesn't work...)
  - In the vaccination scenario, set to 57% ages 9-14 (age groups 2,3). For comparator, could do 90%. Make sure vaxCU etc are all set to 0.
- Submit the jobs through mox as indicated in issue #55 
- Export and process results as indicated in issue #55. A key difference with the export results script in Matlab is that we will not be saving HIV prevalence, and for this analysis we are defining crude incidence and mortality rates for ages 3 : age. We also don't need to save CC counts so ignore that. Change the directory names and filepaths to store results in the OptimalAge folder.