OHDSI Home | Forums | Wiki | Github

Project SCYLLA Protocol Finalized; Community Feedback Sought On Symptoms & Outcomes Within This Thread

(Craig Sachson) #1

The Project SCYLLA (SARS-Cov-2 Large-scale Longitudinal Analyses) protocol has been finalized and is available here. SCYLLA’s objectives are to assess the comparative safety and effectiveness of all emerging drug therapies used in COVID-19 treatments …
… administered during hospitalization and prior to intensive services.
… administered during hospitalization after initiating intensive services.
… administered after COVID-19 positive testing and prior to hospitalization.

We believe this set of analyses could provide a critical impact in informing the overall effectiveness of various treatments being used globally against COVID-19. We have databases from 10 nations across Asia, Europe and North America, so we will be able to generate reliable real-world evidence at scale to inform decision-making around this pandemic and are seeking active participation from around the world. We continue to seek more collaborators to join in this important network study; any researchers with data in the OMOP CDM are encouraged to ‘join the journey.’

We have set up a specific thread on the forum seeking community feedback for the treatments and outcomes under investigation; you can find these in Appendix A & B of the protocol. Please use this thread for all feedback. Thank you!

[COVID-19] May 12 Update
(Alan Goldhammer) #2

Two stylistic comments. I’ve see both Arbidol and umifenovir are used; since the list of treatments uses the generic name, I would suggest using umifenovir. The other point is a minor nit. There is no uniform use of capital letter in the first listed treatment name. Either capitalize them or leave them lower case. This may not bother those of us readily familiar with drug names, but it is inconsistent: Bevacizumab versus eculizumab.

From my reading of pre-prints, there have been so many hydroxychloroquine combinations used. Are the investigators confident that this won’t cause interpretational problems as data is acquired?

The list of different drugs appears to be complete by the list I have been maintaining.

(Vojtech Huser) #3

I have a note similar like outcomes but on larger context.

In section 9.4 (Data sources) we may consider adding some sentences about how datasets are handled.


In order to aggregate data from several datasets, each contributing data source will provide some minimum information about a dataset (MIAD).

For example:

databaseId <- "MDCD"

databaseName <- "Truven Health MarketScan® Multi-State Medicaid Database"

databaseDescription <- "Truven Health MarketScan® Multi-State Medicaid Database (MDCD) adjudicated US health insurance claims for Medicaid enrollees from multiple states and includes hospital discharge diagnoses, outpatient diagnoses and procedures, and outpatient pharmacy claims as well as ethnicity and Medicare eligibility. Members maintain their same identifier even if they leave the system for a brief period however the dataset lacks lab data. [For further information link to RWE site for Truven MDCD."

In addition, some basic characteristics about the database as a whole may be collected (e.g., VA data may contain more data on male patients). Database overall size is another collected metadata parameter. Some dataset may contain only data (or have more data) from certain care settings (e.g., inpatient care vs. outpatient care).

Also, I think the protocol (for feedback) should not be shared within OHDSI in PDF format. PDFs are not FAIR.

Note: Btw, MIAD was more developed here. https://github.com/vojtechhuser/DataQuality/blob/master/extras/archivedReadme.md#generate-miad-minumum-information-about-a-dataset The dependency on Achilles can be easily removed and that function rewritten to be fully self standing.

(Cynthia Sung) #4

Would it be possible to work backwards, ie identify a group of patients who had poor outcomes vs those with milder disease, balance them with baseline characteristics and then do an AI search through their history of medication usage to find differences? This might turn up some unexpected candidates, then work forwards again using the algorithms as described in the paper.

Is there a possibility of studying combinations of certain drugs with single agents as the comparator? Specifically, a suggestion was raised in the ACE/ARB Teams channel about statin+ARB. Would statin alone or ARB alone be a suitable comparator group?

(Alan Goldhammer) #5

@CSung - this has been my thinking for some while since the controversy about whether ARBs were injurious or protective. There are a handful of trials on losartan, valsartan, and more recently a couple of statins. Certainly in the US there are a large number of people being treated for blood pressure and elevated cholesterol that are on these medications. The age range is also quite wide but it’s not clear how many are on monotherapy of statin alone.

I am hoping that SCYLLA might be able to pick this apart and perhaps come up with an age/medication/COVID-19-outcome cohort to see if there might be some protection from these treatment regimens. I’ve seen several recent preprints coming out from several New York hospitals but none of them have looked at this.

(David Vizcaya) #6

If I understand well, you are proposing a nested case-control study (or a series of case-control studies) and do a kind of hypothesis-generating evaluation of treatments and their combinations, right? that would be a great way to analyze different exposures! It would have some limitations of course, but just like any other approach.

(Dalia) #7

@CraigSachson I hope it is not too late to give a couple of suggestions for the protocol:

  • Drugs: I propose also including the antivirals zanamivir and peramivir. These have been used before for H1N1 and may have been trialled in some patients. Also,I am not sure if you plan to characterise the drugs used in terms of doses, frequency of administration and length of treatment course but if not, I suggest trying to do that and not group all regimens together under one drug to help better inform practice. These will also help economic analyses that could be undertaken later.

  • Effectiveness Outcomes: I did not see admission to ICU but you have hospitalisation requiring intensive intervention, so I assume these are the same(?). I would suggest also looking at length of stay (days) in hospital and length of stay in ICU. These are outcomes that were reported I think in the trials of remdesivir and again would be helpful outcomes to report for future economic analysis. VTE events can also occur post discharge, as they may start during hospitalisation as asymptomatic event but change to symptomatic later, so I suggest looking at CTE events up to 90 days post discharge if you will have access to post discharge data.

  • Safety outcomes: these are more difficult to be comprehensive about as they will be specific for each drug/intervention you are looking at. However, I suggest adding secondary bacterial infections. these will be relevant for MAbs and other drugs in the list I imagine

I am really looking forward to contributing to this work and to seeing the results which could be very useful in cost-effectiveness analysis of these candidate interventions too!
All the best

(Dalia) #8

@CraigSachson Forgot to also add the link to this JAMA article on potential treatments for COVID19 if you haven’t seen it

I did not see ethnicity in the covariates and not sure if you capture this in the data you have access to, but would be relevant if you can.


This article provides an interesting insight into objective number 4 under research question and objectives.

Emerging Prophylaxis.pdf (132.9 KB)

(Joshua Vogelstein) #10

For the alpha blockers, we now have specific evidence that tamsulosin is effective, https://arxiv.org/abs/2004.10117. I wonder whether it would be possible to add tamsulosin, in addition to the class of alpha blockers. Also, while the trial we have going for alpha blockers is currently an inpatient trial (https://clinicaltrials.gov/ct2/show/NCT04365257), we are in the process of getting funding to start an outpatient trial, again targetting tamsulosin. So ideally both alpha blockers (in general) and tamsulosin (in specific) could be included for analysis in the outpatient setting. I discussed this with @Patrick_Ryan in teams already, so sorry if I am being redundant here.