Wikidata:Property proposal/medicine marketing authorisation

medicine marketing authorisation edit

Originally proposed at Wikidata:Property proposal/Natural science

Done: medicine marketing authorization (P3464) (Talk and documentation)

Description	medicine is allowed by the EMA to be marketed for the treatment of
Data type	Item
Domain	drug-disease interaction (Q26720101)
Allowed values	product authorization status (Q26720715)
Example	Levetiracetam for epilepsy treatment (Q26720719) → EMA authorized for human use (Q26720713) Source: http://www.ema.europa.eu/ema/index.jsp?curl=pages/medicines/human/medicines/002051/human_med_001531.jsp&mid=WC0b01ac058001d124
Source	http://www.ema.europa.eu/ema/index.jsp?curl=pages/includes/medicines/medicines_landing_page.jsp&mid= https://open.fda.gov/ (has to be queried)

Motivation

The information about drug approvals for therapeutic interventions is very important to many stakeholders. It's complex information and thus easier to be displayed on a separate drug-disease interaction (Q26720101) instead of being displayed on the page of a drug. ChristianKl (talk) 12:20, 30 August 2016 (UTC)[reply]

Notified participants of WikiProject Medicine

Discussion

Comment Terrible model imho. I'd prefer an item for each type of authorization - USA public health authorization, ... . Then associate each of those items to its jurisdiction. You don't get the qualifier.

Support after those changes. author TomT0m / talk page 14:48, 30 August 2016 (UTC)[reply]

Okay I changed this property to be about the EMA medicine marketing authorisation. The EMA publishes data that can be used to fill the property. The FDA at this point in time doesn't. I emailed the FDA and they verified that they don't provide this data.ChristianKl (talk) 09:44, 31 August 2016 (UTC)[reply]

Fortunately it seems that the FDA provides it through the API on https://open.fda.gov/ even when they don't provide it for normal users, so we could import it. ChristianKl (talk) 17:50, 31 August 2016 (UTC)[reply]

I agree with TomT0m. Getting a dataset which verifies these relationships would be difficult. If datasets did exist, then I expect that they would apply only to a certain country and from a certain date. Two different countries may even define the medical condition to be treated differently, or say that a treatment is for the same condition but in different circumstances. This would be difficult to manage. Blue Rasberry (talk) 21:31, 30 August 2016 (UTC)[reply]

At the moment I don't think that a dataset exist that contains information from multiple countries. It would be a chance for Wikidata to be the first dataset that does this.

As far as definition of illnesses go, I'm currently not sure whether the EMA actually deals in definitions. I wrote them a mail to ask to them to clarify. Fortunately there are treaties that suggest that all countries in the WHO have to use ICD codes to classify illnesses so there's hope that in practice the same definition get's used. ChristianKl (talk) 09:44, 31 August 2016 (UTC)[reply]

I am not sure why the instance of drug use is itself the Wikidata item. It seems analogous to the concept of "inserting nails" being a use of a hammer, and I like the idea of tracking all the ways that anything might be used, but I am not sure how this should look in Wikidata. It is an interesting idea but I have trouble thinking through how this works. Blue Rasberry (talk) 21:31, 30 August 2016 (UTC)[reply]

The analogy would be "inserting nails with a hammer" being an item. It might be a valid value for "main subject" of an article that deals with how to use a hammer to insert nails but as long as the item isn't needed it doesn't have to be created.

In the absence of having a drug-disease iteraction item, the information would have to be stored both in the drug item and the disease item and be duplicated.

It would also be more complicated to enter specific start/stop times for approvals.

Having a separate item allows for better information storage. ChristianKl (talk) 09:44, 31 August 2016 (UTC)[reply]

@ChristianKl: An alternative would be to use qualifiers who are suitable to represent some kinds of n-ary relationships. I guess you refer however the lack of ways to reverse the statements whithout using SPARQL - this is likely to be resolved in not so much time as the devteam is currently working on how to deal with queries on Wikidata.

On you modification : I was more thinking on something like

⟨ subject ⟩ authorization Search ⟨ EMA authorization ⟩

, with

⟨ EMA authorization ⟩ décerné par Search ⟨ EMA ⟩

⟨ EMA authorization ⟩ for juridiction Search ⟨ European union ⟩

... Maybe also something to link the authorization (class) item to the (class of) action it authorizes in the jurisdiction - prescription by doctors, selling the stuff.~ author TomT0m / talk page

It's not binary. There are four classes. Authorized, Withdrawn, Denied and Requested. Additionally there's potentially 'no value' which might be interesting to fill for FDA approved drugs where no approval is sought with the EMA.

It's possible that there will be soon a another class of "Conditional marketing authorisation"

I think it's generally useful to keep complexity down. It might be possible to lists a lot of information on the item on a drug with n-array statements but that makes everything more complicated. In computer programming there the general principle of keeping functions as short as possible. It helps to keep the structure of a program simple.

It would be great to be able to point reasonator at a drug-disease interaction and get a list of all drug approvals and all relevant clinical trials. ChristianKl (talk) 11:21, 31 August 2016 (UTC)[reply]

@ChristianKl: The "withdrawn" case, in the case of Wikidata, should be dealt with a "end date" qualifier. I guess the "refused" statement could also be dealt with a generic class "EMA refused medicine"

⟨ the refused cure ⟩ instance of (P31) ⟨ EMA refused medicine ⟩

or with a "no-value".

Complexity can only be measured by comparing the different models in depth and how complicated it's to use them, and is hard to use as an abstract concept. In particular, we don't have to map the EMA concepts with a 1-1 correspondance, we can use wikidatas concept - the classes they define can probably dealt with instance of (P31) and don't have to get a specific property, as we have a property to deal with classes. On the other hand, the concept of authorization of medecine is not Europe centric. author TomT0m / talk page 12:53, 31 August 2016 (UTC)[reply]

I think it's valuable to have a property for marketing authorization. I think it's valuable if the property can model "refused" as a different state as "no value". If I live in the US and take an FDA approved medicine than it tells me something different when the EMA refused to approve it in contrast to the EMA not having ruled on it.

The concept isn't authorization of medicine but authorization of marketing of medicine for a therapeutic area. Doctors are allowed (authorized) to use medicine off-label.

It's a difference concept from whether a substance is legal and the two can be easily confused by laypeople. I'm also not sure whether the split works the same way everywhere in the world. ChristianKl (talk) 14:20, 31 August 2016 (UTC)[reply]

@ChristianKl: Browsing http://www.fda.gov/Drugs/InformationOnDrugs/default.htm a litte bit it appears yes, FDA approved drugs for some diseases as well. It's the same concept. I think we totally can create classes "FDA unauthorized" to deal with rejection. I guess if we actually want to refer to authorization processes we could label stuffs as "authorisation request". This would allow us to create an item for the authorization process and to link to the clinical studies put to the file for example. author TomT0m / talk page 14:33, 31 August 2016 (UTC)[reply]

I edited the concept again. Are you okay with the new way? ChristianKl (talk) 15:00, 31 August 2016 (UTC)[reply]

Support. author TomT0m / talk page 18:32, 31 August 2016 (UTC)[reply]

Notified participants of WikiProject Medicine Does anyone have any input? ChristianKl (talk) 11:29, 30 November 2016 (UTC)[reply]

Not sure, confused I asked about the nails earlier. I do not know what to say. Can I ask for further discussion? Should all of these have Wikidata items?

Levetiracetam for epilepsy treatment (Q26720719)
Levetiracetam for epilepsy treatment (Q26720719)
Keppra for epilepsy treatment (a brand name, assuming that it is relevant to note brand differences)
Racetam (Q1649216) for epilepsy treatment (the class of drug)
5mg Keppra brand tablets for epilepsy treatment in Caucasian males age 18-55 (sometimes there lots of notes)

I am not sure what is reasonable to expect and I really do not know what is appropriate. I am not looking for a long explanation and do not need much convincing, but I am still not sure why an item should be the intersection of "drug" and "use". Is there a Wikidata precedent for combing objects and use, like "ink for pens" versus "ink for newspapers"?

Thanks for experimenting and discussing here. Blue Rasberry (talk) 19:55, 30 November 2016 (UTC)[reply]

I'm not sure what difference you imply between 1 and 2.

Currently we don't have separate items for Keppra and Levetiracetam in the first place in Wikidata ontology. I think we should have separate items for those. I would be happy with both 2 and 3.

There a Wikidata precedent for having items for the borders between two countries that aren't just statements in country A and B. ChristianKl (talk) 11:51, 3 December 2016 (UTC)[reply]