Property talk:P2176

From Wikidata
Jump to navigation Jump to search


drug used for treatment
drug, procedure, or therapy that can be used to treat a medical condition
Representsmedication (Q12140)
Data typeItem
According to this template: diseases
According to statements in the property:
health problem (Q2057971)
When possible, data should only be stored as statements
Exampleskin melanoma (Q18558032)dabrafenib (Q3011604)
According to this template: National Drug File (
According to statements in the property:
When possible, data should only be stored as statements
Robot and gadget jobsWill be added and maintained by User:ProteinBoxBot/Drug_items
Tracking: usageCategory:Pages using Wikidata property P2176 (Q26678450)
See alsopossible treatment (P924), therapeutic area (P4044), may prevent (P4954)
Proposal discussionProposal discussion
Current uses1,217
Search for values
[create Create a translatable help page (preferably in English) for this property to be included here]
Type “health problem (Q2057971): element must contain property “instance of (P31)” with classes “health problem (Q2057971)” or their subclasses (defined using subclass of (P279)). (Help)
Exceptions are possible as rare values may exist.
List of this constraint violations: Database reports/Constraint violations/P2176#type Q2057971, SPARQL, SPARQL (new)
Value type “medication (Q12140): This property should use items as value that contain property “instance of (P31)”. On these, the value for instance of (P31) should be an item that uses subclass of (P279) with value medication (Q12140) (or a subclass thereof). (Help)
Exceptions are possible as rare values may exist.
List of this constraint violations: Database reports/Constraint violations/P2176#Value type Q12140, SPARQL, SPARQL (new)

Please notify projects that use this property before big changes (renaming, deletion, merge with another property, etc.)

Data usage[edit]

What is using this data? Nemo 15:22, 4 September 2016 (UTC)

Inverse or super-property[edit]

How does this property relate to possible treatment (P924)? -- JakobVoss (talk) 10:30, 15 October 2016 (UTC)


The description "drug that is used to treat this medical condition (this isn't a statement about the effectiveness of the drug)" makes no sense. Who would be using a drug to treat a medical condition that is not effective? And even if they were, how could that be described as "treating"? In addition, what use could this property possibly have? There's a value in having a property that lists drugs that are have some effectiveness as a treatment, but what use is there in listing statins as a treatment for pneumonia? {Hint: doctors don't treat pneumonia with statins). We either need another property that conforms with the normal English usage of the verb "to treat", or we need to modify this property to exclude drugs that have no effectiveness as treatment. --RexxS (talk) 22:59, 12 December 2016 (UTC)

this would be OK if it were filled with names of a) things that actually are drugs, and b) that actually are used for treatment. There doesn't have to be any comment on how effective they are (e.g which NSAID is better for a headache). What is problematic, is adding entries based on data in CHEMBL that just records stuff that scientists have done in the lab or in clinical trials, not something that doctors do (or that people do when they buy and use OTC drugs). Jytdog (talk) 02:56, 13 December 2016 (UTC)
@RexxS: In practice many treatment decisions are off-label treatments. Homeopathic drugs get used by doctors to treat various illnesses even when they aren't effective. The normal English meaning of the word treat in no way implies that there's no doctor who treats pneumonia with homeopathic medicine.
Many drugs do have side-effects for which they aren't proven to be effective. Having a list of illnesses which scientists tried to treat with the drug is useful information, because it points at possible side-effects. That's both interesting information if you take a drug and care about what the drug might do to you and it's important information for bioinformatic applications that try to find patterns in large pile's of data. ChristianKl (talk) 17:26, 22 February 2017 (UTC)
@ChristianKl: Actually, no, in practice doctors who want to keep their licences don't prescribe drugs that have no evidence base. Secondly there's no such thing as a "homeopathic medicine". There's either medicine that has an evidence base for its use, or there's "snake oil", and we do nobody any favours by creating lists of drugs that have been tried and failed to have any effect on a given condition. Third parties - as well as Wikipedias - who make use of our data do not see the description for P2176 (which of course could be changed at any time); they see a label in their language corresponding to "drug used for treatment". It is thoroughly irresponsible to present misleading information to third-parties in this way and it needs to stop before serious injury occurs as a result. --RexxS (talk) 18:05, 22 February 2017 (UTC)
I did study bioinformatics. Huge structured databases are valuable in bioinformatics. Using this category isn't something that Wikidata invented. Government databases like the National Drug File - Reference Terminology (NDF-RT) uses the term "May be treated by" for this category. For bioinformatic purposes it's valuable to have this kind of data and that's why NDF-RT hosts the data under such a label. The problem comes when third-party uses don't understand the nature of the represented information. For pneumonia you can see their list:
That problem exists the same way when they deal with the NDF-RT directly or when they deal with that data as it's imported into Wikidata. Wikipedia has a valid concern when it doesn't want to import this data because it might be misleading to the average person who doesn't understand the nature of the data but there's nothing forcing Wikipedia to import this data and represent it in a misleading context.
I think the intention of helping bioinformatic applications by aggregating information like the NDF-RT that's valuable for bioinformatic purposes seems from my perspective a valuable goal for Wikidata.
Wikidata foundational idea is that different kinds of knowledge can coexist. If you don't like what another source says you are free to add "statement disputed by". The way to deal with knowledge that's conceptualized in a different way isn't to simple remove it. In this case we are talking about knowledge from the Western scientific establishment of the field of bioinformatics that's stored in NIH funded databases. Serious scientific sources.ChristianKl (talk) 19:16, 22 February 2017 (UTC)
As far as off-label usage goes Wikipedia says "A 2009 study found that 62% of U.S. pediatric office visits from 2001-2004 included off-label prescribing". The idea that off-label usage by doctors doesn't happen in reality because doctors have to fear that it loses them their license doesn't seem to have a basis in reality.
Even the FDA says "From the FDA perspective, once the FDA approves a drug, healthcare providers generally may prescribe the drug for an unapproved use when they judge that it is medically appropriate for their patient. " ChristianKl (talk) 19:30, 22 February 2017 (UTC)
  • Perhaps change the disclaimer to "(Wikidata can make no warranty as to their effectiveness)".
I also agree that it would be useful to tighten the property to "drugs that have been licensed to treat this condition", since (i) the licensing in itself is useful information; and (ii) it's more likely to be verifiable. Jheald (talk) 18:31, 22 February 2017 (UTC)
@Jheald: It's misleading to say "Wikidata can make no warranty as to their effectiveness" when we basically present the data that NDF-RT aggregated. It suggests that we believe that there's clinical effectiveness but just aren't certain. The way the NDF-RT makes their list isn't based on only listening drugs that have been FDA approved (or EMA approved).
It's worth noting that there are similar issues with the present proposal (
We have recently added properties for the EMA and FDA ID for drugs. That means we can import all their information as well and Wikipedia can hopefully simple import the drugs that passed phase 4 for a condition and are thus FDA/EMA approved. At the same time it's worthwhile for bioinformatic purposes to import more than just the phase 4 data. It's also worthwhile to label it in a way that it isn't believed to be stronger indication that a drug works than it is. [
It's also worthy to note that the property proposal discussion was about a property that represents the NDF-RT data. There wasn't a suggestion in this discussion that the property is supposed to mean something different than what the NDF-RT means. [User:ChristianKl|ChristianKl]] (talk) 19:16, 22 February 2017 (UTC)

Doc James
Daniel Mietchen
Andrew Su
Projekt ANA
Pavel Dušek
Was a bee
Chris Mungall
Dr. Abhijeet Safai
Sami Mlouhi
Netha Hussain
Abhijeet Safai
Shani Evenstein
ZI Jony
Pictogram voting comment.svg Notified participants of WikiProject Medicine ChristianKl (talk) 19:24, 22 February 2017 (UTC)

  • ChristianK I just reviewed this thread. Three responses..
    • 1) Rexx said "doctors who want to keep their licences don't prescribe drugs that have no evidence base." What you wrote about 62% of pediatric office visits supports what Rexx wrote; "medically appropriate" = there is evidence for it.
    • 2) Phase 4 is not required for FDA or EMA approval but you write as though it is. That is disturbing.
    • 3) Not every NIH-funded database is relevant. Again - a bunch of entries were added to this field, based on in vitro experiments that were recorded in CHEMBL. Arguing that some chemical (which may be toxic as hell) that was a hit in some assay belongs in a field called "drugs used for treatment" is ridiculous and dangerous.
    • It is going to be impossible to get consensus in en-WP to import data from WD for anything related to medicine in any regular way as long as there are not better controls on changing field definitions, and until there are much more serious controls on changing data. Jytdog (talk) 20:34, 22 February 2017 (UTC)
agree w/ Jytdog--Ozzie10aaaa (talk) 11:25, 23 February 2017 (UTC)
  • This is not in any reasonable sense a change in the intention of the property proposal. The property proposal was for a property that mirrors NDF-RT. The problem here is that people who read the property definition thought it meant something different than it means in our data-source. I edited the description to be more clear about what the data that's contained in the property is about.
There was never a time when this property was designed to represent anything like "FDA approved drugs" or the description of it spoke of clinical effectiveness.
I went through the pneunomia drugs at NDF-AT. The second I looked up is ALATROFLOXACIN. It was withdrawn from the market in 2006 due to "Liver toxicity; Serious liver injury leading to liver transplant; Death". It's still in the NDF-AT list. ChristianKl (talk) 21:00, 22 February 2017 (UTC)
Please read the proposals for this property and its inverse: Wikidata:Property proposal/Archive/35 #P2175. Note these comments: "This property shall be used to link a drug to the diseases it can be used for treating" and "at least one dictionary described 'treatment' as 'relief or cure' a disease", and "perhaps confusion would reduce if the property would be linked to the fact that 'the drug is accepted a prescription treatment for'?". There is no mention of the modification you have unilaterally introduced. If you want a property that is defined as "drug appears on the NDF-RT database connected to a given condition", then propose it and take your rationale for such a property there. This property is understood to connect a drug with the condition it treats, and drugs should not be connected to conditions that they do not treat. If that means that a particular database can't fit that definition, then we shouldn't be importing from that database into this property. A clear example of the problem is found in the edit history of pneumonia (Q12192), where on 11 December 2016, Wikidata had the claim that pneumonia had a "drug used for treatment" which was simvastatin (Q670131). Now the truth is simple: simvastatin does not treat pneumonia, and no doctor in his right mind would prescribe simvastatin to treat pneumonia. The confusion is caused because there is a single study which found that people who were taking statins long-term had slightly better outcomes in recovering from pneumonia, and that translated into a connection in the record at Frankly, we need to be far better at matching what is imported into Wikidata to what is genuinely useful for the end-user. It is most unhelpful to pollute data about what the vast majority of people understand by "this drug treats this condition" with misleading and irrelevant connections on the grounds that some researchers might want to know whether a drug was once tried in a failed experiment. And it is madness to change the property definition to attempt to fit the data. You appear to think that the property definition has to be changed from what was proposed because the data sources don't match what people expect. That's the wrong way round and I'm going to remove your edit as it is clear that consensus is against you here. --RexxS (talk) 02:46, 24 February 2017 (UTC)
  • @ChristianKl: This is a conversation going in multiple directions. You pinged the medicine project. Did you have a particular question or want a certain kind of comment? Mostly it seems like there are misunderstandings here, and probably everyone is correct in their own sense. Blue Rasberry (talk) 21:07, 22 February 2017 (UTC)
  • It worth to agree on what this property should represent. Given the idea of creating an analogous "may prevent" property it's also worthwhile to form a consensus on what kind of data we store and how we describe the data we store. ChristianKl (talk) 21:19, 22 February 2017 (UTC)
  • "drug used for treatment" Treatment implies effectiveness else it is not a treatment so agree makes no sense. Doc James (talk · contribs · email) (if I write on your page reply on mine) 13:44, 23 February 2017 (UTC)
  • Turning more narrowly to the topic of this thread (a disclaimer of effectiveness) some of this is about... the medical disclaimer issue. The current WD disclaimer is here: Wikidata:General_disclaimer. As folks from en-WP know this is a very old and contentious issue in en-WP with regard to health content, and again as the en-WP folks know, the long term solution (still contested sometimes) has been the medical disclaimer that is included in the general disclaimer. That whole bucket of issues is at play here, I believe. Jytdog (talk) 18:34, 23 February 2017 (UTC)
  • My comment wasn't there as a disclaimer. The database from which we import the data says "may treat". There's assumed uncertainty. If you remove this suggestion than this uncertainty isn't accurately represented to our users. That's bad and not simply an issue of having a disclaimer. Encouraging people to take data has more certain than it is, is harmful. ChristianKl (talk) 07:34, 25 February 2017 (UTC)
There is absolutely no value in having a property titled "drug used for treatment" and then importing a database of drugs that may treat a condition - or may not by implication. If you want a property that represents "drugs that have at some point been considered for treating a given condition, whether or not they had any effect", then by all means propose it. But such a property would not be useful to any of the Wikipedias, and it is not acceptable to hijack a property useful all Wikipedias ("drug that treats this condition") and render it useless to the readers who need it. Your edit was not a "suggestion", but an attempted justification for dumping misleading data into entries on Wikidata. Your "suggestion" doesn't represent any uncertainty to the eventual readers, because they never see the description field. Just who do you think are our users? What is harmful is allowing third-party re-users of Wikidata to read that a particular condition has a certain "drug used for treatment", when that drug has no evidence of effectiveness in treatment whatsoever. --RexxS (talk) 21:09, 28 February 2017 (UTC)