Don’t Use PubMed as a Journal Whitelist

I frequently met colleagues in pharma industries, which did not even knew that PubMed is a limited source for bio-medical publications. What a pity … and what a risk for the departments those colleagues are working in.

This post by “Scholarly Open Access”, a quite proficient blog on scientific literature sources, points out the tiny little (but quite important) difference between free PubMed and its professional sibling Medline.

“Product Literature Database” (PLD) … what the heck is this?

zur deutschen VersionPharmaceutical companies have an inevitable need for regular if not permanent analysis of literature published on their products. They are not only legally obligated with regards to pharmacovigilance (e.g. processing of any undesirable drug effect reported somewhere) and medical information (e.g. answering product inquires by practitioners and pharmacists). Beyond that, product reports within the scientific literature are a full treasure of real-world product behavior findings which support marketing, competitor intelligence, product innovation, and more.

But we are talking about giant pool of millions of publications within thousands of scientific journals which is growing every single day. Ad-hoc searching and analyzing costs reasonable amounts of money:

  1. Professional literature databases are quite expensive with a market dominated by few providers only, which sometimes behave like monopolists.
    (And to make an early clear statement: no, PubMed clearly does not qualify, especially due to its lack of comprehensiveness and poorly finished content.)
  2. It is very time-consuming to find precise and significant answers by searching and analyzing scientific literature. And the cheaper the source, the more (expensive) human workload is needed.


The solution

So-called “product literature databases” (PLDs) or “corporate literature databases” deliver the knowledge about what has been published about your own products … much more efficient than highly redundant and multiplied individual ad-hoc literature searches.

PLDs are sort of subsets of the worldwide literature, including only publications which mention at least one product of the company. Typically, they are filled by automatic search agents (search profiles) or by feeds delivered by database providers.

Well-designed PLDs also provide mechanisms for rating publications, annotating information and signaling predefined events.


External PLD providers

UK-based Pi2 solutions Ltd. is an established vendor for customized pharma and biotech PLD solutions and was acquired by ProQuest Dialog this summer. Pi2 traditionally supports Pfizer, and since 2009/2010 also Wyeth (who had worked with OvidSP® for ad-hoc literature research before). A 2013 poster presented at the 9th Annual Meeting of the International Society for Medical Publication Professionals might give some information on general approaches. Beyond that are no public information available regarding Pi2’s market success or market share, and I am quite curious about impact of the new collaboration with ProQuest.

Other potential providers of PLD solutions are major B2B specialist information database and service providers, like Reed Elsevier, Thomson Reuters and Wolters Kluwer, who are factually dominating the mass market for literature raw data. Particularly Elsevier has already shown strong interest in providing more customized and mature services to industry clients. They quite recently build a kind of customized product literature service for Sanofi’s pharmacovigilance by combining their database content with the QUOSA literature management software.


In-house PLDs

The Novartis pharma group had their own internal PLD since the late 60’s, called “eNova”. This solution has been the most mature and significant PLD I have ever seen. Novartis not only collected literature on their products, they also applied a kind of in-depth ‘digestion’ of reported product findings and clinical data. As a result, the PLD was able to very precisely answer questions on any aspect of real-world drug behavior at the push of a button. “eNova” was finally discontinued and shut-off by Novartis end of 2013, despite the fact that internal analysis had shown substantial positive impact on productivity and individual time savings for product related literature research & analysis of 93% and more.

Roche once also had an internal PLD similar to “eNova”, which was shut-down a couple of years ago already. As a “side effect” corresponding product literature research & analysis activities and workload were distributed across the organisation. For example, each national affiliate had to substitute the service by an own solution to continue mandatory MedInfo deliveries and to comply with regulatory expectations. It goes without saying that this split-up of different solutions and approaches did not really result in an overall productivity increase nor in overall cost savings.

A little later, after negative effects became more and more evident and clear, Roche tried to reactivate their in-house PLD. But unfortunately the reintroduction failed as 2-digit million CHF investments were needed but not provided.

By the way, much more money than continuing the Roche in-house PLD would have costed.


Why do PLDs have such a poor standing?

Watching the developments at Novartis and Roche, one automatically ends up with the question for what reason their PLDs were shut-off … despite obvious downsides for the enterprises? Actually, there are some dependencies and basic conditions for the reasonable operation of an in-house PLD. And those dependencies and basic conditions are sometimes contrary to currently practiced management paradigms.

  1. PLDs need long-term view and sustainability. But currently lived management approaches in pharma enterprises are more near-term and quite volatile. Without a strategic integration, a PLD is always in danger falling victim to a short-term budget or structural decision, similar to other internal services. But in contrast to other internal services, such a decision is much more fatal for a PLD, as it cannot be just shut-on again once negative side effects become obvious.
  2. PLDs save money in the in the field. What a fatal dilemma. As a central service, PLD costs are budgeted with a (global) business unit, which does not necessarily benefit itself by the service. On the other hand, corresponding cost savings, e.g. by higher productivity, cut costs for external providers, synergy effects, etc. pp. are effective within the whole organisation, with completely different business units. As a result, budget and benefit are organizationally decoupled. Overall, the enterprise has a tremendous advantage and cost savings by the PLD. But unfortunately this full picture view is less and less shared with individual budget decisions.
  3. PLDs are IT, they are not. Effective PLDs certainly need a powerful IT infrastructure, databases, and more. Unfortunately, this bears the risk to rashly assign PLDs to the IT department. To be very honest, to my opinion that is the wrong place. I also need a PC to work efficiently and powerful, but I am far away from being a software engineer. For me, a clever PLD implementation includes a clear localization within business, at least within a well-established linking function between business and IT.
  4. PLDs are strong as central functions. Only then they can fully exert resulting synergistic effects. In contrast, there seem to be frequent “waves” with in pharmaceutical enterprises to distribute tasks over the whole organisation. The underlying thought is “we save money at Global (e.g. for the PLD), and the work is shared by all”. Funny thought … unfortunately with fatal implications on productivity of associates.
  5. PLDs are designed by information experts for information experts. True, and there are historical reasons. But this approach does not fit to todays reality within pharmaceutical enterprises anymore. During the past 10-15 years, pharma has consequently reduced the number of educated information professionals. As a result, todays users of PLDs are more and more subject matter experts (e.g. medics) without explicit expertise in using professional information tools. And honestly spoken, I so far have not seen many PLDs which serve those new user groups regarding usability adequately.



An in-house PLD – cleverly designed and implemented – is able to reliably cover the need to know that has been published about a company’s own products. It also prevents troubles with regulatory expectations and authorities, and increases productivity at once.

But “cleverly designed and implemented” also includes a long-term strategic integration within the enterprise as well as a reasonable degree of independence from short-term decisions and tactical changes. Any short-term shut-down of an established in-house PLD bears the risk to create hidden but substantial costs. And in all known cases it had been an irreversible back to zero.

Currently, one of the biggest challenges of PLDs is, to give medics and other non information professionals efficient access to product answers, especially by more productive and intuitive user interfaces. Success will be result of votes by the feet … resp. by the keyboards.

Checklist on pitfalls with bibliographic searches

With a so-called “bibliographic search” you are looking for the abstract or full-text of a scientific publication. This means, you already have at least some citation information on the publication, like author name(s), publication year, title, journal name, volume#, issue#, and/or page#.

There are some known traps and pitfalls with bibliographic searches, that I would like to share with you.

6 pitfalls for bibliographic searches …

1. Always assume a typo

Generally assume typos in either the database record of the publication, or your notes, or the original publication.

2. Do not use special characters

If the known publication title you would like to search for includes any special characters, like hyphens, colons, commas, semicolons, brackets, Greek symbols and so on and so forth, use only those parts of the title for your search which do not include any of those.

“Oral fingolimod (FTY720) in relapsingremitting multiple sclerosis (RRMS): 2Year αData efficacy results; the phase III FREEDOMS I trial”
should be searched as
“Oral fingolimod” AND “multiple sclerosis” AND “efficacy results” AND “phase III FREEDOMS I trial”

However, some literature databases handle brackets, hyphens & co. quite well. When they are phrased.

“Oral fingolimod (FTY720) in relapsingremitting multiple sclerosis (RRMS)”

By the way. In literature databases non-Latin characters (Greek symbols e.g.) are normally translated to the corresponding Latin character (α -> a) or written out (α -> alpha). Similar for local characters, like the French accents (à, á, â) that most likely will be used just as “a”.

3. Do not trust publication titles

Even if the known title of a publication can be the quickest way to identify the reference, always doubt it. If you do not find anything with it, it does not necessarily mean that the publication is not there. The source, where you have it from might have included an error, or there could be an accidental typo.

Also think about the already mentioned different notations for Greek symbols, special characters, numbers (3, III, three) or abbreviations as well as differently use blanks, that are all potential variations resp. sources of mismatches.

If you cannot pass on searching the title, the solution might be to not use the complete but just a fragment of it, which seems to be more valid (= less opportunities for variations) .

“Oral fingolimod (FTY720) in relapsingremitting multiple sclerosis (RRMS): 2-Year αData efficacy results; the phase III FREEDOMS I trial”
could be searched in the title field as
“Oral fingolimod (FTY720) in relapsingremitting multiple sclerosis”

4. Use author’s last names only

For “Jean-Paul Sartre” you would find the following alternative writings in scientific literature databases:

  • Sartre Jean-Paul
  • Sartre Jean Paul
  • Sartre, Jean-Paul
  • Sartre; Jean-Paul
  • Sartre, J-P
  • Sartre, J.-P.
  • Sartre J.-P.
  • Sartre JP
  • Sartre, JP
  • Sartre, Jean
  • Sartre Jean P.
  • Sartre, Jean-P.

Sometimes you find even in a single database notation variations of the same author’s name. So, the only stable and consistent values are the author’s last names.

5. Use sparse search values only

If you know the full citation data, a search with the first authors last name, the publication year and the first page number alone in most cases will be sufficient and bears minimum risk only for mistakes and typos.


  • author last name and starting page number (and publication year if necessary) might be sufficient already
  • volume, starting page and publication year might be sufficient already
  • volume, issue and starting page number might be sufficient already

6. Avoid journal names

Search for journal names only if there is no other opportunity to identify.

But keep in mind that there might be variations of the journal name like “Proceedings of the National Association of Science”, “Proc Nat Assoc Sci”, and “PNAS”. Better limit by clearer values, like volume number, issue number, publication year, first page number … without using the journal name.

So, and now just enjoy your next search! Try those 6 simple rules, and failure should be history.