PVmap™ of the Week

ProSanos has initiated a program to publicly provide a limited set of PVmaps™ generated from the FDA's Adverse Event database. A different map will be posted each week focusing on a drug and adverse event combination that is a current topic of discussion within the industry or in the published literature. For more information about PVmaps or the PVmap of the Week program, .

Progressive Multifocal Leukoencephalopathy as an Adverse Event (5/7/2007)
A recent MedScape^® Hematology-Oncology article discusses an FDA alert with regard to spontaneous, fatal progressive multifocal leukoencephalopathy (PML) in patients taking rituximab for the treatment of systemic lupus erythematosus.⁴¹

PML is a potentially fatal demyelinating disease caused by activation of human polyomavirus JC in immunosuppressed patients. Infection with this virus is asymptomatic in individuals with normal immune function. PML has been seen with AIDS and is a rare consequence of immunosuppression in transplant patients and others. A number of chronic diseases such as lupus, multiple sclerosis, and rheumatoid arthritis are now being treated with immunomodulators, and new, uniquely-targeted immunomodulators are being brought to market for these indications and others. This has raised the level of concern regarding PML as an adverse consequence of treatment with these drugs.

In this weeks PVmap of the Week we illustrate the capabilities (and some limitations) of data mining by performing a top-to-bottom review of the list of drugs that are statistically associated with PML. We begin with an Event-focused PVmap for progressive multifocal leukoencephalopathy. (Note that a more complete analysis would include terms such as JC virus infection as well.):

The complete list of drugs that are statistically associated with this adverse event is shown below:

Most of these are antiviral agents used against HIV. For these drugs, PML is considered a consequence of the underlying disease, rather than the treatment. Eliminating the drugs whose primary indication is in HIV disease, we are left with the following list, in order by PVmaps signal strength:

• saline infusion
• probenecid
• bactrim
• cyclophosphamide
• vincristine
• rituxan
• fludarabine phosphate
• fludara
• septrim
• pentamidine isethionate
• prograf
• azathioprine
• co-trimoxazole
• natalizumab
• ethambutol hcl

The presence of Saline infusion on this list is clearly an anomaly. The nature of the anomaly can be seen by drilling down to the individual safety reports:

We see six identical reports of what is apparently a single case. This is an occasional unfortunate occurrence in the AERS database. PVmaps software contains features for automatic duplicate detection which can be selected by the user. This feature examines the age, gender, and other demographic features of listed patients to identify and eliminate high-probability duplicates. In this particular instance, the required demographic information is missing, as is evident below. When this occurs, the duplicate-detection feature errs on the side of caution, and includes this signal in the list of results.

Note that this heavily-duplicated case included probenecid as well and is responsible for the association of this drug with PML.

For the next drug we encounter, bactrim, we can use a Potential Interactions Map to show that, when associated with PML, this drug generally appears in the context of HIV treatment:

For the small numbers of cases involved, it can be equally effective to visually examine the individual safety report details, an excerpt of which is shown below:

Further Potential-Interactions PVmaps show that cyclophosphamide, vincristine, and rituximab show strong statistical interrelationships in cases of PML. These drugs are heavily co-prescribed as components of a lymphoma chemotherapy regimen known as "R-CHOP". Fludarabine ("fludara") with rituximab is the basis for another lymphoma chemotherapy regimen. The numbers of PML cases are consistent with cyclophosphamide, vincristine, and fludarabine being mainly "bystanders" with rituximab, though there are very small numbers of PML cases for those drugs where rituximab does not appear.

Potential-Interactions PVmaps show that Septrim, pentamidine, co-trimoxazole, and ethambutol are associated with HIV treatments, similar to "bactrim" described above.

From its short list of six ISRs, "Prograf" does not appear to be a bystander and it does not fall into the HIV-associated category. "Prograf" (tacrolimus) is an immunosuppressive agent, but has also been associated with a very rare reversible leukoencephalopathy that is drug-related and distinct from PML.⁴² To investigate this drug-event combination further, one would need to examine case reports carefully to distinguish the two entities.

Azathioprine is another immunosuppressive agent with six PML cases associated with it. In one PML case, it is associated with natalizumab, but in the others, it is not and appears only with "conventional" immunosuppressive agents.

The final drug on the list is natalizumab, with four PML cases. The relationship between natalizumab and PML has been established and is discussed at the FDA web site in the context of the approval, withdrawal, and subsequent reintroduction of this treatment for multiple sclerosis.⁴³

This column has guided you through the complete investigation of a complex real-world adverse event using PVmaps as a productivity tool. We have encountered reporting anomalies (duplication) and bystander effects. These were identified quickly and effectively. All of the maps discussed here were generated interactively, in real time, over the course of one day. This included time spent at the FDA web site, reading labels for the drugs involved, and searching the medical literature with PubMed. In many cases, the output of one map led us to generate another one. For instance, the strong association between rituximab and cyclophosphamide necessitated the generation of a Co-Prescribed Drugs PVmap for rituximab to examine the possibility of a drug interaction. In many situations, we were dealing with small numbers of cases and we had the choice of either generating a Potential Interactions PVmap or drilling down into the cases themselves and visually inspecting the lists of drugs for possible bystander effects and interactions. PVmaps lets you choose either of these approaches, and, if you see something suspicious with one approach, you can easily confirm it with the other.

PML is a serious adverse event and any individual report of this condition in connection with a drug not previously associated with it would receive close scrutiny. However, it is not appropriate to sound a public alarm for every individual case of a rare, severe event, particularly when a patient is using many pharmaceutical products and sorting out the causative one (or ones) is a complex undertaking. Risk/benefit ratio must be considered: false safety alarms can deprive patients of important treatments and can therefore do more harm than good. One of the benefits of a safety data mining program is that it can set a rational threshold for when to escalate the level of concern. In addition, a "top-to-bottom" review of all signals of disproportionate reporting can provide reasonable assurance that no ominous patterns have been missed or ignored. Such a review can be done for a drug or for an adverse event, as shown here. With PVmaps as a tool, systematic reviews such as this are a practical and informative undertaking that can serve to safeguard both the public and the pharmaceutical manufacturer alike.

Event-focused PVmaps
The PVmap shown in the case study above is an Event-focused PVmap that allows you to visualize which drugs are most highly associated with a particular adverse event (rather than the other way around). In this case, the adverse event is the PML and the red dots represent drugs reported in the AERS database to be associated with PML. On the horizontal axis of this graph is the reporting ratio, which compares the number of times that a drug is reported with the specified adverse event to the number expected due to chance alone. The vertical axis expresses the statistical significance of the finding. Dots above the horizontal blue line and to the right of the vertical blue line represent "significant signals". The drugs with the strongest association with PML appear at the top and to the right on the PVmap.

Potential Interactions PVmaps
Above is a Potential Interactions PVmap that allows you to visualize what concomitant drugs are significantly associated with a specified drug/adverse event combination. In this case, the drug/adverse event combination is the drug "bactrim" reported with PML. The red dots on the first map represent concomitant drugs in use when the drug / adverse event combination bactrim / PML occurred. On the horizontal axis of this graph is the reporting ratio, which compares the use of the concomitant drug during bactrim / PML with the use of the concomitant drug expected due to chance alone. The vertical axis expresses the statistical significance of the finding. Dots above the horizontal blue line and to the right of the vertical blue line represent "significant signals". The concomitant drugs that are most highly associated with the drug/event combination of interest appear at the top and to the right of the PVmap.

Sponsor companies have used ProSanos PVMaps for multiple therapeutic areas. To learn more about PVMaps projects in your therapeutic area or indication, please .

Disclaimers

1. ProSanos is not affiliated with the authors of cited references and this article does not imply endorsement of their findings, content, or offerings.
2. Potential risks highlighted by drug safety analysis must be balanced against the clinical benefit attained by the use of a pharmaceutical product in a given clinical situation. Nothing in these analyses is intended to influence the practice of medicine, nor to weigh the benefits of one product over another.
3. Whether the reporting ratio of an adverse event is high enough to influence the decision to use a given product or products can only be determined by a complete analysis of the benefits, risks, and therapeutic alternatives.
4. Use of the publicly available FDA AERS data does not imply endorsement or agreement of the findings by the FDA Center for Drug Evaluation and Research.
5. There are many factors that can influence how the adverse events are reported in the AERS database and may impact the resulting safety signal. These include but are not limited to: publicity and media attention, litigation, length of time drug is on the market, whether the event in question has been previously attributed to the drug, the source of the report, etc.
6. AERS data must often be "cleaned" prior to analysis. This process may include de-duplication, reconciliation of misspelled product names, mapping of adverse events terms, and other manipulations which could introduce bias into the analysis.
7. PVmaps has been evaluated as a safety signal investigation tool for over two years.

References

41. Fox RI. FDA Alert for Rituximab in Patients With Systemic Lupus Erythematosus. Accessed on the Internet at http://www.medscape.com/viewarticle/551351_1, May 2, 2007.
42. Singh N, Bonham A, Fukui M. Immunosuppressive-associated leukoencephalopathy in organ transplant recipients. Transplantation. 2000 Feb 27;69(4):467-72.
43. Natalizumab (marketed as Tysabri) Information. Accessed on the Internet at http://www.fda.gov/cder/drug/infopage/natalizumab/, 2 May 2007.