Deciphering the adverse event reporting system: the FDA's database of side effects reported in association with drugs contains a wealth of information for plaintiff attorneys. But finding and using what you need takes considerable skill and expertise.

What's my cholesterol level? Most people ask this question when they have an annual physical, and for some, the answer means adding a new drug to their daily routine, a statin drug like Lipitor or Zocor.

But like most drugs, statins have risks as well as benefits. One of those risks is damage to muscle tissue, ranging from myopathy (weakening and wasting of muscles) to an especially severe form of myopathy known as rhabdomyolyis, or "rhabdo." In some rhabdo patients, large protein molecules from destroyed muscle tissue enter the bloodstream and clog up the kidneys, leading to kidney failure and even death.

All statins can cause myopathy and rhabdo, but Baycol (cerivastatin) is particularly deadly. On August 8, 2001, Baycol was pulled off the market because far more rhabdo-related deaths were reported for it than for other statins.

Six months later, three FDA scientists published a letter in the New England Journal of Medicine explaining how their analysis of data in the agency's Adverse Event Reporting System (AERS) showed that the "rate of fatal rhabdomyolysis associated with cerivastatin therapy ... is 16 to 80 times as high as the rates for any other statin." (1)

The authors said that some of the difference appeared to result from a particularly vicious interaction between Baycol and gemfibrozil, another cholesterol drug sometimes prescribed in combination with a statin. Even by itself, however, in so-called monotherapy, Baycol's fatal rhabdo rate was "10 to 50 times as high as the rates associated with the other statins." (2)

Additional analysis of the AERS data has revealed that when both fatal and nonfatal rhabdo cases are considered, Baycol patients were 16 times as likely to develop the disease as patients taking other statin drugs. (3) And a recently published study based on more detailed and refined data concluded that the "risk of rhabdomyolysis with cerivastatin monotherapy was tenfold greater than with other statins." (4)

AERS data has clearly played a prominent role in the story of Baycol's demise and in the ensuing litigation about the drug. And AERS can provide plaintiff lawyers in other pharmaceutical product cases with a treasure trove of information that can help establish credibility and causation. But for lawyers to use this information effectively, they must understand it and explain their findings clearly to a jury.

History and anatomy of AERS

Many people believe--incorrectly--that the FDA review-and-approval process is so thorough and rigorous that it effectively prevents unsafe drugs from ever reaching the pharmacist's shelf. But recent news reports about the deadly side effects of approved drugs, like Vioxx, show why nothing could be further from the truth. As two scientists noted in a 2002 article on statins and rhabdomyolysis,

   One of the limitations of the premarketing
   phase of drug development is that rare
   adverse events go undetected, primarily
   due to the small sample size involved, the
   short duration of the studies, and the controlled
   setting of the trials. For example,
   to detect an adverse event that occurs in 1
   out of 1,000 patients with a 95 percent confidence
   level, 3,000 patients need to receive
   the drug. (5)

Because the number of subjects in clinical trials typically ranges from several hundred to 3,000, it is not surprising that many problems are missed. (6) A 2002 review found that at least 10 percent of drugs approved between 1975 and 1999 later either required a black-box warning or were removed from the market. (7)

To address this problem, the FDA and regulatory agencies in other countries now track adverse drug reactions after a drug is approved. Most of these systems trace their roots to the infamous thalidomide disaster in 1961. In December of that year, an Australian physician wrote a letter to the editor of The Lancet noting a strong association between limb-deformity birth defects and the mothers' use of the antinausea drug while pregnant. (8)

In 1962, amendments to the Food, Drug, and Cosmetic Act of 1938 "required pharmaceutical manufacturers to report [adverse events] to the FDAfor any of their products [for which a new drug application had been submitted]." (9) Today, the World Health Organization and more than 60 countries maintain reporting systems. (10) In this country, the system is AERS.

Accessing and analyzing the database properly requires considerable understanding and sophistication. (11) The current database is the 1997 successor to earlier systems, a technical wrinkle that can create some minor difficulties if you need pre-1998 information.

The information in AERS comes from reports submitted to the FDA. Anyone aware of an adverse drug reaction can submit a report, but there is no submission requirement, and it is generally recognized that because the system is completely voluntary, only a small percentage of adverse events actually gets reported. Doctors and other health care professionals (for example, pharmacists) have no obligation to report at all.

If a report is made, there are two possible routes--direct to the FDA or via a drug's manufacturer. More than 90 percent of the reports come through manufacturers. (12)

Once a manufacturer receives or becomes aware of a report, it usually must be submitted to the FDA on something called a MedWatch Form 3500A, (13) although recent changes now allow electronic submission using a format known as E2B. (14) This change is part of a worldwide attempt to harmonize adverse-event-reporting methodologies to facilitate global analyses. Currently, every country determines on its own how to collect and report adverse events, but within a few years, all manufacturers will be required to submit reports in electronic form. (15)

Under FDA regulations, whether and when a report is sent in depends on several factors. Is the event serious or nonserious? (16) Expected (that is, already included in the drug's labeling) or unexpected? (17) Is it foreign or domestic? All reports of adverse events that are both serious and unexpected must be reported within 15 days of receipt by a manufacturer, whether they are foreign or domestic. (18) Other adverse events are reported at regular periodic intervals, (19) except that nonserious or expected foreign adverse events do not have to be included in periodic reports. Thus, a report of serious and expected (that is, labeled) foreign event need not be sent to the FDA. (20)

After the FDA receives a report, the information is reviewed before it gets added to the AERS database. To provide some degree of standardization, the agency goes through several steps. The most important is assigning a standard set of terms contained in a dictionary known as MedDRA, a hierarchical classification system consisting of four primary levels of specificity and over 20,000 terms. (21)

To illustrate, rhabdomyolysis is a preferred term (the most specific level). The high- level group term is "muscular disorders," and the system organ class is "musculoskeletal and connective tissue disorders." Manufacturers do not have to use MedDRA internally, but the FDA makes an effort to ensure that the AERS information is compatible with the MedDRAsystem.

The assignment of MedDRA terms by the FDA is important because it largely eliminates one of the potential limitations of adverse event reporting: Companies may assign different terms to the same type of event. For example, one company might code a statin adverse event as "muscle pain" and another company, selling a different statin, might code the same adverse event symptoms as "rhabdomyolysis." The agency's skilled medical coders use the MedDRA dictionary to determine what goes into AERS. Often, they select terms different from those the manufacturer used.

In the case of a serious report, the manufacturer is obligated to follow up with the original reporter and then to forward any additional information to the FDA. (22) Thus, the AERS database will contain the initial version of a report as well as revisions, and if the revisions are not taken into account, overcounting of adverse events almost certainly will occur.

To avoid this problem, it is best to apply a concept known as the "last best case," which eliminates data from all but the latest follow-up of a reported event while continuing to use the date of the event's first appearance in a report or fob low-up report. This procedure ensures that an event is described as accurately as possible and that any analysis nonetheless reflects when the event actually was reported.

To illustrate, suppose that the first report for an adverse even t for Drug A is received on 6/15/01. It gives certain symptoms, but it does not establish the occurrence of Disease X. Then on 8/17/01, a follow-up report is filed that includes additional diagnostic test results establishing that the adverse event is indeed a case of Disease X.

Two more reports, with still more details, are filed on 10/23/01 and 12/28/01. The last report--filed on 12/98/01--would be the sole source of information about the nature of the adverse event, but the date of the report for analytical purposes relating to Disease X would be 8/17/01, the date on which the disease first appeared in the AERS system for this particular adverse event.

Another potential problem with AERS is the listing of some drugs under several names. The AERS database contains about 150,000 different drug names, but some entries are either misspellings of a drug's correct name or different names for the same drug. For example, entries for "Cerivastatin," "Lipobay" (the name underwhich Baycol was sold in Europe), "Baycol (Cerivastatin)," and "Baycol 800mcg"would have to be counted along with those under "Baycol" to get a total number of adverse event reports involving the drug.

Pluses and minuses

AERS data has several advantages over clinical trial data, but it also suffers from several disadvantages, and both the positive and the negative features must be understood to use AERS effectively in litigation. On the plus side, AERS monitors the entire population of patients taking a drug, and it does so at a relatively modest cost. (23) In other words, the data reflects what happens during real-world use of the drug, which is quite different from the limited and controlled environment of a clinical study.

On the negative side, the data's uses may be limited by the voluntary nature of the system and lack of control over the criteria that doctors use for submitting a report. Even for serious adverse reactions, it has been estimated that rarely more than 10 percent get reported. (24) Moreover, doctors may use different diagnostic criteria, and their reasons for attributing causality (and hence reporting) may differ as well. This can make the quality of reports uneven. (25)

Also, publicity about suspected adverse events associated with a drug may boost reporting, skewing results for a drug that has gained notoriety. (26) The length of time a drug has been on the market also may be a factor. It is sometimes said that physicians tend to submit reports about a drug more frequently at the end of a drug's second year of marketing. (27) Whether such a relationship exists for the last drug to go on the market in a class of drugs is an open question. The focus maybe more on the first drug in the class and much less on subsequent drugs.

Finally, differences in the way drug companies seek out and classify adverse events may affect the validity of comparisons among drugs based on AERS data. (28) For example, how much detailing (direct personal marketing to physicians and medical practice groups) does one company do compared with another, and what kind of instructions are the detailers given with regard to asking questions and seeking out adverse reactions? How do different companies reach decisions about what counts as serious or not? Is one company more aggressive than another about listing adverse events in its drug's labeling, thus taking some adverse events out of the 15-day priority-reporting category?

All that said, adverse event data can be extremely useful. A recent article published by four doctors who have served as plaintiff experts in the Baycol litigation makes clear that suspected adverse drug reaction data is "one source, sometimes the only source, of timely information about the adverse events associated with recently marketed drugs." (29)

Using AERS to detect signals

For the most part, AERS data is used to determine if there was a "signal" of a problem warranting a pharmacoepidemiologic study. (30) Sometimes it is possible to go beyond signal detection to determine that one drug carries a higher risk of an adverse event than others in its class or that a drug in fact causes a specific disease, but establishing when a signal was or should have been seen can itself be critical in litigation.

Almost anything that raises serious suspicion of a problem is a signal. For rare and unexpected adverse reactions, it "has been suggested that a temporal relationship ... coupled with positive dechallenge and rechallenge, can make isolated reports conclusive as to a product-event association." (31) In other words, if an adverse event goes away when use of a drug is stopped and returns when a person resumes taking the drug, an association between use of the drug and the adverse event can be inferred. Indeed, if there also is evidence of "biological plausibility and reasonable strength of association," the relationship may be deemed causal. (32)

Other signals that can warrant further investigation include:

* an observed increase in the severity of a labeled event

* more than a small number of reports of an adverse event that was thought to be rare

* new interactions with drugs, food, or dietary supplements 

* the appearance of an adverse event in a population not thought to be at risk

* concerns about the way a product is understood or used by patients. (33)

When questions arise about the safety of one drug in a class compared with another in the same class--the situation with Baycol and the other statins--the signal might be a higher adverse event reporting rate for the suspect drug. Of course, the reporting rate will not be a measure of actual incidence because the majority of adverse events are not reported.

Determining even reporting rates can be difficult, however. One hundred cases of Disease X per year with Drug A compared with 25 cases with Drug B does not necessarily mean A is worse than B. What if only 100,000 people used B and 1 million used A? That would mean a rate of 10 per 100,000 people for A but 25 per 100,000 people for B.

Then again, what if the average duration of use for B was two and a half months, and the average duration for A was one month? In that case, on a "person-time" basis--which epidemiologists prefer--the two would be the same. (34) The rate for both would be 10 per 100,000 patient-months (100 divided by 1,000,000 equals 25 divided by 250,000). Such details are not trivial and must be understood if AERS data is to be used effectively as evidence.

Still another potential problem is the determination of denominators for the kind of comparison described in the previous paragraph. If person or person-time data is unavailable, it may be possible to use surrogates, "such as numbers of prescriptions or kilograms of product sold." (35) Again, however, care must be taken. For example, differences in the average length of prescriptions for the drugs being compared may affect the results, biasing them against drugs with longer average prescriptions. In the case of kilograms sold, the relative potency of the drugs must be taken into consideration.

Proportional reporting ratio

An innovative signal-detection technique called proportional reporting ratio (PRR) analysis does away with the problem of finding denominator data and also with possible biases due to the higher rate of reporting during a drug's early years on the market. (36) Using PRR, one looks at all reports received for a drug and determines what percentage contains a specific adverse event term. That percentage is the proportional reporting rate for that drug. A similar calculation is done to determine what percentage of adverse event reports for all other drugs contains the adverse event term. Then the PRR--the ratio of the proportional reporting rates--is calculated.

To give a simple hypothetical example, suppose that 25 percent of all adverse event reports for Drug A are for Disease X. And suppose that Disease X appears in only 5 percent of the adverse event reports for all other drugs. The PRR would be 5 (25 percent divided by 5 percent). This does not necessarily mean Disease X is five times more likely to occur with Drug A than with other drugs. The PRR should not be confused with the more commonly cited relative risk (RR), which is based on a ratio of incidence rates. (37)

Nonetheless, "the higher the PRR, the greater the strength of the signal." (38) If one wants to see if a drug is different from other drugs in the same class, the PRR can be slightly modified by focusing only on the class. To continue with the hypothetical, suppose there are 1,000 re ports for Drug A, and 250 of them include the dreaded Disease X. That would be 25 percent.

Now, if 300 of 3,000 reports (10 percent) for other drugs in the class included Disease X, one might conclude that the disease is more strongly associated with Drug A than the others. Moreover, while not quantifying the strength of the association (that would require an RR), the PRR would roughly indicate its magnitude. In the hypothetical, the PRR for Drug A versus other drugs in its class would be 2.5 (25 percent divided by 10 percent). Note that this PRR is not as high as the PRR for Drug A against all other drugs (not just the class). This indicates Disease X is associated with the class, but especially associated with Drug A.

To do PRR comparisons in a meaningful manner, it is critical to eliminate as much as possible the differences in reporting procedures between manufacturers. For example, the FDA allows companies to request waivers from the regulation requiring periodic submission of nonserious reports. Some companies request the waiver, and some do not. To account for this possibility, nonserious reports should be eliminated from PRR comparisons.

As a matter of good practice, reports also should be culled where the drug in question was not considered suspect for the event (for example, where the report was submitted for another drug but the patient also was taking the suspect drug). The same applies for reports in which there has been "polytherapy" with more than one member of the class of drugs at issue.

An analysis of how the PRR evolves over time can be used to see how far back a signal may have developed. This evolution can be displayed effectively with a graph. Table 1 lists data by quarter for reports of Disease X with Drug A and with all other drugs, and Figure 1 shows this information graphically. (See page 37.)

In this hypothetical, Drug A came on the market during the first quarter of 2000, and the data is shown quarterly for 2000 through 2002. Once reports started coming in, it became clear, certainly by the third quarter of the first year, that Drug A was generally more often associated with Disease X than other drugs. Perhaps extra cases of Disease X were expected, but if not, the PRR provides a signal that should be investigated.

When comparing drugs within a class, it may be helpful to graph the evolution of the proportional reporting rate rather than the PRR. This approach allows for more easily understood visual comparison of information about each drug in the class. Table 2 contains data on three hypothetical drugs in a hypothetical class, and Figure 2 is the graph of this data. (See page 38.)

Note that drug A went on the market in the first quarter of 2000, Drug B had already been on the market for some time as of the first quarter of 2000, and Drug C did not come onto the market until the fourth quarter of 2000. Also note how much higher the rate is for Drug A than for either Drug B or Drug C, a signal that Drug A entails greater risk for Disease X.

Legal implications of signal detection

Determining when a signal first appeared can be crucial to your effort to establish when a drug company defendant knew or should have known of a problem. Sometimes AERS data for a suspect drug will reveal a signal, as when a completely unexpected and rare event occurs. Because most of this information comes through the drug company (and if not, would be sent to the company per FDA rules), there is no way a defendant can hide from its logical and legal consequences.

A signal of a drug-drug interaction also may be evident solely from a defendant's data. The data in Table 3 and the graph in Figure 3 illustrate this point and show how a PRR analysis can be used when the comparison is not with all other drugs. (See page 40.) The table shows that when Drug A is used in combination with Drug B, the association with Disease X becomes much stronger. Assuming that Drug B is not by itself associated with Disease X, this relationship indicates some kind of interaction is occurring. The data necessary to do the kind of analysis shown in Figure 3 can be extracted from the AERS database, but the manufacturer of Drug A would have all the reports in its own database and should not be able to plead it was ignorant of them.

Because AERS data on all drugs is readily and cheaply available, one can also argue persuasively that even if a signal required comparison with other drugs, the manufacturer should have caught it. Missing a signal may constitute negligent "pharmacovigilance." It may also violate FDA regulations, which require companies selling FDA-approved products to "promptly review all adverse drug experience information obtained or otherwise received ... [and] develop written procedures for the surveillance, receipt, evaluation, and reporting of postmarketing adverse drug experiences to FDA." (39)

Going from a signal to proof of causation or to proof that a drug entails greater risk than others in its class is a more difficult issue than establishing the existence of a signal. The FDA discourages rigorous comparisons based on AERS data. (40) However, under some circumstances, adverse event data can establish more than just a signal. Causation for very rare and unexpected adverse events can in some circumstances be proved with just a few reports. For example, adverse event reports conclusively established that the nonsteroidal anti-inflammatory drug suprofen caused flank pain syndrome. (41)

Also, when differences between drugs are big enough, adverse event reporting rates, by themselves, may establish causation. As the four Baycol plaintiffs' experts have written, "although a number of potential biases make [them] difficult to interpret, [adverse event reporting rates] for cerivastatin [in comparison with other statins] were so [great] that few alternative explanations are credible, and an inference of cause and effect seems warranted." (42)

Even if disparities are not so large, when adverse event data is consistent across multiple databases and with other kinds of information, it can be an important element in the totality of the evidence against a drug. Consequently, in most pharmaceutical cases, research should include an analysis of the information in AERS. Often the effort will be rewarded with new insights and more powerful evidence of both liability and causation, making it easier to obtain justice for clients injured by pharmaceutical products.

Table 1: Data for PRR Comparison of Disease X for Drug
A Versus All Other Drugs

                                                        Tot. Adv.
Quarter    Tot. Adv.      Drug A Adv.    % of Drug A      Event
Ending    Event Repts.   Event Repts.    Adv. Event      Repts.
           for Drug A     with Dis. X    Repts. With    for All
                                           Dis. X      Other Drugs

Dec-99         0               0            0.00         10,000
Mar-00         25              2            8.00         10,000
Jun-00         75             12            16.00        10,500
Sep-00        100             30            30.00        12,500
Dec-00        120             37            30.8         12,000
Mar-01        125             29            23.2         11,500
Jun-01        120             31            25.8         13,000
Sep-01        135             32            23.7         13,500
Dec-01        130             35            26.9         12,500
Mar-02        120             30            25.0         14,000
Jun-02        140             35            25.0         14,000
Sep-02        135             34            25.2         14,500
Dec-02        140             32            22.9         14,500

Quarter    Other Drug     % of Other         PRR
Ending     Adv. Event     Drug Adv.
          Repts. With    Event Repts.
             Dis. X       with Dis. X

Dec-99        500             5.0            0.0
Mar-00        520             5.2            1.5
Jun-00        520             5.0            3.2
Sep-00        650             5.2            5.8
Dec-00        580             4.8            6.4
Mar-01        570             5.0            4.7
Jun-01        650             5.0            5.2
Sep-01        650             4.8            4.9
Dec-01        630             5.0            5.3
Mar-02        700             5.0            5.0
Jun-02        710             5.1            4.9
Sep-02        710             4.9            5.1
Dec-02        730             5.0            4.5

Table 2: Data for Proportional Reporting Rate Comparison
of Disease X for Drug A, Drug B & Drug C

Quarter   Tot. Adv.    Drug A Adv.   % of Drug A
Ending      Event        Event       Adv. Event
          Repts. for   Repts. with   Repts. with
            Drug A       Dis. X        Dis. X

Dec-99        0             0            0.0
Mar-00        25            2            8.0
Jun-00        75           12           16.0
Sep-00       100           30           30.0
Dec-00       120           37           30.8
Mar-01       125           29           23.2
Jun-01       120           31           25.8
Sep-01       135           32           23.7
Dec-01       130           35           26.9
Mar-02       120           30           25.0
Jun-02       140           35           25.0
Sep-02       135           34           25.2
Dec-02       140           32           22.9

Quarter    Tot Adv.    Drug B Adv.   % of Drug B
Ending      Event         Event      Adv. Event
          Repts. for   Repts. with   Repts. with
            Drug B       Dis. X        Dis. X

Dec-99       150            4            2.7
Mar-00       165            4            2.4
Jun-00       160            5            3.1
Sep-00       170            5            2.9
Dec-00       180            6            3.3
Mar-01       175            5            2.9
Jun-01       150            4            2.7
Sep-01       160            4            2.5
Dec-01       170            5            2.9
Mar-02       165            4            2.4
Jun-02       180            5            2.8
Sep-02       180            6            3.3
Dec-02       175            5            2.9

Quarter   Tot. Adv.    Drug C Adv.   % of Drug C
Ending      Event         Event      Adv. Event
          Repts. for   Repts. with   Repts. with
            Drug C       Dis. X        Dis. X

Dec-99         0            0            0.0
Mar-00         0            0            0.0
Jun-00         0            0            0.0
Sep-00         0            0            0.0
Dec-00        35            0            0.0
Mar-01        50            1            2.0
Jun-01        75            1            1.3
Sep-01        80            2            2.5
Dec-01        90            1            1.1
Mar-02        95            2            2.1
Jun-02       100            3            3.0
Sep-02        95            2            2.1
Dec-02       105            2            1.9

Table 3: Data for PRR Comparison of Disease X
for Drug A & Drug B Used Together Versus Drug A Used Alone

Quarter    Tot. Adv.       Drug A &     % of Drug A &
Ending    Event Repts.   Drug B Adv.     Drug B Adv.
          for Drug A &   Event Repts.   Event Repts.
             Drug B      with Dis. X     with Dis. X

Dec-99          0              0              0.0
Mar-00          5              1             20.0
Jun-00         10              6             60.0
Sep-00         12             12            100.0
Dec-00         15             15            100.0
Mar-01         14             12             85.7
Jun-01         11             10             90.9
Sep-01         12             11             91.7
Dec-01         14             13             92.9
Mar-02         15             13             86.7
Jun-02         12             11             91.7
Sep-02         15             14             93.3
Dec-02         12             11             91.7

Quarter    Tot. Adv.     Drug A Alone
Ending    Event Repts.    Adv. Event
           for Drug A    Repts. with
             Alone          Dis. X

Dec-99         0               0
Mar-00         20              1
Jun-00         65              6
Sep-00         88             18
Dec-00        105             22
Mar-01        111             17
Jun-01        109             21
Sep-01        123             21
Dec-01        116             22
Mar-02        105             17
Jun-02        128             24
Sep-02        120             20
Dec-02        128             21

Quarter   % of Drug A     PRR-Drug A
Ending     Alone Adv.      & Drug B
          Event Repts.   versus Drug
          with Dis. X      A Alone

Dec-99         0.0           0.0
Mar-00         5.0           4.0
Jun-00         9.2           6.5
Sep-00        20.5           4.9
Dec-00        21.0           4.8
Mar-01        15.3           5.6
Jun-01        19.3           4.7
Sep-01        17.1           5.4
Dec-01        19.0           4.9
Mar-02        16.2           5.4
Jun-02        18.8           4.9
Sep-02        16.7           5.6
Dec-02        16.4           5.6

Notes

(1.) Judy A. Staffa et al., Letter to the Editor, Cerivastatin and Reports of Fatal Rhabdomyolysis, 346 NEW ENG. J. MED. 539,540 (2002).

(2.) Id.

(3.) The AERS database contains 279 rhabdo reports for Baycol from August 1, 1999, through July 31,2001. For all other statins combined during that period, there were 237 rhabdo reports. An article published in The Washington Post the day after Baycol was withdrawn from the market reported that the drug's peak market share was about 6.7 percent. David Brown, Cholesterol Drug Taken off Market, WASH. POST, Aug. 9, 2001, at A1, available at www.washingtonpost.com/ac2/ wp-dyn? pagename=article&node=&contentId= A50968-21/01Aug8 (last visited Jan. 27, 2005). Conservatively assuming Baycol held that market share during the period in question, the relative rhabdo reporting rate for the drug compared with the other statins would be (279/6.7) / (237/ 93.3), which equals 16.4. The actual relative reporting rate is probably higher.

(4.) David J. Graham et al., Incidence of Hospitalized Rhabdomyolysis in Patients Treated with Lipid-Lowering Drugs, 292 JAMA 2585, 2589 (2004). Three of the study's 10 authors are FDA scientists.

(5.) Mohamed A. Omar & James E Wilson, IDA Adverse Event Reports on Statin-Associated Rhabdomyolysis, 36 ANNALS PHARMACOTHERAPY 288, 288 (2002).

(6.) Michelle Meadows, The FDA's Drug Review Process: Ensuring Drugs Are Safe. and Effective, FDA CONSUMER MAG. (July-Aug. 2002, rev. Sept. 2002), at www.fda.gov/fdac/features/2002/ 402_drug.html (last visited Jan. 27, 2005).

(7.) Robert J. Temple & Martin H. Himmel, Editorial, Safety of Newly Approved Drugs: Implications for Prescribing, 287 JAMA 2273, 2273 (2002) (citing Karen E. Lasser et al., Timing of New Black-Box Warnings and Withdrawals for Prescription Medications, 297 JAMA 2215 (2002)).

(8.) Bengt-Erik Wiholm et al., ,Spontaneous Reporting Systems Outside the United ,States, in PHARMACOEPIDEMIOLOGY 175 (Brian L. Strom ed., 3d ed. 2000).

(9.) See Dianne L. Kennedy et al., ,Spontaneous Reporting in the United States, in PHARMACOEPIDEMIOLOGY 151, 153 (Brian L. Strom ed., 3d ed. 2000).

(10.) Wiholm et al., supra note 8, at 175-76.

(11.) A single year's worth of data costs only $414. To obtain data, go to www.ntis.gov and type "adverse event" in the search box. A list of year: specific ordering instructions will appear. The data is also available for free at www.fda.gov/cder/ aers/extract.htm (last visited Jan. 27, 2005).

(12.) Roger Goetsch, E2B (M) : FDA Perspective, Presentation to the Drug Industry Association, Slide 9 (June 15, 2004), available at www.fda.gov/ cder/present/DIA2004/Goetsch.ppt (last visited Jan. 27, 2005).

(13.) Applications for FDA Approval to Market a New Drug, 21 C.F.R [section] 314.80 (f) (2004).

(14.) See FDA, Welcome to the FDA AERSElectronic Submissions Web Site, at www.fda.gov/cder/ aerssub (last visited Jan. 27, 2005).

(15.) See generally International Conference on Harmonisation; Guidance on Data Elements for Transmission of Individual Case Safety Reports, 63 Fed. Reg. 2396 (Jan. 15, 1998).

(16.) 21 C.F.R. [section] 314.80(a).

(17.) Id.

(18.) Id. [section] 314.80(c) (1).

(19.) Id. [section] 314.80(c) (2).

(20.) Id. [section] 314.80(c)(2)(iii).

(21.) The dictionary is available online at www. meddramsso.com/NewWeb2003/index.htm (last visited Jan. 8, 2005).

(22.) 21 C.F.R. [section] 314.80(c) (1)(ii).

(23.) Kennedyet al., supra note 9, at 165-66.

(24.) Id. at 167.

(25.) Id. at 167-68.

(26.) Id. at 168.

(27.) Id. at 167.

(28.) See generally Carlene Baum et al., Differences in Manufacturer Reporting of Adverse Drug Reactions to the FDA in 1984, 21 DRUG INFO. J. 257 (1987); Gerald A. Faich et al., Sources of Spontaneous Adverse Drug Reaction Reports Received by Pharmaceutical Manufacturers, 21 DRUG INFO. J. 251 (1987).

(29.) Bruce M. Psaty et al., Potential for Conflict of Interest in the Evaluation of Suspected Adverse Drug Reactions: Use of Cerivastatin and Risk of Rhabdomyolysis, 292 JAMA 2622, 2625 (2004).

(30.) Kennedy et al., supra note 9, at 166.

(31.) Id.

(32.) Id.

(33.) FDA, DRAFT GUIDANCE FOR INDUSTRY ON GOOD PHARMACOVIGILANCE PRACTICES AND PHARMACOEPIDEMIOLOGIC ASSESSMENT 9-10 (May 2004), available at www.fda.gov/ohrms/dockets/98fr/04d-0189gd10001-5767dft.pdf (last visited Jan. 27, 2005) [hereinafter DRAFT GUIDANCE].

(34.) Kenneth J. Rothman & Sander Greenland, MODERN EPIDEMIOLOGY 31-32 (2d ed. 1998).

(35.) DRAFT GUIDANCE, supra note 33, at 10.

(36.) Wiholm et al., supra note 8, at 180-81.

(37.) See Bert Black & David E. Lilienfeld, Epidemiologic Proof in Toxic Tort Litigation, 52 FORDHAM L. REV. 732, 757-58 (1984).

(38.) Id. at 181.

(39.) 21 C.F.R. [section] 314.80(b).

(40.) Staffa et al., supra note 1, at 540, tbl.l.

(41.) Allen C. Rossi et al., The Importance of Adverse Reaction Reporting by Physicians: Suprofen and the Flank Pain Syndrome, 259 JAMA 1203 (1988).

(42.) Psaty et al., supra note 29, at 2625.

BERT BLACK is a partner with Lockridge Grindal Nauen in Minneapolis.

KEITH ALTMAN is the director of adverse event analysis with the law firm of Finkelstein and Partners in Newburgh, New York.