Historical evidence exercise. Decision date February 8, 2001

Rofecoxib in VIGOR: the 2001 formulary record beside what came later

The left column uses only sources dated on or before the decision date. The right column holds a company announcement from 2004 and a journal notice from 2005. Those later items were not knowable on February 8, 2001, and nothing on this page treats them as if they were.

Dated timeline

November 23, 2000 VIGOR article published in NEJM. VX1, dated 2000-11-23 Available by the decision date.
February 1, 2001 FDA cardiovascular safety review memo dated. VX2, dated 2001-02-01 Available by the decision date.
February 8, 2001 Decision date for this evidence sheet. The formulary call is fixed here.
September 30, 2004 Merck withdrawal announcement citing a separate trial, APPROVe. VX3, dated 2004-09-30 Later; not knowable on the decision date.
December 8, 2005 NEJM Expression of Concern published online. VX4, dated 2005-12-08 Later; not knowable on the decision date.
December 29, 2005 The same notice in the print issue. VX4, dated 2005-12-08, print 2005-12-29 Later; not knowable on the decision date.

Two evidence columns

Dated on or before 2001-02-08

Knowable on February 8, 2001: trial article and FDA review

Formulary call: Yes: disclose

The gastrointestinal benefit should not stand alone. The myocardial infarction comparison belongs in the study description and a cardiovascular statement belongs in Warnings. With no placebo group, naproxen protection could not be told apart from rofecoxib harm.

Basis, the FDA reviewer: It would be difficult to imagine inclusion of VIGOR results in the rofecoxib labeling without mentioning cardiovascular safety results in the study description as well as the Warnings sections. VX2, dated 2001-02-01, PDF p. 39

Trial articleVIGOR randomized 8,076 patients to rofecoxib 50 mg daily or naproxen 500 mg twice daily: two active arms. VX1, dated 2000-11-23, abstract
Trial articleConfirmed gastrointestinal events: 2.1 with rofecoxib versus 4.5 with naproxen per 100 patient-years, relative risk 0.5. VX1, dated 2000-11-23, abstract
Trial articlePrinted myocardial infarction incidence: 0.4 percent with rofecoxib versus 0.1 percent with naproxen, relative risk 0.2 (naproxen to rofecoxib). VX1, dated 2000-11-23, abstract
FDA review, updated application dataAcute myocardial infarction: 20 of 4,047 rofecoxib participants (4.94 per 1000) versus 4 of 4,029 naproxen participants (0.99 per 1000). Crude risk ratio 4.98; 3.95 more per 1000 participants over trial follow up. VX2, dated 2001-02-01, PDF p. 20
FDA reviewerAs there is no placebo group, it will be difficult to assess the CV thrombotic risk with rofecoxib use compared with no therapy at all. VX2, dated 2001-02-01, PDF p. 36
Open question in 2001The printed 0.4 percent sits below the 0.49 percent implied by the FDA count of 20 of 4,047. These two sources do not explain the gap. VX1, dated 2000-11-23; VX2, dated 2001-02-01

Dated 2004-09-30 and 2005-12-08

Later: company announcement and editors' notice, not knowable on February 8, 2001

Merck, company announcementMerck &. Co., Inc. today announced a voluntary worldwide withdrawal of VIOXX® (rofecoxib), its arthritis and acute pain medication. The company’s decision, which is effective immediately, is based on new, three-year data from a prospective, randomized, placebo-controlled clinical trial, the APPROVe (Adenomatous Polyp Prevention on VIOXX) trial. VX3, dated 2004-09-30
Merck, company announcementIn APPROVe, increased relative risk for confirmed cardiovascular events began after 18 months of treatment, compared with placebo. VX3, dated 2004-09-30
Our reading of VX3APPROVe is a separate study with a placebo arm, unlike the two-arm VIGOR trial. Its outcome does not show that VIGOR identified a causal mechanism, and the excerpt names no mechanism. VX3, dated 2004-09-30
Journal editors statedThree myocardial infarctions, all in the rofecoxib group, were not included in the data submitted to the Journal. VX4, dated 2005-12-08, p. 1
Journal editors, Tables 1 and 2Manuscript window, omitting the events: 17 infarctions in 2,315 rofecoxib person-years versus 4 in 2,316 naproxen person-years, relative risk 4.25. Included-event window: 20 in 2,698 versus 4 in 2,699, relative risk 5.00. VX4, dated 2005-12-08, p. 1
Our arithmetic, not the editors' wordsThe included window adds 3 rofecoxib events, no naproxen events, and 383 rofecoxib and 383 naproxen person-years. The FDA 2001 tally of 20 already equals the editors' included count, so the three events are not added a second time. VX4, dated 2005-12-08; VX2, dated 2001-02-01

Two windows in the editors' tables

These rates come from the 2005 notice. They are myocardial infarctions per 1000 person-years, a different unit from the FDA risk per participant, so the editors' ratio of 5.00 and the FDA crude ratio of 4.98 are not interchangeable.

Manuscript window, rofecoxib: 7.34 per 1000 person-years
Manuscript window, naproxen: 1.73 per 1000 person-years
Included window, rofecoxib: 7.41 per 1000 person-years
Included window, naproxen: 1.48 per 1000 person-years

Bars start at zero on a scale of 0 to 8 per 1000 person-years. Rates are our division of the editors' counts by their person-years; the ratios 4.25 and 5.00 are as reported. VX4, dated 2005-12-08, p. 1

Open the table difference: manuscript window versus included window

Exact excerpt from the editors: Three myocardial infarctions, all in the rofecoxib group, were not included in the data submitted to the Journal. VX4, dated 2005-12-08, p. 1

Editors' Table 1 and Table 2, with the difference computed here
Measure	Table 1, omitting the three	Table 2, including the three	Difference
Rofecoxib infarctions	17	20	plus 3
Rofecoxib person-years	2,315	2,698	plus 383
Naproxen infarctions	4	4	none
Naproxen person-years	2,316	2,699	plus 383
Relative risk, as reported	4.25	5.00	plus 0.75

The included window is longer as well as fuller: both arms gain 383 person-years, so the naproxen rate falls from 1.73 to 1.48 per 1000 while its count stays at 4. The FDA's 2001 updated table already counted 20 rofecoxib infarctions among 4,047 participants. Adding the three omitted events to that tally would give 23, a figure no source reports, so it is not used.

FDA 2001 counts beside the editors' 2005 included table
Measure	FDA, 2001	Editors, 2005	Reading
Rofecoxib infarctions	20 of 4,047 participants	20 in 2,698 person-years	Same count; no double counting.
Naproxen infarctions	4 of 4,029 participants	4 in 2,699 person-years	Same count.
Ratio	4.98, risk per participant	5.00, rate per person-year	Different units.

Sources: VX2, dated 2001-02-01, PDF p. 20; VX4, dated 2005-12-08, p. 1

What the editors said versus what we infer

The editors' characterization

The editors stated only that three rofecoxib infarctions were not included in the data submitted to the Journal. VX4, dated 2005-12-08, p. 1

Our inferences, made from the later record

Arithmetic consistency only: 17 of the FDA rofecoxib denominator of 4,047 is 0.42 percent, which rounds to the 0.4 percent printed in 2000, while 20 of 4,047 is 0.49 percent. This suggests, but does not prove, that the article's figure matches the manuscript window. VX4, dated 2005-12-08; VX1, dated 2000-11-23
On February 8, 2001, this gap was open and unexplained; the reconciliation above depends on the 2005 notice.

What we do not infer

These excerpts do not say who left the three events out, when, or why.
We draw no inference about any author's knowledge or intent.
APPROVe does not show that VIGOR identified a causal mechanism.
No population-wide number of harmed people is estimated, and gastrointestinal rates are not netted against infarction risks.

Source ledger

VX1. Comparison of Upper Gastrointestinal Toxicity of Rofecoxib and Naproxen in Patients with Rheumatoid Arthritis. New England Journal of Medicine, dated 2000-11-23. Abstract. PubMed record, dated 2000-11-23
VX2. Review of cardiovascular safety database; NDA 21-042 S-007. FDA Division of Cardio-Renal Drug Products, dated 2001-02-01. Memo pages 1, 18 and 34 to 37, preserved in a later FDA correspondence scan. FDA scan, dated 2001-02-01; FDA meeting briefing docket
VX3. Merck Announces Voluntary Worldwide Withdrawal of VIOXX. Merck in SEC Exhibit 99, dated 2004-09-30. Opening announcement and APPROVe paragraphs. SEC exhibit, dated 2004-09-30
VX4. Expression of Concern: Bombardier et al. VIGOR trial. New England Journal of Medicine, dated 2005-12-08 online, print issue 2005-12-29. Pages 1 and 2. Journal PDF, dated 2005-12-08