Carter-Glogau Laboratories, Inc., now known as Retrae, Inc., Ronald M. Carter, Sr. (Carter), and Larry K. Hiland appeal from judgments of conviction entered by the district court1 following a two-month jury trial. The convictions resulted from the manufacture and distribution of E-Fer-ol Aqueous Solution (E-Ferol), a pharmaceutical product administered intravenously to premature infants. Each of the appellants was convicted on one count of conspiracy to commit mail and wire fraud, and to violate the Federal Food, Drug, and Cosmetic Act (FDCA), in violation of 18 U.S.C. § 371; six counts of introducing a new drug not approved by the Food and Drug Administration (FDA) into interstate commerce with the intent to defraud or mislead, in violation of 21 U.S.C. §§ 331(d), 333(a)(2);2 and six counts of introducing a misbranded drug into interstate commerce with the intent to defraud or mislead, in violation of 21 U.S.C. §§ 331(a), 333(a)(2). Hiland was also convicted on five counts of committing mail fraud, in violation of 18 U.S.C. § 1341. Carter and Carter-Glogau were acquitted on these counts. All of the appellants were acquitted on seven counts of committing wire fraud, in violation of 18 U.S.C. § 1343.
The other two defendants, O’Neal, Jones and Feldman, Inc. (OJF), now known as O’Neal, Inc., and James B. Madison, entered guilty pleas pursuant to plea agreements. On the first day of trial, OJF pled guilty to one count of conspiracy, four counts of mail fraud, and twelve FDCA felony counts.3 Midway through the trial, Madison pled guilty to two counts of wire fraud and one FDCA felony count.4 Madison subsequently testified for the government. OJF and Madison have not appealed their convictions or sentences.
The appellants allege numerous grounds for reversal of their convictions.5 They join in contending that (1) the jury was not properly instructed on the knowledge required for the FDCA offenses; (2) the district court erred in giving a willful blindness instruction; and (3) the court erred in admitting prejudicial medical testimony regarding the effects E-Ferol had on premature infants. Carter also argues that (1) FDA policy actively led him to believe E-Ferol could be lawfully marketed without a new drug approval; (2) this policy was so vague and indefinite that he lacked fair warning his conduct was illegal; and (3) a statement in the prosecutor’s closing argument was improper.6 Hiland also argues that (1) the court’s instruction on his theory of defense impermissibly shifted the burden of proof; (2) the court erred in allowing the government to cross-examine him about certain instances prior to E-Ferol in which OJF had marketed drugs without FDA approval; (3) the court had no authority to resubmit the misbranding counts against him to the jury, and doing so was both coercive and a violation of the double jeopardy clause; (4) the mail fraud verdict against him cannot stand because the court failed to poll the jurors individually on that verdict; and (5) the court erred in failing to give a specific unanimity instruction in*1119forming the jury that its verdict must be unanimous with respect to each defendant on each count. Carter and Hiland do not challenge their sentences.7 Carter-Glogau challenges the portion of its sentence requiring it to pay $100,000 toward the costs of prosecution.8 We vacate this part of Carter-Glogau’s sentence and remand for a redetermination of taxable costs. We affirm all of the convictions.
I.
Carter-Glogau, located in Glendale, Arizona, was a manufacturer of generic (or “me-too”) injectable drugs. Carter was the corporation’s president and chief operating officer. OJF, located in Maryland Heights, Missouri, was a distributor of prescription pharmaceutical products, primarily generic drugs. Hiland was OJF’s president and Madison was its executive vice-president of operations. Almost all of the injectable drugs distributed by OJF were manufactured by Carter-Glogau. In most cases, the drugs manufactured by Carter-Glogau for OJF were generic copies of innovator (or “benchmark”) drugs that were formulated by other companies and approved by the FDA.
Carter-Glogau began manufacturing E-Ferol for OJF in the fall of 1983. E-Ferol was a high potency vitamin E solution for intravenous administration to premature infants. It consisted of a type of vitamin E and two types of polysorbates, which are emulsifying agents. Carter-Glogau manufactured and shipped to OJF three commercial lots of E-Ferol, totaling approximately 40,000 vials. OJF distributed approximately 26,000 vials of the product from November 1983 to April 6, 1984, when the FDA requested that shipments of E-Ferol be ceased because of reports linking it to the illness and death of premature infants. OJF initiated a total recall of E-Ferol on April 11, 1984.
A. Development of E-Ferol
The development of E-Ferol was prompted. by the need for an intravenous form of vitamin E to combat retrolental fibroplasia (RLF), a disease that causes impaired vision or permanent blindness in premature infants. Many neonatologists, physicians specializing in the care of premature infants, considered vitamin E to be useful in reducing the incidence and severity of RLF. The two principal forms of vitamin E available to neonatologists in the early 1980’s were an oral preparation and an intramuscular injection, both of which were sold as nutritional supplements and not represented as safe and effective for use in premature infants.9 For some years prior to the development of E-Ferol, Carter-Glogau had been manufacturing a vitamin E intramuscular product (“E-Ferol IM”) for OJF. This product was labeled as a nutritional supplement and was not approved by the FDA. Although its labeling contained no reference to RLF, E-Ferol IM was used by some neonatologists to treat this disease.10 However, like the oral form of vitamin E, • E-Ferol IM had drawbacks that made it difficult to administer to premature infants.
In April 1982, one of Carter-Glogau’s customers wrote Carter to ask whether an intravenous form of vitamin E could be developed, noting that “[tjhere must be a Hell of a market out there.” Carter expressed a reluctance to develop such a product. In his responses to the customer’s inquiry, he stated that the amount of polysorbates needed “may be detrimental,” and pointed out that “fat emulsions for IV *1120use ... are very tricky products and fraught with particular size problems.”
In August 1982, Madison wrote Carter to see if he could develop for OJF a high potency intravenous form of vitamin E for use in premature infants. He informed Carter that Hoffmann-LaRoche, a large pharmaceutical company, was testing an injectable vitamin E product for the treatment of RLF in an effort to obtain FDA approval of the product. Madison wrote that he was “afraid that when Roche gets their Vitamin E approved, we will lose the business, unless you can come up with something.” Madison’s letter clearly indicated that the primary purpose of the product he was proposing would be to treat RLF, and stated, “We could always label it for Vitamin E supplementation.” Hiland received a copy of this letter. Several weeks later, Madison sent Carter a followup letter recommending a dosage level sufficient to alleviate RLF.
In his responses to Madison’s inquiries, Carter expressed serious safety concerns regarding the development of an intravenous vitamin E product, stating in part: “If we make some attempt to solubilize the Vitamin E and use the wrong proportions and kill a few infants, we’d have some serious problems.” Carter was specifically concerned about developing such a product without proper clinical testing. He wrote Madison that:
The administration of this product intravenously in neonatals without appropriate clinical work concerning toxicity will undoubtedly lead to an exposure in terms of product liability which neither you nor we may wish to assume.
After all, one neonatal death is one too many.
Carter suggested to Madison that the best way to proceed would be to wait until Hoffman'n-LaRoche came out with the “benchmark” product so that they could then copy its formulation, thereby taking advantage of Hoffman-LaRoche’s toxicity studies. This was Carter-Glogau’s usual mode of operation.
Carter and Madison resumed their dialogue about vitamin E early in the summer of 1983. Notwithstanding the strong safety concerns he had expressed less than a year earlier, Carter went ahead and developed a high potency intravenous vitamin E product (E-Ferol) for OJF. He did so despite knowledge that Hoffmann-LaRoche was still engaged in clinical testing of its intravenous vitamin E product and had not received approval from the FDA to market it. Carter personally made the final decisions as to E-Ferol’s formulation, including the type of vitamin E and the types and proportions of polysorbate the product would contain. Carter admitted that at the time he made these decisions, he did not even know what level of polysorbates was safe for injection into premature infants. At no time did Carter do testing or research of any kind to determine whether E-Ferol’s formulation would be safe and effective for use in premature infants. Nor did he ever request OJF to do any testing. Carter testified that he instead relied on Madison as the “expert” on E-Ferol, even though Madison had no formal scientific training.
In the summer of 1983, Madison requested and received authorization from Hiland to explore the possibility of having Carter-Glogau supply OJF with the high potency vitamin E intravenous product being formulated by Carter. Madison informed Hi-land that its primary use would be for the treatment of RLF and that Hoffmann-La-Roche was testing a similar product under FDA guidelines. Hiland subsequently approved the addition of E-Ferol to OJF’s product line. Only Hiland had the authority to authorize such an addition. OJF never did any testing to determine whether E-Ferol would be safe and effective for use in premature infants. Nor did Hiland or Madison ever ask Carter-Glogau to perform such testing.
B. Labeling and Marketing of E-Ferol
In August 1983, Carter and Madison prepared the labeling for E-Ferol. Carter wrote the first draft of the package insert that was to accompany the product. He used the E-Ferol IM label as a model even though he knew the new product would *1121have an intravenous route of administration. The “Indications” section of the insert set forth uses of E-Ferol as a nutritional supplement, but at Madison’s request, Carter put a reference to RLF in the insert’s “Clinical Pharmacology” section. Carter also accepted Madison’s suggestion that the dosage recommendation in the label be the same as the E-Ferol IM dosage that Madison understood was being used by physicians to treat RLF (25 to 50 milligrams), and included that recommended dosage in the draft insert he sent Madison.
Madison made revisions to E-Ferol’s package insert, including additional references to RLF, all of which he sent to Carter for comments. The reference to RLF in the clinical pharmacology section of the final draft of the insert read as follows: “Reports in the literature indicate that substantial doses of Vitamin E will reduce the severity of retrolental fibroplasia in neona-tals, which are administered oxygen because of their low weight, under 1500 grams.* ”11 The asterisk referred the reader to the insert’s “References” section, in which Madison had placed a list of scientific articles pertaining to the use of vitamin E as a treatment for RLF. There was nothing in the labeling that indicated E-Ferol had never been tested for safety and effectiveness.
In a memorandum dated July 20, 1983, Madison advised Hiland that the new intravenous vitamin E product would be used for nutritional purposes and “[f]or Retro-lental Fibroplasia (RLF) for which the dose is 25 mg. every day. (We cannot label it for this use — it would make it a New Drug).” Madison also forwarded Hiland an August 8,1983 memorandum from Samuel Fainberg, OJF’s technical and regulatory consultant, in which Fainberg stated, “I do not believe that legally we can make a claim that Vitamin E is effective against Retrolental Fibroplasia if given to infants. This will require a[n] NDA [new drug application] approval for the use of Vitamin E as a drug.” On August 15, 1983, Madison sent Hiland a memorandum discussing the labeling for E-Ferol and a marketing plan for the product. In this memorandum, he told Hiland that the package insert made reference to the product’s use for RLF. Madison then explained, “We have to be careful how we show the ‘Indications’ for RLF. See insert; we list it under Clinical Pharmacology.” According to Madison’s testimony, he included with the memorandum a copy of the draft insert for Hiland to review. Hiland claimed at trial that he did not read, E-Ferol’s package insert until January 1984.
In September 1983, Hiland approved the marketing campaign for E-Ferol, which consisted of a mass mailing of “Dear Doctor” letters accompanied by a brochure with a copy of the package insert on the reverse side. These promotional materials were mailed in late October 1983 and again in January 1984 to approximately one thousand directors of neonatal intensive care units. This group was chosen because they were involved in the treatment of RLF. Nothing in the promotional materials indicated that E-Ferol had never been tested to determine whether it was safe and effective for the treatment of RLF in premature infants.
At trial, physicians and pharmacists testified that based on E-Ferol’s labeling, they believed the product was being promoted for the treatment of RLF in premature infants. In addition to the references to RLF, the high dosage recommended in the labeling was a clear indication to them that E-Ferol was to be used for RLF because the dosage was well above any possible nutritional needs of premature infants. Pharmacists and physicians also testified that they were led to believe E-Ferol had been tested and proven as safe for use in premature infants, especially because the labeling contained no warnings or suggestions to the contrary and it was extremely unusual to see a product specifically labeled and marketed for use in premature infants. They testified further that they assumed E-Ferol had been approved by the *1122FDA because it was targeted for the treatment of RLF in premature infants and had an intravenous route of administration, which distinguished it from the other single-entity vitamin E products on the market.
C. Adverse Reaction Reports and Revision of Labeling
During the five months that E-Ferol was on the' market, OJF received numerous inquiries about the product and three reports of severe adverse reactions associated with its use. Each of the reports related similar unusual symptoms, and two of them included infant deaths associated with E-Ferol.
Inquiries and complaints about OJF products were directed to Madison.- In December 1983, he received a telephone call from Dr. David Easa, a neonatologist at Kapiola-ni Hospital in Honolulu, Hawaii, who reported that two premature infants under his care had developed unusual symptoms after receiving E-Ferol. Madison testified that he immediately advised Hiland of this report. Hiland denied this, claiming that he did not learn of the Hawaii report until after the recall of E-Ferol in April 1984.
On January 26, 1984, Dr. Carl Boden-stein, a neonatologist at Sacred Heart Medical Center in Spokane, Washington, telephoned Madison to report that E-Ferol may have caused the deaths of three premature infants and serious illness in a fourth. Bodenstein told Madison that the surviving infant’s blood contained excessively high levels of vitamin E. Madison informed Hiland of Bodenstein’s call and sent him a memorandum describing the report the next day.
Hiland halted the distribution of E-Ferol and took charge of the investigation into the validity of the Spokane report. He instructed Madison to notify Carter, consult with Fainberg, do a literature search on the safety of vitamin E and polysor-bates, contact other neonatal units to see if they had experienced similar problems, and obtain more information from Bodenstein. No effort was made to advise hospitals to which E-Ferol had been distributed that OJF had received a report associating the product with the deaths of premature infants. The OJF sales force was instead instructed to tell customers that E-Ferol was unavailable because it was on back order. Madison testified that when he asked Hiland what he should say to hospitals inquiring about the product, Hiland told him not to mention the Spokane report. Hiland denied giving Madison such a directive. Twelve days after the distribution of E-Ferol had been suspended, Hiland made the decision to resume all shipments of the product.
The evidence at trial established that Madison’s investigation into the Spokane incident was woefully inadequate, even when measured by the limited instructions Hi-land had given him. For his part, Hiland knew when he ordered the resumption of shipments that Bodenstein had not in fact been successfully contacted for additional information. Hiland did not ask Madison for a written report of his investigation. Nor did he inquire into the extent of Madison’s literature search, or ask him how many hospitals he had contacted. The Spokane report was far and away the most serious adverse reaction report Hiland had ever encountered. Yet, he assigned the investigative tasks to Madison despite knowing that Madison had no medical or other scientific training to evaluate such reports, aside from his experience in the pharmaceutical industry. Five months pri- or to the Spokane incident, Hiland wrote an evaluation criticizing Madison for jumping to conclusions, reaching conclusions or opinions without giving enough thought and consideration to the overall problem, and working on too many different areas without concluding on any one issue.
Carter was also informed by Madison on January 26, 1984 of the Spokane report. A few days later, Madison sent Carter a copy of the same memorandum he had given Hiland regarding the report. From this memorandum, Carter knew that Boden-stein was questioning both E-Ferol’s dosage recommendation and its high polysor-bate content. Madison informed Carter that OJF had temporarily stopped shipping E-Ferol while the Spokane report was be*1123ing investigated. Carter did nothing to determine the accuracy of this report, even though it was the most serious adverse reaction report he had ever received, and was precisely the type of incident he had feared when he expressed serious concerns in the summer and fall of 1982 about the potential for infant deaths. Carter testified that he simply forgot the safety concerns he had previously put in writing. In late February, Carter asked Madison what he should do with the third lot of E-Ferol Carter-Glogau was holding. When Madison told him to ship the lot to OJF, Carter did so without asking for any details about OJF’s investigation into the Spokane report.
On January 26 and 27, 1984, Hiland discussed the Spokane report with John Lemker, an attorney who represented OJF in food and drug matters. After reading E-Ferol's package insert, which he had not previously been asked to review, Lemker advised Hiland that the references to RLF made the product a drug and that the FDA would probably consider it to be a new drug. Lemker further advised that OJF stop shipping E-Ferol and contact experts for an evaluation of the drug. Fainberg also recommended to Hiland that shipments of E-Ferol be ceased. Hiland never informed Lemker and Fainberg of his subsequent decision to resume shipments of E-Ferol with its existing package insert.
After his discussions with Lemker and Fainberg, Hiland directed Madison to work with them and Carter to revise E-Ferol’s labeling. After seeing one of the proposed insert revisions, Hiland returned it to Madison with a note stating, “This insert really reads as if it’s a drug.” With Hiland and Carter aware of the changes being made, Madison prepared a revised package insert that deleted all references to RLF, greatly reduced the recommended dosage so it conformed with nutritional dosages, and added a warning that levels of vitamin E in the blood should not exceed a certain amount. The revised labeling was finalized in March 1983, but was never included with any shipments of E-Ferol. Hiland testified that he allowed the shipments to resume before the revised insert was packaged because it had no “safety issues,” but rather only “technical changes.”
In a memorandum dated March 27, 1984, Madison advised Hiland of a third adverse reaction report. This report came from Melanie Sandlin, a pharmacist at the University of Tennessee Hospital, who informed Madison that the hospital had stopped administering E-Ferol because of four premature infant deaths associated with its use. Hiland testified he did not order that shipments of E-Ferol be stopped because he thought the manner in which OJF had received the report was unusual.12 Hiland’s only response to the Tennessee report was to write Madison a note asking him to follow up with the hospital and make sure that the pharmacist had received some literature Madison had sent her.
Shipments of E-Ferol with the old package insert continued until April 6, 1984, when the FDA contacted OJF and requested that shipments be ceased because of reports associating the product with serious adverse reactions and deaths. Five days later, OJF initiated a total recall of E-Ferol, which resulted in its permanent removal from the market.
D. Indictment
The grand jury indictment against the appellants and their codefendants was returned on July 9, 1987. Trial commenced before a jury on August 1, 1988 and concluded on September 30, 1988.
The mail fraud counts charged in the indictment were based primarily on promotional mailings sent by OJF to hospital neonatal units. The FDCA counts were based on six shipments of E-Ferol from OJF to hospitals after the defendants had received the Spokane adverse reaction report in late January 1984. Two of these shipments were made in February 1984; the other four were made in early April 1984, after OJF had received the adverse reaction report from Tennessee.
The FDCA counts charged that E-Ferol was an unapproved “new drug” and “mis-*1124branded” under the FDCA, and that the defendants introduced it into interstate commerce with the intent to defraud and mislead. E-Ferol was alleged to have been misbranded in five ways in that its labeling (1) omitted certain material facts; (2) failed to bear adequate directions for use; (3) failed to bear adequate warnings for when its use might be dangerous; (4) failed to bear adequate warnings against unsafe dosage, administration, and application; and (5) recommended and suggested conditions of use under which E-Ferol was dangerous to the health of premature infants.13 The essence of the fraud against neonatologists and other medical professionals alleged by the government was that the defendants intentionally represented E-Ferol as safe and effective for intravenous use in premature infants to treat RLF despite knowing no testing had been done to establish it as safe and effective for such use, and continued to do so after receiving reports associating E-Ferol with severe adverse reactions and deaths.
II.
A. Due Process
Carter argues that his conviction on the new drug counts violated due process because (1) FDA policy actively led him to believe that E-Ferol could be marketed lawfully without a new drug approval, and (2) this same policy was so vague and indefinite as to deprive him of fair warning that his conduct was illegal. Before addressing these arguments, we will set forth relevant provisions of the FDCA and briefly outline the legislative and regulatory background of the FDA policy on which Carter focuses his due process challenges.
A list of acts prohibited by the FDCA is set out in 21 U.S.C. § 331. Section 331(d) prohibits “[t]he introduction or delivery for introduction into interstate commerce of any article in violation of section 344 or 355” of title 21. Section 355(a) states in turn that “[n]o person shall introduce or deliver for introduction into interstate commerce any new drug, unless an approval of an application filed pursuant to subsection (b) or (j) of this section is effective with respect to such drug.” An article is a drug if it is “intended for use in the diagnosis, cure, mitigation, treatment, or prevention of disease in man.” 21 U.S.C. § 321(g)(1)(B). The term “new drug” means “[a]ny drug ... the composition of which is such that such drug is not general*1125ly recognized, among experts qualified by scientific training and experience to evaluate the safety and effectiveness of drugs, as safe and effective for use under the conditions prescribed, recommended, or suggested in the labeling thereof.” § 321(p)(l). The penalties for violations of § 331 are contained in 21 U.S.C. § 333. Section 333(a)(1) provides for misdemeanor punishment of “[a]ny person who violates a provision of section 331.” Carter was convicted under § 333(a)(2), which makes it a felony to “commit[] such a violation with the intent to defraud or mislead.”
Regulation of the drug industry was first made a part of federal law by the Food and Drugs Act of 1906. That statute did not give the government authority to require manufacturers to obtain FDA approval before introducing a drug into the market. Instead, the government had to initiate legal action to remove a drug from the market if it was adulterated or misbranded. This scheme changed in 1938 with the enactment of the FDCA, which required manufacturers to obtain an approved new drug application (NDA) before marketing a new drug. The NDA had to contain data establishing that the drug was safe for its intended uses. If not rejected by the FDA within a certain period of time, an NDA was automatically deemed approved. Due to a grandfather clause, no approval was required for drugs marketed under the 1906 Act if their conditions of use remained unchanged.
The FDCA was amended in 1962 to require manufacturers to demonstrate that a new drug was not only safe, but also effective for its intended uses before it could be marketed. In addition, manufacturers now had to obtain affirmative FDA approval of an NDA. A narrow grandfather clause exempted certain drugs from the new effectiveness requirement. Because this requirement applied retroactively to all drugs for which an approved NDA had been obtained before 1962, the FDA had to evaluate all such products for effectiveness. With respect to prescription drug products, this review became known as the Drug Efficacy Study Implementation (DESI) program.
At the time E-Ferol was placed on the market, the FDA’s compliance policy regarding the marketing of unapproved drug products covered by the DESI review was set forth in Compliance Policy Guide (CPG) 7132c.02, a publicly available document. CPG 7132c.02 established priorities for regulatory action on these products once final determinations were made as to their effectiveness. Under this CPG, the FDA deferred enforcement action against unapproved drugs marketed after 1962 that were identical, similar, or related to existing pre-1962 drugs of unresolved regulatory status, unless it received significant new information questioning the safety or effectiveness of a drug. [Hereinafter referred to as “ISR policy.”] Although technically not covered by CPG 7132c.02, the FDA applied the same compliance policy to drug products that were not included in the DESI review, such as single-entity vitamin E products.14
Carter argues that the FDA’s ISR policy affirmatively led him to believe that E-Ferol could be lawfully marketed without a new drug approval because it was similar or related to existing pre-1962 vitamin E products of unresolved regulatory status. Even assuming one could have reasonably considered E-Ferol to be similar or related to those products, we find no merit in Carter’s argument. The district court allowed Carter to introduce extensive evidence concerning the FDA’s policy and his alleged reliance on it, including testimony from Dr. Avrum Mark Novitch, acting FDA Commissioner from September 1983 to July 1984, as well as evidence that there were approximately 5,150 unapproved prescription drugs on the market as of June *11261984. The court also instructed the jury on Carter’s defense to the new drug counts.15 Carter does not contend that these instructions were inadequate, or that he was not permitted to fully present his defense. Under these circumstances, we can only conclude the jury rejected Carter’s assertion that in marketing E-Ferol, he relied on and was misled by the FDA’s ISR policy. This conclusion renders unavailing Carter’s reliance on United States v. Pennsylvania Indus. Chem. Corp., 411 U.S. 655, 93 S.Ct. 1804, 36 L.Ed.2d 567 (1973) (PICCO), the principal case upon which he bases his due process argument. In PICCO, the Supreme Court simply held it was error to deny a corporate criminal defendant the opportunity to present evidence that it was affirmatively misled by the responsible government agency into believing its actions were not a violation of statute. Id. at 673-75, 93 S.Ct. at 1816-17; see also Heckler v. Community Health Servs., 467 U.S. 51, 60 n. 12, 104 S.Ct. 2218, 2224 n. 12, 81 L.Ed.2d 42 (1984) (describing PICCO as holding that “criminal defendant may assert as a defense that the [gjovernment led him to believe that its [sic] conduct was legal”). The Ninth Circuit decisions cited by Carter also require a defendant to establish actual reliance on official representations before they can provide grounds for precluding or reversing a conviction. See United States v. Clegg, 846 F.2d 1221, 1224 (9th Cir.1988) (per curiam); United States v. Tallmadge, 829 F.2d 767, 774 (9th Cir.1987).
There are additional reasons why Carter’s argument must fail, aside from the jury’s rejection of his defense. Unlike the agency regulations at issue in PICCO, 411 U.S. at 670-74, 93 S.Ct. at 1814-16, the FDA’s ISR policy did not purport to modify existing statutory requirements. The policy in no way suggested that it was lawful under the FDCA to market a new drug without an approved NDA. It simply established a set of enforcement priorities in an effort to best allocate limited FDA resources. Indeed, CPG 7132c.02 was adopted by the FDA after a federal court decision overturned its prior policy of permitting certain classes of new drugs to be marketed without an approved NDA. See Hoffmann-LaRoche, Inc. v. Weinberger, 425 F.Supp. 890, 894 (D.D.C.1975) (holding that this policy contravened “the clear statutory requirement of preclearance mandated by 21 U.S.C. § 355”). CPG 7132c.02 expressly recognized that “all drugs in the DESI review are ‘new drugs’ under the law,” and stated further:
It has been decided to reaffirm that all products marketed as drugs under the DESI program are new drugs and therefore require an approved NDA or ANDA [abbreviated new drug application] for marketing. In view of this reaffirmation of this policy, it is necessary that the Agency proceed to remove from the market any current DESI-effective prescription products not subject of an approved NDA or ANDA, and to prevent in the future the marketing of any such unapproved products.
CPG 7132c.02 at 1 (April 1, 1981). Finally, we note that even if the ISR policy could somehow have been construed as making it *1127legal to market certain new drugs without an approved NDA, it certainly could not have been read as making such action lawful when done with the intent to defraud or mislead.
We reject Carter’s vagueness challenge on similar grounds. Carter argues that even if he was not affirmatively misled by the ISR policy, the policy was so vague and indefinite that he lacked adequate notice of the illegality of his conduct.16 Due process requires that “[pjenal statutes must define the criminal offense with sufficient clarity that ordinary people understand, what conduct is prohibited.” United States v. Ferguson, 776 F.2d 217, 223 (8th Cir.1985), cert. denied, 475 U.S. 1020, 106 S.Ct. 1207, 89 L.Ed.2d 320 (1986). Vagueness challenges not involving first amendment rights “are to be evaluated in light of the particular facts of each case.” United States v. Maull, 806 F.2d 1340, 1344 (8th Cir.1986), cert. denied, 480 U.S. 907, 107 S.Ct. 1352, 94 L.Ed.2d 522 (1987). Carter does not allege the evidence in this case was insufficient to establish that EFerol was a new drug and that he acted with the intent to defraud or mislead. And as discussed above, nothing in the ISR policy even remotely suggested that it was lawful to market an unapproved new drug with such an intent. Under these circumstances, the statutory requirement of fraudulent intent strips Carter’s vagueness argument of whatever merit it might have in the absence of such a mens rea requirement. See Hygrade Provision Co., Inc. v. Sherman, 266 U.S. 497, 501-03, 45 S.Ct. 141, 142-43, 69 L.Ed. 402 (1925); United States v. Peden, 556 F.2d 278, 280 (5th Cir.), cert. denied, 434 U.S. 871, 98 S.Ct. 216, 54 L.Ed.2d 150 (1977); see also Colautti v. Franklin, 439 U.S. 379, 395, 99 S.Ct. 675, 685, 58 L.Ed.2d 596 (1979) (noting Court’s long-standing recognition that the constitutionality of a vague standard is “closely related to whether that standard incorporates a requirement of mens rea ”). Even assuming the ISR policy made it unclear whether it was legal to market an unapproved new drug that was similar or related to a pre-1962 drug of unresolved regulatory status, we have no difficulty concluding § 333(a)(2) provided Carter with ample warning that it was unlawful to do so with the intent to defraud or mislead.17
B. Instructions as to Knowledge Required for FDCA Offenses
Carter and Hiland contend that their convictions on the FDCA counts must be reversed because the district court denied their request to instruct the jury that (1) knowledge that E-Ferol was an unapproved “new drug” was an essential element of the new drug offense, and (2)' knowledge that E-Ferol was “misbranded” was an essential element of the mis-branding offense. The court instructed the jury that the essential elements of the new drug offense were (1) the defendants introduced E-Ferol into interstate commerce; (2) E-Ferol was an unapproved new drug; and (3) the defendants acted with the intent to defraud or mislead. The elements instruction for the misbranding offense was the same except that the court substituted the term “misbranded” for “unapproved new drug.”
Under § 333(a)(1), neither knowledge nor intent is required for a misdemeanor viola*1128tion of § 331. United States v. Park, 421 U.S. 658, 668-73, 95 S.Ct. 1903, 1909-12, 44 L.Ed.2d 489 (1975). On its face, the only-additional element required by § 333(a)(2) for a felony violation is “the intent to defraud or mislead.” Based on the express statutory language, as well as policy considerations embodied in the FDCA, the government argues that § 333(a)(2) does not require knowledge of the facts comprising the underlying misdemeanor violation of § 331. In essence, this argument holds that a § 333(a)(2) violation consists of a completed misdemeanor plus an intent to defraud or mislead that need' not be connected to the predicate violation of § 331. The government argues in the alternative that the court’s instructions adequately directed the jury to find that the defendants had the knowledge necessary to form the requisite intent to defraud or mislead.
As is reflected in numerous mail fraud cases, knowledge of the essential nature of the alleged fraud is a component of the intent to defraud. See, e.g., United States v. Sedovic, 679 F.2d 1233, 1238-39 (8th Cir.1982); United States v. Stull, 743 F.2d 439, 442 (6th Cir.1984), cert. denied, 470 U.S. 1062, 105 S.Ct. 1779, 84 L.Ed.2d 838 (1985); United States v. Pearlstein, 576 F.2d 531, 537 (3d Cir.1978). The only decision directly addressing the meaning of § 333(a)(2)’s intent requirement adheres to this basic principle. See United States v., Industrial Laboratories Co., 456 F.2d 908, 909-11 (10th Cir.1972).18 Given the fraud that the government alleged and sought to prove in the instant case, we think it is quite clear that Carter and Hiland-could not have acted with the intent to defraud or mislead absent (1) knowledge that E-Ferol was a “drug” which was not approved by the FDA and had not been established as safe and effective for use in premature infants to treat RLF (i.e., was an unapproved “new drug”);19 and (2) knowledge that E-FeroFs labeling contained misrepresentations and misleading omissions (i.e., was “misbranded”).20 Thus, we need not decide whether knowledge of the facts constituting the misdemeanor violation of § 331 would be a separate and essential element of a § 333(a)(2) violation in a case where the defendants could have acted with the intent to defraud or mislead without such knowledge. Our inquiry here is whether the court’s instructions were adequate to prevent the jury from convicting Carter and Hiland on the FDCA counts without finding that they had the knowledge necessary for the intent required by § 333(a)(2).
A district court is afforded broad discretion in choosing the form and language of jury instructions. United States v. Jerde, 841 F.2d 818, 820 (8th Cir.1988). In assessing the adequacy of instructions, we must consider them as a whole. United States v. Figueroa, 900 F.2d 1211, 1216 (8th Cir.), cert. denied, — U.S.-, 110 S.Ct. 3228, 110 L.Ed.2d 675 (1990). Moreover, the jury *1129charge must be viewed in the context of the entire trial. Park, 421 U.S. at 674-75, 95 S.Ct. at 1912-13. In this respect, it is significant that much of the. evidence presented at trial was directed at either establishing or refuting that the defendants knew E-Ferol was marketed as a drug, knew it could not be considered as having been established as safe for use in premature infants to treat RLF, and knew its labeling contained misrepresentations and misleading omissions.
Although not a model of clarity, we conclude that when viewed as a whole and in the context of the entire trial, the district court’s instructions fairly advised the jury that Carter and Hiland could not have acted with the intent to defraud or mislead without knowledge that E-Ferol was an unapproved new drug and misbranded. In accordance with the definition of “drug” under § 321(g)(1)(B), the court instructed the jury that E-Ferol was a drug if it was intended for use in the treatment or prevention of disease, including RLF. The obvious import of this instruction was that whether E-Ferol was a drug depended on the purposes for which the defendants intended it to be used, as evidenced by the product’s labeling and marketing. See National Nutritional Foods Ass’n v. Mathews, 557 F.2d 325, 333 (2d Cir.1977) (“The vendors’ intent in selling the product to the public is the key element in [§ 321(g)(l)(B)’s] definition.”). The court instructed the jury that to act with the intent to defraud “means to act with the specific intent to deceive or cheat.” Sept. 28 Tr. at 83. In light of the evidence and arguments presented at trial, the jury could not have failed to understand that the intentional deception concerned facts' which Carter and Hiland knowingly misrepresented to or concealed from medical professionals. Cf. United States v. Huckaby, 698 F.2d 915, 919 (8th Cir.1982) (allegation in indictment that defendant acted with intent to defraud implied defendant “knew what she was concealing”), cert. denied, 460 U.S. 1070, 103 S.Ct.. 1526, 75 L.Ed.2d 948 (1983). The same connection between intent and knowledge was communicated to the jury by the court’s instruction that to act with the intent to mislead “means to act intentionally and to omit information from a statement, and thereby cause a portion of such statement to be misleading, or to intentionally conceal material facts and' thereby create a false impression by such statement.” Sept. 28 Tr. at 83. The mis-branding charge in the indictment, which was read to the jury, specifically alleged that the defendants knew of the material facts they failed to reveal in E-Ferol’s labeling. The court instructed the jury that as to the misbranding counts, Carter and Carter-Glogau “contend that they did not have knowledge of the information that was allegedly omitted from the product.” Id. at 86. The court also informed the jury of Hiland’s defense that he did not know E-Ferol was a -new drug and misbranded. In addition, the court gave instructions on specific intent, stating that if a defendant believed he was acting properly with respect to the events alleged in the indictment, even if mistaken in that belief, he could not have formulated the specific intent to violate the law.21 Finally, the guilty verdict against Carter and Hiland on the conspiracy count is a clear indication the jury did find that both of them knowingly misrepresented E-Ferol as safe and effective for intravenous use in premature infants to treat RLF.
C. Willful Blindness Instruction
Both Carter and Hiland allege that the district court committed reversible error in giving a willful blindness instruction.22 They contend that the instruction *1130permitted the jury to employ a negligence standard and was not supported by the evidence. We disagree.
In essence, a willful blindness instruction “allows the jury to impute knowledge to [the defendant] of what should be obvious to him, if it found, beyond a reasonable doubt, a conscious purpose to avoid enlightment.” United States v. Zimmerman, 832 F.2d 454, 458 (8th Cir.1987) (per curiam). As the First Circuit has noted, “[t]he purpose of the willful blindness theory is to impose criminal liability on people who, recognizing the likelihood of wrongdoing, nonetheless consciously refuse to take basic investigatory steps.” United States v. Rothrock, 806 F.2d 318, 323 (1st Cir.1986).
We find no reversible error in the language used to instruct the jury on willful blindness. In United States v. Massa, 740 F.2d 629, 642-43 (8th Cir.1984), cert. denied, 471 U.S. 1115, 105 S.Ct. 2357, 86 L.Ed.2d 258 (1985), we approved the giving of a willful blindness instruction essentially identical to the one given in this case. Viewed in the context of the entire jury charge, which included instructions on acts done knowingly, specific intent, and intent to defraud, the district court’s willful blindness instruction did not permit the jury to convict the defendants on the basis of negligent conduct. We reject Carter’s assertion that such an instruction must specifically state that a defendant has knowledge of a certain fact only if he is aware of a high probability of its existence, unless he actually believes that it does not exist. Although the Second Circuit has taken this position, see United States v. Feroz, 848 F.2d 359, 360 (2d Cir.1988) (per curiam), our decisions have not required the inclusion of such language in a willful blindness instruction. See Massa, 740 F.2d at 642-43; United States v. Graham, 739 F.2d 351, 352-53 (8th Cir.1984) (per curiam); United States v. Kershman, 555 F.2d 198, 200-01 (8th Cir.), cert. denied, 434 U.S. 892, 98 S.Ct. 268, 54 L.Ed.2d 178 (1977); accord United States v. DeVeau, 734 F.2d 1023, 1028 (5th Cir.1984) (per curiam) (rejecting argument that willful blindness instruction must include “high probability” language), cert. denied sub nom. Drobny v. United States, 469 U.S. 1158, 105 S.Ct. 906, 83 L.Ed.2d 921 (1985). In any event, Carter neither requested the additional language nor objected to its absence, and its omission certainly does not constitute plain error. See United States v. Martin, 815 F.2d 818, 824 (1st Cir.), cert. denied, 484 U.S. 825, 108 S.Ct. 89, 98 L.Ed.2d 51 (1987).
Even if adequately worded, “[a] conscious avoidance instruction is ‘properly given only when the defendant claims a lack of guilty knowledge and there are facts [in] evidence that support an inference of deliberate ignorance.’ ” United States v. White, 794 F.2d 367, 371 (8th Cir.1986) (quoting United States v. McAllister, 747 F.2d 1273, 1275 (9th Cir.1984), cert. denied, 474 U.S. 829, 106 S.Ct. 92, 88 L.Ed.2d 76 (1985)). The first requirement is clearly met here because both Carter and Hiland claimed that they did not know EFerol was dangerous and falsely labeled. The point of contention is whether there was enough evidence to justify the instruction. If the evidence in a case points solely to either actual knowledge or no knowledge of the facts in question, a willful blindness instruction should not be given. See United States v. Alvarado, 838 F.2d 311, 314 (9th Cir.), cert. denied, 487 U.S. 1222, 108 S.Ct. 2880, 101 L.Ed.2d 915, 488 U.S. 838, 109 S.Ct. 103, 102 L.Ed.2d 778 (1988); United States v. Manriquez Arbizo, 833 F.2d 244, 248-49 (10th Cir.1987). However, even where there is evidence of actual *1131knowledge, as is the case here,23 a willful blindness instruction is proper if there is sufficient evidence to support an inference of deliberate ignorance. See United States v. Perez-Padilla, 846 F.2d 1182, 1183 (9th Cir.1988) (per curiam); Manriquez Arbizo, 833 F.2d at 249. We agree, with the Seventh Circuit that in reviewing a district court’s decision to give a willful blindness instruction, we must review the evidence and any reasonable inference from that evidence in the light most favorable to the government. United States v. Herrero, 893 F.2d 1512, 1538 (7th Cir.), cert. denied, — U.S.-, 110 S.Ct. 2623, 110 L.Ed.2d 644 (1990).
Applying this standard, we find there was sufficient evidence of deliberate ignorance on the part of both Carter and Hiland to support the willful blindness instruction. Although the evidence in, this regard was not overwhelming, taken as a whole it provided the jury with a reasonable basis for inferring that if Carter and Hiland did not actually know E-Ferol was dangerous and falsely labeled, it was only because they consciously chose to be ignorant of those facts. This inference could reasonably be drawn from the evidence concerning their responses to serious indications that E-Fer-ol was associated with the illness and deaths of premature infants. See supra at 1122-23. Despite the prescient warnings he wrote in 1982, Carter personally took no action to investigate the Spokane adverse reaction report of January 1984, and did not inquire into the details of Madison’s investigation. Instead, he went ahead and shipped the third and final lot of E-Ferol to OJF. Three weeks before the Spokane incident, Madison sent Carter-a drug problem report from a pharmacist expressing concerns about the large amount of polysor-bates in E-Ferol. And in March 1984, Madison sent Carter another drug problem report from a pharmacist concerning the fact that E-Ferol’s labeling could be construed as representing the product to be appropriate for the treatment of RLF in premature infants without any studies to support’ this claim. As with the Spokane incident, Carter also did not follow up on either of these reports because he did not think a response from him was necessary. Yet, at trial he admitted the FDA had repeatedly told him prior to E-Ferol that as the manufacturer, he was responsible for products custom made for a specific distributor.
Hiland’s responses to adverse reaction reports likewise provided a basis for inferring deliberate ignorance. There was evidence that he entrusted the entire investigation into the Spokane report to Madison knowing that Madison lacked the capacity to determine the report’s validity. There was also evidence that Hiland had previously been informed of the adverse reaction report from Hawaii. Several weeks after the Spokane report, Hiland was told by OJF’s national sales manager, albeit with doubts about the physician’s competency, that Dr. Bodenstein was continuing to hold E-Ferol responsible for the infants’ deaths. Most significant is the fact that Hiland took no action to investigate the March 1984 report of four infant deaths in Tennessee, which also shed light on his attitude in January 1984. Four of the six FDCA counts were based on shipments of E-Ferol made after OJF received the Tennessee report.
Finally, we think it is relevant that- neither Carter nor Hiland ever consulted the FDA to find out if it was necessary to test E-Ferol for safety.24 Although Carter and Hiland were under no legal duty to make such an inquiry, we believe their decision not to consult the FDA constituted at least some evidence of deliberate ignorance. See United States v. Vesterso, 828 F.2d 1234, 1244-45 (8th Cir.1987).
D. Burden of Proof
Hiland sought a theory of defense instruction informing the jury that it must *1132acquit him on the FDCA counts if the government failed to prove beyond a reasonable doubt that he knew E-Ferol was an unapproved new drug or a misbranded drug. The district court instead instructed the jury that if Hiland “did not know that E-Ferol was an unapproved new drug or a misbranded drug, and had no intent to defraud or mislead, you must find Mr. Hi-land not guilty of the charges.” Sept. 28 Tr. at 81-82 (emphasis added).25 Hiland argues that reversal of his FDCA convictions is required because the court’s alteration of his theory of defense instruction impermissibly shifted the burden of proof to him and permitted acquittal only if the jury found he lacked both knowledge and intent. We find no merit in these arguments.
First, the challenged instruction itself did not “speak directly or indirectly to burden of proof.” Nelson v. Solem, 640 F.2d 138, 136 (8th Cir.1981). It did not state that Hiland had to disprove the presence of knowledge and intent in order to be acquitted. Second, any possibility that the jury would apply the instruction in this way was negated by other instructions which specifically addressed the burden of proof. The court instructed the jury that the government had the burden of proving the defendant’s guilt beyond a reasonable doubt, bore this burden throughout the trial, and was required to prove each element of an offense beyond a reasonable doubt. The court also instructed the jury that the defendant was presumed innocent until proven guilty and never had the burden of calling any witness or producing any evidence. Viewed in light of the jury charge as a whole, the district court’s instruction regarding Hiland’s theory of defense clearly did not shift the burden of proof to him to prove his innocence.26
Contrary to Hiland’s assertion, the instruction did not necessarily imply its converse; that is, it did not tell the jury that it should convict Hiland on the FDCA counts if he had the requisite knowledge but lacked the intent to defraud or mislead, and vice versa. To the extent such confusion could have been engendered by the court’s use of the conjunctive, it was cured by the instructions as a whole, which adequately apprised the jury of the need to find both the intent to defraud or mislead and the knowledge necessary for that intent before it could convict Hiland.
E. Cross-Examination of Hiland
Hiland alleges that the district court erred in allowing the government to cross-examine him regarding certain instances prior to E-Ferol in which OJF had marketed drugs without FDA approval. Given the district court’s broad discretion in this area, we find that the challenged cross-examination was permissible under Fed.R.Evid. 611(b), which provides that “[cjross-examination should be limited to the subject matter of the direct examination and matters affecting the credibility of the witness.”27
The principles guiding our review are well settled. “If a defendant takes the stand, his credibility is placed in issue, and the [government is entitled to attack it by cross-examination.” United States v. Wallace, 722 F.2d 415, 416 (8th Cir.1983). “Cross-examination may embrace any matter germane to direct examination, qualifying or destroying it, or tending to elucidate, modify, explain, contradict or rebut testi*1133mony given by the witness.” Roberts v. Hollocher, 664 F.2d 200, 203 (8th Cir.1981). “The permissible extent of cross-examination is a matter within the broad discretion of the district court.” United States v. Schepp, 746 F.2d 406, 410 (8th Cir.1984), cert. denied, 469 U.S. 1215, 105 S.Ct. 1190, 84 L.Ed.2d 336 (1985). Accordingly, we will reverse the court’s ruling only for abuse of discretion. United States v. Brown, 794 F.2d 365, 366 (8th Cir.1986).
On direct examination, Hiland testified that he had very little involvement with regulatory matters at OJF and had little or no knowledge of when a product required FDA approval as a new drug. He repeatedly insisted that for these reasons he had relied on Madison’s determination of whether FDA approval was needed for E-Ferol. The district court ruled that by so testifying, Hiland had opened the door to cross-examination about his participation in incidents prior to E-Ferol that tended to contradict his professed ignorance of when FDA approval for a product was necessary. The court accordingly permitted the government to question Hiland regarding the sale of Dalalone P.R., a combination steroid drug that was the subject of an FDA seizure action in 1981 because OJF had marketed it without approval. This part of the cross-examination centered on two documents. The first was a letter written by Hiland to his superior at OJF’s parent company recommending that OJF defend the continued distribution of Dalal-one P.R. by arguing that it was not a new drug. The second was a 1981 memorandum, written by Madison at Hiland’s request, which discussed the history of Dalal-one P.R. and stated that the company was taking a risk by marketing it without a new drug approval. The court also allowed the government to question Hiland about a 1981 memorandum regarding a telephone conversation he had with a manufacturer of some of OJF’s injectable products. The conversation concerned the fact that some of the products did not have new drug approval. In the memorandum, Hiland described the potential loss to OJF if the FDA caused the manufacturer to stop making these products, and wrote that he doubted the manufacturer would succeed with its argument that the products did not require new drug approval.28
The line of inquiry permitted by the court was relevant to showing that Hiland was better informed than he claimed to be regarding OJF’s regulatory affairs in general, and in particular about when a product would or might be considered a new drug by the FDA. The Dalalone P.R. example had special relevance in that it tended to show Hiland was aware prior to E-Ferol that a product comprised of two established and well-known ingredients might very well constitute a new drug.29 We do not accept Hiland’s contention that his involvement in the controversy with the FDA over Dalalone P.R. was irrelevant since this product was not a single-entity vitamin product like E-Ferol. Despite this and related arguments, we cannot say that the district court abused its discretion in determining that the challenged cross-examination was within the scope of direct examination, relevant, and not unfairly prejudicial.30
*1134F. Prejudicial Medical Testimony
Both Carter and Hiland contend that the district court abused its discretion in admitting the testimony of several physicians and pharmacists regarding the effects E-Ferol had on premature infants. The challenged testimony consisted primarily of descriptions of the unusual pattern of symptoms observed in premature infants who had received the drug, and concerned adverse reaction episodes of which Carter and Hiland were not actually aware prior to the recall of E-Ferol.31 Carter and Hiland argue that (1) the court should have required the government to accept a stipulation they offered because it would have eliminated any need for the testimony, and (2) the testimony was inadmissible under Fed.R. Evid. 403 due to its limited relevance and prejudicial effect. Neither argument persuades us that the court committed error.
The challenged medical testimony helped to show that E-Ferol was dangerous when administered in accordance with dosage directions and other recommendations for use in its labeling. It was therefore plainly relevant to establishing that E-Ferol was both misbranded and a new drug, essential elements of the government’s case.32
At trial, the defendants offered to stipulate “that at some time after April 9, 1984 it was determined by others that E-Ferol Aqueous Solution was dangerous to health in certain cases.” The proposed stipulation said nothing about whether E-Ferol was dangerous when used as labeled. It is well established in this circuit that as a general rule, the government is not bound by a defendant’s offer to stipulate. United States v. Bass, 794 F.2d 1305, 1312 n. 6 (8th Cir.), cert. denied sub nom. Price v. United States, 479 U.S. 869, 107 S.Ct. 233, 93 L.Ed.2d 159 (1986); United States v. Ellison, 793 F.2d 942, 949 (8th Cir.), cert. denied, 479 U.S. 937, 107 S.Ct. 415, 93 L.Ed.2d 366 (1986); United States v. Peltier, 585 F.2d 314, 324 (8th Cir.1978), cert. denied, 440 U.S. 945, 99 S.Ct. 1422, 59 L.Ed.2d 634 (1979). The rationale for this rule is “to enable the government to present to the jury a complete picture of the events constituting the crime charged. ‘To substitute for such a picture a naked admission might ... rob the evidence of much of its fair and legitimate weight.’ ” Ellison, 793 F.2d at 949 (quoting Peltier, 585 F.2d at 324). Nevertheless, the general rule is subject to the strictures of Rule 403. Id. Thus, “a proper [Rjule 403 balancing analysis will incorporate some assessment of the need for the allegedly prejudicial information in light of a valid stipulation.” Bass, 794 F.2d at 1312 n. 6. A critical aspect of this assessment is whether the challenged evidence is probative of issues other than the one to which the defendants offered to stipulate. See id. In the instant case, acceptance of the defendants’ proposed stipulation would not have satisfied the government’s burden of proving that E-Ferol was not only dangerous, but dangerous when used as recommended in its labeling. Given this deficiency in the proposed stipulation and the rationale underlying the general rule regarding such offers, we conclude that the district court did not err in refusing to exclude the challenged medical testimony “merely because of [the defendants’] offer to stipulate.” Peltier, 585 F.2d at 325.
As relevant evidence, this testimony was inherently prejudicial in the sense *1135of being detrimental to the defendants’ case. Rule 403 is not directed at this type of prejudice, but rather protects only against unfair prejudice. See Wade v. Haynes, 663 F.2d 778, 783 (8th Cir.1981), aff'd sub nom. Smith v. Wade, 461 U.S. 30, 103 S.Ct. 1625, 75 L.Ed.2d 632 (1983). Evidence is unfairly prejudicial to the extent it creates “ ‘an undue tendency to suggest decision on an improper basis, commonly, though not necessarily, an emotional one.’ ” United States v. Weddell, 890 F.2d 106, 108 (8th Cir.1989) (quoting Rule 403 advisory committee’s note). Testimony regarding the effects of E-Ferol undoubtedly presented the potential for such prejudice. However, Rule 403 permits the exclusion of evidence only if its probative value is “substantially outweighed” by the danger of unfair prejudice. United States v. Abodeely, 801 F.2d 1020, 1025 (8th Cir.1986). In reviewing a discretionary decision not to exclude evidence under Rule 403, “we give great deference to the trial judge who saw and heard the evidence.” Peltier, 585 F.2d at 321. We will not reverse that decision absent a clear and prejudicial abuse of discretion. Abodeely, 801 F.2d at 1026.
Although Carter and Hiland make much of references in the medical testimony to “oozing,” “bleeding,” and abdominal “bloating,” the record reveals that such terminology was used in only a few isolated instances. For the most part, the witnesses in question limited themselves to strictly clinical terms when describing the symptoms they had observed, and their testimony did not unduly emphasize E-Ferol’s association with certain infant deaths. We are satisfied from the record that the medical testimony was carefully controlled and not presented in an inflammatory manner. We note also that to a certain extent, some of the challenged testimony was intertwined with other relevant testimony of the witnesses, such as their explanations of the purposes for which they had used E-Ferol based on its labeling. The fact that E-Fer-ol had been associated with infant deaths was disclosed to the jury by other properly admitted evidence. Indeed, it was Carter who introduced into evidence several documents containing a statement that E-Fer-ol had been associated with the deaths of thirty-eight infants. Finally, we are impressed by the diligent efforts of the district court to limit the potential for unfair prejudice. During the trial, the court twice cautioned the jury that the defendants were not being tried for harming or causing the death of infants. The court reiterated this same point in no uncertain terms in its final instructions to the jury.33 In sum, after a careful review of the challenged medical testimony, we cannot say the district court committed a clear abuse of discretion in determining that the danger of unfair prejudice presented by this testimony did not substantially outweigh its probative value.
G. Prosecutorial Misconduct
Carter contends that the following statement made at the conclusion of the prosecution’s closing argument was improper and requires reversal:
In sum, you’ve heard how Hoffman[n-]LaRoche was testing their product. They were doing animal studies and they were doing clinical studies. I submit that these defendants were testing their product too, on the babies in Spokane, Tennessee, in Cincinnati, and other places.
Although the final sentence of this statement was improper, defense counsel failed to object. Our review is therefore limited to determining whether the prosecutor's remark amounted to plain error. See Fed.R.Crim.P. 52(b); United States v. Skarda, 845 F.2d 1508, 1511 (8th Cir.1988). Under this standard, reversal is warranted only if the remark was “such as to undermine the fundamental fairness of the trial and con*1136tribute to a miscarriage of justice.” United States v. Young, 470 U.S. 1, 16, 105 S.Ct. 1038, 1047, 84 L.Ed.2d 1 (1985). As noted previously, the district court emphasized to the jury that the defendants were not on trial for injuring infants. The court also instructed the jury that the attorneys’ closing arguments were not evidence. Moreover, the evidence admitted against Carter was substantial and persuasive. We are satisfied that, viewed in the context of the entire trial, the prosecutor’s improper remark did not contribute to a miscarriage of justice.
H. Supplemental Charge, Jury Poll, and Unanimity of Verdict
The indictment alleged that E-Ferol had been misbranded in five different ways. The first was that the labeling for E-Ferol was “false and misleading” because it failed to reveal one or more of the “material facts” identified in the indictment.34 The jury instructions characterized these five grounds as options to be considered individually. The jury was told that, as to each defendant and each misbranding count, it had to specify at least one of the five options in order to return a guilty verdict. However, the final misbranding instructions described the first option as simply “false and misleading in any particular,”35 and did not require the jury to indicate which of the material facts listed in the indictment had not been revealed. During discussions with counsel while the jury deliberated, the district court recognized that a misbranding verdict specifying option one might not be sufficiently precise to ensure unanimity as to the way in which E-Ferol had been misbranded. Accordingly, the court informed counsel that it would direct the jury to deliberate further if such a verdict were returned. The jury subsequently returned with a misbranding verdict against Hiland that specified only option one.36 The court told the jury that its verdict was ambiguous due to the wording of the instructions, and directed it to retire to the jury room to see if it could unanimously agree on another misbranding option as to Hiland. The jury returned shortly thereafter with a verdict specifying another option.37
Before giving its supplemental charge on the misbranding counts against Hiland, the court read the other verdicts to the jury and asked it as a body whether the various verdicts were correct. The jury confirmed them by silence. After the second mis-branding verdict was returned, the court polled the jurors individually but inadvertently omitted the mail fraud verdict against Hiland. Earlier, the court had neglected to augment its general unanimity instruction with the specific unanimity instruction requested by Hiland, which would have told the jury that its verdict must be unanimous with respect to each defendant on each count.
Hiland contends that the jury’s second verdict on the misbranding counts cannot stand. He suggests that the district court had no authority to resubmit these counts to the jury and argues that doing so was both coercive and a violation of the fifth amendment’s prohibition against double jeopardy. We are not persuaded by these arguments.
Under the sixth amendment, a federal criminal defendant has a non-waivable right to a unanimous jury verdict. United States v. Eagle Elk, 820 F.2d 959, 961 (8th *1137Cir.), cert. denied, 484 U.S. 867, 108 S.Ct. 191, 98 L.Ed.2d 143 (1987). The district court was justified in its concern that the jury’s first misbranding verdict against Hi-land might not satisfy this constitutional requirement because of the possibility that the jurors did not agree on the particular way in which E-Ferol had been misbrand-ed. See id. (sixth amendment requires that jurors be in “substantial agreement as to the nature of the defendant’s guilty act”); McKoy v. North Carolina, — U.S. -, 110 S.Ct. 1227, 1237 n. 5, 108 L.Ed.2d 369 (1990) (Blackmun, J., concurring) (must be “unanimous agreement as to the nature of the defendant’s violation, not simply the fact that a violation has occurred”). Indeed, Hiland asserts that the jury’s first misbranding verdict was defective for this very reason.38 Nevertheless, Hiland suggests that the court had no authority to resubmit the misbranding counts to the jury in the absence of a jury poll indicating that the first verdict was not unanimous. A federal district court “has authority to require redeliberation in cases in which there is uncertainty, contingency, or ambiguity regarding the jury’s verdict.” United States v. Rastelli, 870 F.2d 822, 835 (2d Cir.), cert. denied sub nom. Agar v. United States, — U.S.-, 110 S.Ct. 515, 107 L.Ed.2d 516 (1989). Rule 31(d) of the Federal Rules of Criminal Procedure expressly authorizes the court to direct the jury to retire for further deliberation if a poll shows a lack of unanimity.39 However, “the language of [Rjule 31(d) does not delimit the only circumstance in which a trial judge may require redeliberation.” Rastelli, 870 F.2d at 835; see also United States v. Mears, 614 F.2d 1175, 1179 (8th Cir.) (where verdict of “not guilty” read in open court and jury foreman then told court verdict was incorrectly signed, court did not err in directing jury to resume deliberations), cert. denied, 446 U.S. 945, 100 S.Ct. 2174, 64 L.Ed.2d 801 (1980). Furthermore, in all cases not provided for by rule, the district court may proceed “in any manner not inconsistent with” the Federal Rules or local rules. Fed.R.Crim.P. 57; see also United States v. Jerry, 487 F.2d 600, 604 (3d Cir.1973) (“[njothing in the Rules limits the power of the court to correct mistakes made in its handling of a case so long as the court’s jurisdiction continues”). We conclude that the district court had authority to direct the jury to deliberate further on the misbranding counts against Hiland" even though the only ambiguity in the jury’s first verdict was due to the court’s instructions.
“Any criminal defendant ... being tried by a jury is entitled to the uncoerced verdict of that body.” Lowenfield v. Phelps, 484 U.S. 231, 241, 108 S.Ct. 546, 552, 98 L.Ed.2d 568 (1988)." Hiland argues that the second misbranding verdict was coerced because the jury received the court’s supplemental charge late on a Friday afternoon after deliberating for two days, and then returned with its second verdict approximately ten minutes after receiving the charge. The fact that the jury returned so soon with its verdict “suggests the possibility of coercion.” Id. at 240, 108 S.Ct. at 552; see also United States v. Webb, 816 F.2d 1263, 1267 (8th Cir.1987) (finding of coercive effect “supported by” fact that verdict returned only fifteen minutes after Allen charge). However, we are certainly not required to draw such an inference, as is shown by our decisions addressing supplemental Allen charges.40 See United States v. Dawkins, 562 F.2d 567, 570 n. 4 (8th Cir.1977) (per curiam) (citing cases). We consider it significant that defense counsel did not object to the supplemental charge.41 “[S]uch an omis*1138sion indicates that the potential for coercion argued now was not apparent to one on the spot.” Lowenfield, 484 U.S. at 240, 108 S.Ct. at 552; see also Amos v. United States, 496 F.2d 1269, 1273 (8th Cir.) (lack of objection by defense counsel to court’s procedures in polling jury and directing jury to resume deliberations “permits an inference that the procedures utilized did not appear coercive at the time”), cert. denied, 419 U.S. 896, 95 S.Ct. 174, 42 L.Ed.2d 140 (1974). The fact that the charge was not directed at jurors holding a minority view, if indeed there were any, further reduces the likelihood of coercion. See Lowenfield, 484 U.S. at 237-38, 108 S.Ct. at 550-51; United States v. Porter, 881 F.2d 878, 889 (10th Cir.), cert. denied, — U.S. -, 110 S.Ct. 348, 107 L.Ed.2d 336 (1989). In instructing the jurors to resume deliberation on the misbranding counts against Hiland, the district court stressed that they were free to find no other misbranding option applied to Hiland. Having considered the district court’s supplemental charge “ ‘in its context and under all the circumstances,’ ” Lowenfield, 484 U.S. at 237, 108 S.Ct. at 550 (quoting Jenkins v. United States, 380 U.S. 445, 446, 85 S.Ct. 1059, 1060, 13 L.Ed.2d 957 (1965) (per curiam)), we find that the jury’s second misbranding verdict was not coerced.
Hiland argues that the double jeopardy clause prohibited resubmission of the misbranding counts to the jury because by specifying only option one, the jury’s first verdict implicitly acquitted him of misbranding on the basis of any of the other four options. Hiland relies on Green v. United States, 355 U.S. 184, 190, 78 S.Ct. 221, 225, 2 L.Ed.2d 199 (1957), which held that a verdict of guilty on a lesser included offense constituted an “implicit acquittal” of the defendant on the greater offense charged, barring retrial on that offense. We find Green inapposite because the misbranding options were equivalent in that they were simply presented as the various ways in which E-Ferol was alleged to have been misbranded. The jury was instructed that it need agree on only one of the options in order to find a defendant guilty of misbranding. It was not instructed that it had to indicate each of the options on which it unanimously agreed. Thus, there is no reason to conclude that the jury’s first mis-branding verdict necessarily reflected a refusal to find Hiland guilty of that offense on the basis of any of the options not designated by the jury. Furthermore, a verdict is not final for purposes of double jeopardy simply because it is announced by the jury foreman in open court. United States v. Love, 597 F.2d 81, 84-85 (6th Cir.1979). If not accepted by the trial court, a verdict is not final for purposes of double jeopardy. See United States v. Chinchic, 655 F.2d 547, 550 (4th Cir.1981). The record here establishes that the district court expressly refrained from accepting the verdicts until after the jury had returned from its redeliberation on the mis-branding counts against Hiland.
Hiland contends that the district court’s failure to poll the jurors individually on the mail fraud verdict against him requires automatic reversal of that verdict. The right to poll the jury is provided by Fed.R.Crim.P. 31(d), but is not of constitutional dimension. Government of the Virgin Islands v. Hercules, 875 F.2d 414, 417 (3d Cir.1989); United States v. Carter, 772 F.2d 66, 67 (4th Cir.1985); United States v. Beldin, 737 F.2d 450, 455 (5th Cir.), cert. denied, 469 U.S. 1075, 105 S.Ct. 572, 83 L.Ed.2d 512 (1984); United States v. Shepherd, 576 F.2d 719, 724 (7th Cir.), cert. denied, 439 U.S. 852, 99 S.Ct. 158, 58 L.Ed.2d 155 (1978); Jaca Hernandez v. Delgado, 375 F.2d 584, 585 (1st Cir.1967). Denial of a timely request for a poll under Rule 31(d) is reversible error. Hercules, 875 F.2d at 417 n. 5, 419; Shepherd, 576 F.2d at 724. It is also reversible error not to allow the defendant a reasonable opportunity to make such a request. Miranda v. United States, 255 F.2d 9, 18 (1st Cir.1958). On the other hand, it is well settled that a *1139poll under Rule 31(d) is not required unless requested and is waived if the request is not timely. Beldin, 737 F.2d at 455; United States v. Hockridge, 573 F.2d 752, 759 n. 19 (2d Cir.), cert. denied, 439 U.S. 821, 99 S.Ct. 86, 58 L.Ed.2d 112 (1978); United States v. Marr, 428 F.2d 614, 615 (7th Cir.1970); Mull v. United States, 402 F.2d 571, 574 (9th Cir.1968), cert. denied, 393 U.S. 1107, 89 S.Ct. 917, 21 L.Ed.2d 804 (1969); United States v. Neal, 365 F.2d 188, 190 (6th Cir.1966); Miranda, 255 F.2d at 17. Silence constitutes a waiver if adequate time is allowed for a request. Beldin, 737 F.2d at 455. '
Hiland asserts that he specifically requested a poll, citing an exchange which took place after the court had asked the jury collectively whether its verdicts were correct and directed it to retire for further deliberation on the misbranding counts against Hiland:
A VOICE [presumably Hiland's counsel]: Your Honor, can I ask a question?
THE COURT: Yes.
A VOICE: Is that the way the Court customarily polls the jury-
THE COURT: No, I still think the formal polling individually of the jury, but with so many different things here, we'd have to poll it 25 times 8, and that was too much, so I read that at first and I'm going to ask them to po'l over again when they come back, individually to respond.
Sept. 30 Tr. at 9. Counsel’s question can hardly be characterized as a request for a poll. More importantly, when the court later inadvertently omitted the mail fraud verdict against Hiland in polling each juror individually, Hiland neither objected nor brought the oversight to the court’s attention. The record reveals that Hiland had ample time to do so. Although not at fault for the court’s oversight, “counsel failed to take reasonable, available steps” to remedy the situation. Beldin, 737 F.2d at 455. “The time to protest has passed.” Id. We hold that Hiland waived his right to a poll on the mail fraud verdict.42
Hiland proposed a specific instruction informing the jury that its verdict must be unanimous with respect to each defendant on each count. The district court expressed willingness to give this instruction in addition to its general unanimity instruction, but ultimately failed to do so despite several requests and a timely objection by Hiland.43 Hiland argues that the court’s failure to give the requested specific unanimity instruction constitutes reversible error. Because the court did give such an instruction on the mis-branding counts and in its supplemental charge to the jury, this allegation of error has no merit with respect to the second misbranding verdict.
It is well settled that a general unanimity instruction is usually sufficient to protect a defendant's sixth amendment right to a unanimous verdict. See, e.g., United States v. Hernandez-Escarsega, 886 F.2d 1560, 1572 (9th Cir.1989), cert. denied,-U.S.-, 110 S.Ct. 3237, 111 L.Ed.2d 748 (1990); United States v. Phillips, 869 F.2d 1361, 1866-67 (10th Cir.1988), cert. denied, - U.S.-, 109 S.Ct. 2074, 104 L.Ed.2d 638 (1989); United States v. Murray, 618 F.2d 892, 898 (2d Cir.1980). As an exception to this rule, courts have held that a specific unanimity instruction is required in cases where there is a genuine risk of jury confusion. See, e.g., United States v. Bryan, 868 F.2d 1032, 1039 (9th Cir.), cert. denied, U.S.-, 110 S.Ct. *1140167, 107 L.Ed.2d 124 (1989); United States v. Duncan, 850 F.2d 1104, 1114 (6th Cir.1988), cert. denied sub nom. Downing v. United States,—U.S.-, 110 S.Ct. 732, 107 L.Ed.2d 751 (1990).44 The only danger of confusion alleged by Hiland is that the jury may have believed it only needed to agree unanimously on a general verdict, but not on the individual counts. Citing no case law that supports his position, Hiland urges us to hold that a specific unanimity instruction such as he proposed must be given whenever it is requested in a case involving multiple defendants and counts. Although such a practice is preferable, we are not prepared to say that reversal is required whenever it is not followed. As we have emphasized in a related context, “[t]he mere fact ... that an instruction could conceivably permit a jury to reach a non-unanimous verdict is not sufficient to require reversal when the jury has been instructed that it must reach a unanimous verdict.” Berrisford v. Wood, 826 F.2d 747, 754 (8th Cir.1987) (quoting Fryer v. Nix, 775 F.2d 979, 992 (8th Cir.1985)), cert. denied, 484 U.S. 1016, 108 S.Ct. 722, 98 L.Ed.2d 671 (1988).
Other courts have recognized a number of factors that may result in a genuine danger of jury confusion, necessitating a specific unanimity instruction. See, e.g., Duncan, 850 F.2d at 1114. The only one of these factors that Hiland suggests is present here is factual complexity of the evidence.45 The evidence in this case can indeed be characterized as complex, if only because of its volume. However, Hiland does not attempt to explain how this complexity translated into a genuine risk that the jury would be confused as to whether it had to agree unanimously with respect to each defendant on each count, or only on an overall verdict. We are therefore not inclined to hold that such a risk existed in this case.
We need not definitively decide this issue, however, because if the district court erred in failing to give Hiland’s proposed unanimity instruction, the error was harmless beyond a reasonable doubt. See Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 828, 17 L.Ed.2d 705 (1967). Before polling the jurors individually, the court told them that it was doing so “[bjecause a verdict must be the verdict of all the jury and on [sic] each individual juror.” Sept. 30 Tr. at 10. The poll showed that the verdicts against Hiland were unanimous on the conspiracy and new drug counts, curing the alleged error as to those counts. See United States ex rel. Riffert v. Rundle, 464 F.2d 1348, 1351 (3d Cir.1972) (even if failure to instruct jury that unanimity required was plain error, error was harmless because subsequent jury poll indicated verdict was unanimous), cert. denied, 415 U.S. 927, 94 S.Ct. 1434, 39 L.Ed.2d 484 (1974); cf. United States v. Flaherty, 668 F.2d 566, 600 (1st Cir.1981) (use of phrase “collective judgment” in reasonable doubt instruction not reversible error in light of other instructions and jury poll which confirmed verdict was unanimous). Additional circumstances support this conclusion, and convince us that the verdict against Hiland on the mail fraud counts was also unanimous. The court instructed the jurors that it was their duty to “give separate, personal consideration to the case of each individual defendant,” and specifically told them that “[t]he fact that you may find one of the accused guilty or not guilty of one charge should not control your verdict as to another defendant or another charge.” Sept. 28 Tr. at 15. The *1141jury was given a separate verdict form for each defendant with a blank line to mark “guilty” or “not guilty” as to each count. That the jury gave individual consideration to Hiland with respect to the mail fraud counts is further evidenced by its acquittal of the other defendants on those charges. See Shepherd, 576 F.2d at 725 (jury “sophistication” evidenced by acquittal of one defendant and conviction of others reduced likelihood of non-unanimity). Moreover, when questioning the jury as a whole, the court indicated it would consider the jurors’ silence as affirming that they had all found each defendant guilty as stated in the verdicts.46 After the jury returned from its final deliberation, the court informed the jurors that it would repeat these verdicts before polling them individually. Notwithstanding the court’s ensuing failure to repeat the mail fraud verdict against Hiland, we consider it highly unlikely that the jurors did not think they were also confirming this verdict by their individual answers.47 Under these circumstances, we are satisfied beyond a reasonable doubt that the verdicts against Hiland were unanimous on each count.
I. Costs of Prosecution
In addition to fining Carter-Glogau the maximum amount permitted by statute, $10,000 on each count for a total of $130,-000, the district court ordered the corporation to pay $100,000 toward the costs of prosecution. The court later denied as moot Carter-Glogau’s motion to vacate this portion of the sentence. Carter-Glogau contends that the court exceeded its authority to assess costs of prosecution.
Under 28 U.S.C. § 1918(b),48 the district court has the discretionary authority to tax the costs of prosecution against the defendant in any non-capital case. United States v. Burchinal, 657 F.2d 985, 997 (8th Cir.), cert. denied, 454 U.S. 1086, 102 S.Ct. 646, 70 L.Ed.2d 622 (1981). Contrary to Carter-Glogau’s assertion, § 1918(b) permits the court to assess costs of prosecution in addition to the maximum fine permitted by the statute under which the defendant is convicted, even if that statute does not specifically provide for such an assessment. Id. at 998; see also United States v. Gering, 716 F.2d 615, 626 (9th Cir.1983) (following Burchinal).49 The court may not, however, assess costs attributable solely to the prosecution of counts on which the defendant was acquitted. United States v. DeBrouse, 652 F.2d 383, 391 (4th Cir.1981); see also United States v. Palmer, 809 F.2d 1504, 1508-09 (11th Cir.1987) (applying 26 U.S.C. § 7203). Nor may the court assess costs associated exclusively with the unsuccessful prosecution of a codefendant. See United States v. Fowler, 794 F.2d 1446, 1450 (9th Cir. *11421986) (applying § 7203), cert. denied, 479 U.S. 1094, 107 S.Ct. 1309, 94 L.Ed.2d 153 (1987). Generally, the district court’s taxation of costs may be set aside only upon a finding of abuse of discretion. United States v. Dougherty, 810 F.2d 763, 773 (8th Cir.1987).
Section 1918(b) must be read in conjunction with 28 U.S.C. § 1920, which lists the expenses that may be taxed as costs. Section 1920 applies not only in civil but also criminal cases. United States v. Procario, 361 F.2d 683, 684 (2d Cir.1966) (per curiam). Absent explicit statutory or contractual authorization to the contrary, federal district courts may tax as costs only those expenses listed in § 1920. Crawford Fitting Co. v. J. T. Gibbons, Inc., 482 U.S. 437, 107 S.Ct. 2494, 96 L.Ed.2d 385 (1987). Just as Fed.R.Civ.P. 54(d) does not provide such authorization, id. at 441-45, 107 S.Ct. at 2497-99, we find that the discretion granted by § 1918(b) does not authorize federal district courts to order a criminal defendant to pay costs not enumerated in § 1920. See Gering, 716 F.2d at 626 (taxation of items as costs under § 1918(b) must be authorized by statute); United States v. Pommerening, 500 F.2d 92, 101-02 (10th Cir.) (taxable costs in criminal prosecutions dependent upon statutory provisions), cert. denied, 419 U.S. 1088, 95 S.Ct. 678, 42 L.Ed.2d 680 (1974). The Crawford Court concluded that § 1920 defines the term “costs” as used in Rule 54(d). 482 U.S. at 441, 107 S.Ct. at 2497. We believe that § 1920 likewise defines the term “costs of prosecution” as used in § 1918(b). See 21 U.S.C. § 844(a) (providing that person who violates this subsection “shall be fined the reasonable costs of the investigation and prosecution of the offense, including the costs of prosecution of an offense as defined in sections 1918 and 1920 of Title 28 ”) (emphasis added); United States v. Vaughn, 636 F.2d 921, 922 (4th Cir.1980) (“costs of prosecution” referred to in 26 U.S.C. § 7201 are those set forth in § 1920). To conclude otherwise would render § 1920 superfluous. See Crawford, 482 U.S. at 441, 107 S.Ct. at 2497. Therefore, since § 1920 does not include the costs of investigation leading to indictment, Carter-Glogau is correct in contending that it may not be assessed those costs. See Vaughn, 636 F.2d at 922.50
Section 1920 states that “[a] bill of costs shall be filed in the case.” This language is mandatory. See Mason v. Belieu, 543 F.2d 215, 222 (D.C.Cir.), cert. denied, 429 U.S. 852, 97 S.Ct. 144, 50 L.Ed.2d 127 (1976). In addition, 28 U.S.C. § 1924 requires that cost items be verified by affidavit. See Wahl v. Carrier Mfg. Co., Inc., 511 F.2d 209, 216 (7th Cir.1975). The government did not file a verified bill of costs in this case. As a result, it is not possible to determine whether the $100,000 assessed by the district court includes costs of investigation, other costs not listed in § 1920, or costs attributable solely to the unsuccessful prosecution of codefendants or to counts on which Carter-Glogau was acquitted. Accordingly, we find that the district court abused its discretion in ordering Carter-Glogau to pay $100,000 toward the costs of prosecution.
III.
For the foregoing reasons, the convictions are affirmed. The district court’s assessment of costs against Carter-Glogau is vacated and the case remanded for a rede-termination of taxable costs in accordance with this opinion and subject to any valid objections or defenses asserted by Carter-Glogau.
. The Honorable Clyde S. Cahill, United States District Judge for the Eastern District of Missouri.
. At the time of the defendants' indictment in 1987, the language of § 333(a)(2) was set forth in § 333(b). In 1988, § 333(b) was redesignated as § 333(a)(2), which is the citation we shall use throughout this opinion.
. The plea was placed under seal and not accepted by the court until the trial had concluded. OJF was fined $115,000 and ordered to pay $100,000 in costs.
. Madison was sentenced to eight years imprisonment, all but six months of which was suspended, and fined $12,000. In lieu of the fine, Madison could pay $10,000 to the neonatal intensive care units of certain hospitals victimized by the sale of E-Ferol.
. The appellants do not directly challenge their conspiracy convictions. They argue only that these convictions must also be overturned if the FDCA convictions are reversed.
. Carter-Glogau joins in all of the arguments made by Carter. To avoid undue repetition, we will refer only to Carter when discussing their arguments.
. Carter and Hiland were each sentenced to nine years imprisonment, all but six months of which was suspended, and fined $130,000. In lieu of the fine, each could pay $65,000 within one year of judgment to the neonatal intensive care units of certain hospitals victimized by the sale of E-Ferol. Both Carter and Hiland were denied release on bail pending appeal.
. Carter-Glogau was also fined $130,000.
. There was also a multivitamin intravenous injection, containing a small amount of vitamin E, that was used as a nutritional supplement.
. There was trial testimony indicating that it is not uncommon for physicians, in the exercise of their professional judgment, to use a prescription pharmaceutical product for uses other than those recommended or suggested in the product's labeling.
. The administration of oxygen to premature infants was thought to be the primary cause of RLF.
. The Tennessee hospital did not contact OJF directly, but rather reported the infant deaths to an OJF salesman, who then put Sandlin in contact with Madison.
. Specifically, the indictment charged that E-Ferol was misbranded:
A.Under [21 U.S.C. §§ 352(a), 321(n) ] in that the labeling for said drug was false and misleading in that said drug was recommended and suggested as appropriate for intravenous use in premature infants, when defendants knew of, and failed to reveal in the labeling, the following material facts:
(1) The drug had been reported to defendants to be associated with serious, life threatening injuries to premature infants, even when E-Ferol was used under the conditions of use prescribed, recommended, and suggested in the labeling of the drug product;
(2) No adequate investigation was conducted by the defendants to determine the validity of the associations of E-Ferol to said injuries that had been reported to them;
(3) No scientifically adequate or well controlled studies existed that established that E-Ferol could be safely and effectively used under the conditions of use prescribed, recommended, and suggested in the labeling of the drug product; and
(4) The safety and effectiveness of E-Ferol had not been established for use in premature infants.
B. Under [§ 352(f)(1) ] in that said drug, a prescription drug subject to the requirements of 21 C.F.R. 201.100(c) and (d), failed to bear adequate directions for its use, including dosage and administration, contraindications, side effects, hazards, and precautions, under which medical professionals licensed by law to administer the drug could use it safely for the purposes for which defendants intended it.
C. Under [§ 352(f)(2) ] in that the labeling of said drug failed to bear adequate warnings against use in premature infants, where its use might be dangerous to their health.
D. Under [§ 352(f)(2) ] in that the labeling of said drug failed to bear adequate warnings against unsafe dosage and methods and duration of administration and application, in such manner and form, as was necessary for the protection of users of the drug.
E. Under [§ 352(j) ] in that said drug, when used in the dosage, and with the frequency and duration prescribed, recommended, and suggested in the labeling was dangerous to the health of premature infants.
. As a direct result of the E-Ferol incident, the FDA amended the ISR policy in September 1984 by restricting it to only identical drug products. Notice of Revised Compliance Policy Guide 7132C.02, 49 Fed.Reg. 38190-01 (Sept. 27, 1984). For discussions of the above legislative and regulatory background, see generally id.; Weinberger v. Hynson, Westcott & Dunning, Inc., 412 U.S. 609, 612-15, 623-24, 93 S.Ct. 2469, 2474-76, 2480, 37 L.Ed.2d 207 (1973); and United States v. Articles of Drug—Lannett Co., Inc., 585 F.2d 575, 577-79 (3d Cir.1978).
. The court gave the following instructions:
Carter-Glogau Laboratories, Incorporated and Mr. Carter have introduced evidence that the regulation policies and practices of the United States Food and Drug Administration led these defendants to believe that E-Ferol Aqueous Solution did not require an approved new drug application, an NDA, before it could be legally introduced into interstate commerce.
There has been evidence that E-Ferol Aqueous Solution was a single entity vitamin drug which was similar or related to other vitamin products that themselves were not the subject of an approved new drug application, and that it [sic] had been marketed prior to 1938.
Carter-Glogau Laboratories, Incorporated and Mr. Carter, [sic] they were led to believe that E-Ferol Aqueous Solution could be used on equal terms with these other unapproved Vitamin E products.
Sept. 28 Tr. at 76-77.
Later, the court again instructed the jury that as to the new drug counts, "these defendants contend that the policies and practices of FDA led them to believe that E-Ferol Aqueous Solution could be legally introduced in interstate commerce without an approved NDA.” Id. at 85.
. We question whether a vagueness attack can be based on an agency policy that sets enforcement priorities but does not purport to define illegal conduct. Cf. United States v. Wiveli, 893 F.2d 156, 159 (8th Cir.1990) (vagueness doctrine presumes "a law that attempts to proscribe or prescribe conduct”). However, we need not resolve this issue given that there are other reasons why Carter’s argument must fail.
. We find no merit in Carter’s assertion that the district court’s refusal to give certain tendered instructions denied him a fair determination of this due process issue by the jury. The issue whether the vagueness doctrine precluded conviction of Carter on the new drug counts presented a question of law for the court to decide, not the jury. See United States v. Gleason, 726 F.2d 385, 388 (8th Cir.1984) (per cu-riam) (it is "not the jury’s duty to interpret the law”). The court did instruct the jury'that Carter and Carter-Glogau “contend that the policy and practices concerning the introduction in commerce of a single entity vitamin product, such as E-Ferol Aqueous Solution, was [sic] so vague and indefinite that these defendants did not understand that an approved NDA was required.” Sept. 28 Tr. at 85-86.
.We recognize that the Tenth Circuit’s subsequent decision in United States v. Cattle King Packing Co., Inc., 793 F.2d 232 (10th Cir.), cert. denied sub nom. Stanko v. United States, 479 U.S. 985, 107 S.Ct. 573, 93 L.Ed.2d 577 (1986), might be read as conflicting with Industrial Laboratories. In affirming a felony conviction under 21 U.S.C. § 676(a), a provision of the Federal Meat Inspection Act very similar to § 333(a)(2), the Cattle King court found no reversible error in the giving of a “responsible relationship” instruction nearly identical to the one upheld in Park, a misdemeanor FDCA case, because the jury was also instructed that it had to find the intent to defraud as required by § 676(a). Cattle King, 793 F.2d at 240-41. We are not convinced that Cattle King and Industrial Laboratories are necessarily irreconcilable. In any event, Cattle King is not directly applicable in this case because the district court did not give a Park instruction.
. The court instructed the jury that it could find E-Ferol was not generally recognized by qualified experts as safe and effective for use under the conditions prescribed, recommended, or suggested in its labeling, and therefore was a new drug under § 321(p)(l), if there were no adequate and well-controlled scientific studies establishing it as safe and effective for such use. Carter and Hiland do not challenge these instructions.
. We therefore reject Carter's contention that under the principles discussed in Holdridge v. United States, 282 F.2d 302 (8th Cir.1960), the indictment violated due process because it was entirely devoid of any scienter element related to the facts comprising the alleged violations of the FDCA.
. The court did not, however, instruct the jury that the defendants had to know their actions violated particular provisions of the FDCA. Hi-land concedes that such knowledge was not required, but Carter seems to suggest that it was. We reject this suggestion because there is nothing in § 333(a)(2), or elsewhere in the FDCA, that evidences a congressional intent to create such an exception to "the general rule that ignorance of the law is no excuse.” United States v. International Minerals & Chem. Corp., 402 U.S. 558, 563, 91 S.Ct. 1697, 1701, 29 L.Ed.2d 178 (1971).
. The court instructed the jury as follows:
*1130The element of knowledge may be satisfied by inferences drawn from proof that a defendant deliberately closed his eyes, which [sic] would have otherwise been obvious to him. A finding beyond a reasonable doubt with [sic] a conscious purpose to avoid in light [sic] would permit an inference of knowledge.
Stated another way, a defendant's knowledge of a fact may be inferred from willful blindness to the existence of the fact. It's entirely up to you as to whether you find any deliberate closing of the eyes and inferences to be drawn from any such evidence.
The showing of negligence of a statement [sic] is not sufficient to support a finding of willfulness or knowledge.
Sept. 28 Tr. at 83-84.
. Neither Carter nor Hiland alleges the evidence was insufficient to prove that they possessed actual knowledge E-Ferol was dangerous and falsely labeled.
. Madison testified that it was OJF’s policy to never call the FDA unless they knew what the answer would be.
.Immediately prior to this particular instruction, the court instructed the jury that:
As to the [new drug counts], defendant Hi-land contends that he had no knowledge that E-Ferol was an unapproved drug that required FDA approval, and had no intent to defraud or mislead.
As to the [misbranding counts], defendant Hiland contends that he had no knowledge that E-Ferol was a misbranded drug, no intent to defraud or mislead.
Sept. 30 Tr. at 81.
. Hiland’s reliance on Sandstrom v. Montana, 442 U.S. 510, 524, 99 S.Ct. 2450, 2459, 61 L.Ed.2d 39 (1979), is misplaced because it does not stand for the proposition that reversal is necessary whenever an instruction standing alone created any possibility that the jury misal-located the burden of proof.
. We need not address the government's argument that the testimony elicited from Hiland was also admissible under Fed.R.Evid. 404(b) as evidence of similar acts.
. The conversation was with Carter, but this was not disclosed to the jury in order to avoid prejudice to Carter and Carter-Glogau. In addition, no reference was made to the "seizure" of Dalalone P.R. to accommodate the concern of Hiland’s counsel that this term was prejudicial. The court also sustained counsel’s objections to the government’s efforts to cross-examine Hi-land regarding the level of his compensation at OJF.
. As this and other courts have recognized, even if the component parts of a drug are generally recognized as safe, the combination of those parts may not be safe. See International Nutrition, Inc. v. United States Dep’t of Health & Human Servs., 676 F.2d 338, 341 (8th Cir.1982); United States v. Articles of Drug ... Promise Toothpaste, 826 F.2d 564, 566 (7th Cir.1987); United States v. An Article of Drug ... "Entrol-C Medicated", 513 F.2d 1127, 1129-30 (9th Cir.1975); see also 21 C.F.R. § 310.3(h)(2) ("newness of a drug” may arise from "combination of two or more substances, none of which is a new drug”).
.Hiland asserts that the government’s inquiries were particularly prejudicial because the court failed to give a limiting instruction. Under Fed.R.Evid. 105, the burden was on Hiland to request such an instruction. See United States v. Gilmore, 730 F.2d 550, 555 (8th Cir. *11341984). He failed to do so and has not demonstrated plain error on appeal. See Fed.R. Crim.P. 52(b).
. The district court clearly did not abuse its discretion in admitting testimony concerning adverse reactions and deaths actually reported to the defendants prior to the recall. Such testimony was highly probative of their knowledge and intent. The suggestions in Carter’s brief that testimony of this kind was unduly prejudicial are devoid of merit.
. Although a drug's status as a "new drug” under § 321(p)(1) does not require evidence that it is actually unsafe, see, e.g., United States v. Undetermined Quantities of Various Articles of Drug ... Equidantin Nitrofurantoin Suspension, 675 F.2d 994, 1000 (8th Cir.1982), cert. denied sub nom. Performance Products, Inc. v. United States, 460 U.S. 1051, 103 S.Ct. 1497, 75 L.Ed.2d 929 (1983), a drug certainly cannot be considered generally recognized as safe if it has never been tested and is in fact dangerous when used as recommended in its labeling.
. For example, the court told the jury:
I don’t want the jury to make a decision based on the babies. What the decision must be based on is the charge before you. Now I've tried to emphasize that in the clearest and most emphatic language I could use, because I don’t want the defendants to be able to say they have not received a fair trial. I don’t want the defendants to be blamed for something they are not charged with.
Sept. 28 Tr. at 6-7.
. See supra note 13.
. The government contended that this language was appropriate because it tracks that used in 21 U.S.C. § 352(a).
. With respect to both Carter and Carter-Glo-gau, the jury specified options two, four and five, which stated that: (2) "the labeling for E-Ferol failed to bear adequate directions for use”; (4) "the labeling for E-Ferol failed to bear adequate warnings against unsafe dosage and methods and duration of administration and application in such a manner and form as was necessary for the protection of users of the drug”; and (5) “E-Ferol when used in the dosage and with the frequency and duration prescribed, recommended and suggested in the labeling was dangerous to health.”
.The jury specified option three, which stated that "the labeling for E-Ferol failed to bear adequate warnings against its use in premature infants where its use might be dangerous to their health.”
. We need not decide this issue given our conclusions regarding the second misbranding verdict.
. Rule 31(d) reads as follows:
Poll of Jury. When a verdict is returned and before it is recorded the jury shall be polled at the request of any party or upon the court's own motion. If- upon the poll there is not unanimous concurrence, the jury may be directed to retire for further deliberations or may be discharged.
. So named after Allen v. United States, 164 U.S. 492, 17 S.Ct. 154, 41 L.Ed. 528 (1896).
. We reject Hiland’s claim to the contrary. Although his counsel objected to the mis-branding instructions preceding the first verdict *1138on the ground that they would not ensure unanimity, the record shows that counsel did not object to resubmission of the misbranding counts or the language of the supplemental charge.
. Hiland does not argue that the court’s initial questioning of the jury as a whole constituted a ■■poll" and that the court abused its discretion in employing this method of "polling” the jury. Even if we were to read Hiland’s briefs as making such an argument, we would find no reversible error. See United States v. Dotson, 817 F.2d 1127, 1130 n. 1 (5th Cir.1987) (trial court’s demand for a nodding of heads not reversible error because defendants did not object to this informal method of polling jury), modified in part on other grounds, 821 F.2d 1034 (5th Cir.1987) (per curiam); see also Carter, 112 F.2d at 67-68 (no reversible error where trial court polled jury by asking for show of hands, even though court denied defendant’s subsequent request for an individual poll).
. It appears from the record that the court misunderstood Hiland’s requests for a specific unanimity instruction as relating only to the misbranding counts.
. Courts have also held that a specific unanimity instruction is required where a conviction may occur as a result of different jurors concluding that the defendant committed different acts. See United States v. Busacca, 863 F.2d 433, 437 (6th Cir.1988) (per curiam), cert. denied, -U.S.-, 109 S.Ct. 1640, 104 L.Ed.2d 156 (1989); United States v. Beros, 833 F.2d 455, 461 (3d Cir.1987). This exception to the general rule would only apply here, if at all, to the first misbranding verdict against Hiland.
. The other factors that courts have identified include a variance between the indictment and the proof at trial, and a tangible indication of jury confusion, such as a request for clarification. See United States v. Anguiano, 873 F.2d 1314, 1319-20 (9th Cir.), cert. denied, — U.S. -, 110 S.Ct. 416, 107 L.Ed.2d 381 (1989); United States v. Ryan, 828 F.2d 1010, 1020 (3d Cir.1987).
. The court began this inquiry with the conspiracy verdict: “All of you have found all three of the [defendants guilty as to count one, which is conspiracy. That’s the first group. Is there anyone here who did not find any of these defendants guilty of conspiracy?” Upon receiving no response, the court stated: "I take it by your silence, that all of you are confirming the fact that you have found each of the three defendants guilty, that is the corporation and Ronald Carter and Larry Hiland, guilty of conspiracy.” Sept. 30 Tr. at 6. The court followed a similar pattern of inquiry with respect to the remaining counts.
. We note that the Third Circuit reached a similar conclusion under analogous circumstances. See United States v. Aimone, 715 F.2d 822, 833 (3d Cir.1983), aff'g United States v. Musto, 540 F.Supp. 318, 340 (D.N.J.1982), cert. denied, 468 U.S. 1217, 104 S.Ct. 3586, 82 L.Ed.2d 883 (1984).
. Section 1918(b) provides that:
Whenever any conviction for any offense not capital is obtained in a district court, the court may order that the defendant pay the costs of prosecution.
.Carter-Glogau and the government disagree whether payment of the costs of prosecution may be imposed as a condition of probation. In United States v. Dougherty, 810 F.2d 763, 773-74 (8th Cir.1987), we concluded that the district court did not abuse its discretion in taxing costs to the defendant, an individual, as a special condition of probation. But see United States v. Interstate Cigar Co., Inc., 801 F.2d 555, 556-57 (1st Cir.1986) (holding that district court erred in ordering corporation convicted of mail fraud to serve probation in addition to paying maximum statutory fine). In any event, despite the implication of the parties’ arguments, we can find nothing in the record indicating that the district court in fact placed Carter-Glogau on probation.
. The record seems to indicate that the district court chose the figure $100,000 in order to have Carter-Glogau bear a portion of the costs of the government’s investigation in this case.