Last Friday, we provided our readers with a “breaking news” (that is to say, quick and dirty) overview of the FDA’s just-released proposal to require manufacturers of generic drugs to use the “changes being effected” (“CBE”) process (also referred by the FDA as “CBE-0”) to update their labels in the same fashion as innovator/reference listed drug manufacturers.  That proposal appeared in the Federal Register on November 13, 2013 (pages 67985–67999), and here are links to it online and exactly as it appears in the hard copy of the Federal Register.  The 60-day comment period (unless extended) thus runs from 11/13.  Let’s see, 30 days hath November….  We think that the comment period ends on January 13 (the 12th is a Sunday), 2014.

Today we’re taking a closer look as a couple of places where we think the FDA’s proposal comes up short because it fails to address significant problems.  Those are:  (1) the Agency’s rather terse cost benefit analysis, and (2) the FDA’s failure to evaluate the relevant portion of Hatch/Waxman, 21 U.S.C. §355(j) – or indeed any part of Hatch/Waxman at all − in its analysis of its legal authority to make these changes without any Congressional amendments to the statute
itself.

As far as the merits of the proposal, we’ve already discussed what we think should be done to fix the inconsistent and backwards application of preemption to the FDCA under Riegel v. Medtronic, Inc., 552 U.S. 312 (2008), Wyeth v. Levine, 555 U.S. 555 (2009), and PLIVA v. Mensing, 131 S. Ct. 2567 (2011), so we’re not going over that again.  The FDA didn’t go that route, nor did we expect it to.  Instead, it lunged for whatever it could devise to reduce preemption the most.  We’ll leave to others to comment on why that was the case, but we must admit that we’re disappointed.  For decades the FDA has taken the position that it does not tailor its regulations to affect civil liability:

  • “Tort liability cannot be a major consideration for FDA which must be guided by the basic principles and requirements of the act in its regulatory activities.”  63 Fed. Reg. 66378, 66383 (FDA Dec. 1, 1998).
  • “It is not the intent of FDA to influence the civil tort liability of the manufacturer or the physician.”  44 Fed. Reg. 37434, 37437 (FDA June 26, 1979).
  • “[W]hether particular labeling may alter a manufacturer’s liability in a given instance cannot be considered as a dispositive factor by the Commissioner in reaching a decision.”  42 Fed. Reg. 37636, 37637 (FDA July 22, 1977).

Now, the FDA has done a screeching four-wheel drift 180 and proposes to modify its regulations explicitly for the purpose of changing the scope of civil tort liability – in order to eliminate preemption of generic product liability warning (but not design) claims recognized by the Supreme Court under Hatch/Waxman in Mensing and recently reaffirmed in Mutual Pharmaceutical Co. v. Bartlett, 133 S. Ct. 2466 (2013).  See 78 Fed. Reg. 67985, 67988-89 (FDA Nov. 13, 2013) (discussing “recent court decisions”).

We turn to the FDA’s cost benefit “analysis” (use of quotation marks intentional), which consists of only one paragraph, so we’ll quote it in full:

The economic benefits to the public health from adoption of the proposed rule are not quantified.  By allowing all application holders to update labeling based on newly acquired information that meets the criteria for a CBE-0 supplement, communication of important drug safety information to prescribing health care providers and the public could be improved.  The primary estimate of the costs of the proposed rule includes costs to ANDA and NDA holders for submitting and reviewing CBE-0 supplements.  The Agency estimates the net annual social costs to be between $4,237 and $25,852.
The present discounted value over 20 years would be in the range of $63,040 to $384,616 at a 3 percent discount rate, and in the range of $44,890 to $273,879 at a 7 percent discount rate.

78 Fed. Reg. 67985, 67986 (FDA Nov. 13, 2013) (emphasis added).

The mind boggles.  The FDA has had this proposal under consideration for close to two years and it has been unable to identify a single quantifiable benefit from imposing CBE obligations on generic drugs.  The most it can say is that this change “could improve” communication of drug safety information.  But on the other hand, it might not improve communication at all:

FDA acknowledges that there may be concerns about temporary differences in safety-related labeling for drugs that FDA has determined to be therapeutically equivalent, especially if multiple ANDA holders submit CBE-0 supplements with labeling changes that differ from each other and from the RLD.  FDA also recognizes that health care practitioners are unlikely to review product labeling for each of the generic drugs that may be substituted for the prescribed product when making treatment decisions with their patients based on the balance of potential benefits and risks of the drug product for that patient.

Id. at 67989 (emphasis added).

In other words, multiple, differing versions of labeling for what’s supposed to be the same drug can’t help but be confusing, and doctors might not read all those different labels anyway.  Thus, the FDA proposes a website where all of these differing labels will be collected.  If confusion isn’t a problem, then what’s the need for the website?  This appears to be another flip flop, as the government told the Supreme Court in Mensing that differing warnings between innovator and generic drugs “could inaccurately imply therapeutic differences between the generic drug and its [listed equivalent] that do not exist, and therefore be misleading.”  Brief for the United States as Amicus Curiae Supporting Respondents, 2011 WL 741927, at *19, see Mensing, 131 S. Ct. at 2576 (accepting government’s argument on this proposition).  We don’t see how the same sort of differences among differing generics, or between the innovator and generic, contemplated in the current proposal could be any less “misleading.”

The Agency’s failure to address costs in any meaningful fashion is even worse.  First of all, if there’s no preemption, then there’s likely to be an explosion of litigation.  The FDA has previously determined that litigation is inefficient and much more expensive than regulation:

[T]he costs of private monitoring and enforcement of safety using the tort system in an unregulated market are probably substantially greater than the social costs of regulatory enforcement.

65 Fed. Reg. 43269, 43279 (FDA July 13, 2000).  In prior regulatory activity, the Agency has considered reduction of litigation to be a net benefit.  61 Fed. Reg. 8432, 8437 (FDA Mar. 4, 1996) (“FDA agrees that clearer labeling may in some instances help shield manufacturers from product liability.  However, regardless of any effect on product liability, improved labeling, which may help reduce the incidence of injury and death is important”); 59 Fed. Reg. 3944, 3947 (FDA Jan. 27, 1994) (“increase in product liability and medical malpractice litigation” considered a cost that justified preemption); 42 Fed. Reg. 37636, 37637 (FDA July 22, 1977) (“reduced potential manufacturer liability, owing to improved patient compliance and a corresponding decrease in drug-induced injury” considered a benefit of patient package labeling).

Putting aside the current proposal’s various inconsistencies with what the FDA has said in the past (e.g. “changes in labeling that involve warnings or other similar risk information . . . must be included in the labeling of generic competitors,” 59 Fed. Reg. 50338, 50357 (FDA Oct. 3, 1994)), the Agency has simply ignored the cost of increased product liability litigation, which it nowhere alludes to, let alone addresses.  That’s simply inconceivable – particularly since a major purpose of Hatch/Waxman was to ensure that generic drugs were both cheap and plentiful.  See Mensing, 131 S. Ct. at 2582 (intent was “bringing more [generic] drugs more quickly and cheaply to the public”).   How much more litigation?  We can’t begin to guess, but our own generic PMA preemption scorecard currently has 109 decisions on it, and that’s only in the 2+ years since Mensing – and just on that one (albeit important) issue.  We think it’s safe to say that the litigation costs to industry of the FDA’s proposal would be three to four orders of magnitude greater than the trivial costs that the FDA’s recent proposal actually mentions.

That’s not all.  The FDA states, 78 Fed. Reg. at 67989, that generic manufacturers have always had the obligation to report adverse events.  However, the new proposal would require much more than just reporting – by extending the CBE process to generic manufacturers, it would require for the first time that they conduct actual medical and statistical analysis on the material they report to determine if anything in those reports meets the current CBE standard for seeking a label change:  “to add or strengthen a contraindication, warning, precaution, or adverse reaction for which there is reasonable evidence of a causal association.”  Id. at 67996.  This analysis can get pretty complicated – don’t take out word for it; check out the CBE flow chart that the FDA includes.  Id. at 67992.

Opinions can differ on whether or not it’s a good idea to require such analysis from generic manufacturers.  On the one hand, it’s what innovator manufacturers have always been required to do.  On the other hand, for innovator drugs, there is only one manufacturer, so that one entity receives all the adverse reports.  With generics there can be many manufacturers, so the already limited value of adverse reports (recognized by the FDA to be “dubious,” “inefficient,” of “low quality,” “biased,” and “not [to] be used to calculate incidences or estimates of drug risk,” see here), would be further diluted by any generic manufacturer seeing only a sliver of the reports.  How is a generic manufacturer with only, say, a 5% market share to evaluate such a limited universe of reports against this “reasonable evidence of a causal association” standard?  We don’t know, and apparently neither does the FDA.  The difference is, we’re not supposed to.

What we can anticipate is that most, if not all, generic manufacturers would have to hire new professional staff (statisticians, medical analysts, etc.) in order to have any hope of complying with this regulation.  Right now, bioequivalence is the sine qua non of the generic drug industry.  They primarily concentrate on producing properly manufactured, bioequivalent pills with the lowest possible overhead, and in particular without the large and expensive science departments maintained by innovator drug manufacturers.  With the new obligations imposed by the FDA’s proposal, that would have to change.  The FDA seems totally oblivious to the institutional cost that complying with its new scientific analysis requirements would impose on the generic industry.  Whether it’s good or bad to do this, that cost certainly would exist.  We don’t have any data, but we wouldn’t be surprised if some generic manufacturers decide to abandon some drugs (particularly small niche markets) altogether as unprofitable in light of these increased costs.  That outcome – which the FDA’s proposal doesn’t even discuss – would be difficult to square with the intent of Hatch/Waxman to ensure widespread availability of cheaper generic products.

Moreover, pervasive uncertainty as to when the CBE standard for proposing label changes is triggered will only add to the litigation costs previously mentioned, in ways that go beyond innovator drug situations.  We’ve seen something similar in cases where there are multiple drugs in the same class (such as Cox-II inhibitors, statins, and SSRIs).  When one manufacturer seeks a label change, plaintiffs use that request as a sword to argue that every other manufacturer of the same class of drug that didn’t make that request at that time (or earlier) was somehow negligent.  With, say, ten manufacturers of an identical (as opposed to merely similar) product, the same problem could only get worse.  Based on whoever goes first, as somebody has to, plaintiffs would surely argue that “the devil take the hindmost” – that is, that everybody else was negligent from (at least) that moment forward in not making a similar proposal.  Generic manufacturers with small market shares, and thus access to less data, would be particularly disadvantaged, another reason they might decide to stop manufacturing such products.

For the above reasons (and there may well be others), we think that the FDA’s proposal is extremely vulnerable for failure to conduct anything that approaches a plausible cost-benefit analysis.

Our second area of concern (that we are addressing in this post, anyway) is the FDA’s superficial analysis of its legal authority to make these changes, given the current language of the statute itself.  The FDA’s proposal correctly states:  “Under section 505(j) of the FD Act, FDA will approve an ANDA only if the drug is, with limited exceptions, the same as a drug previously approved . . . with respect to . . . labeling. . . .”  78 Fed. Reg. at 67995.  Okay, then how does the FDA’s proposal, which admits to at least “temporary” dissimilarity of labeling, square with the statute itself?

As with cost/benefit, the FDA’s proposal essentially says nothing.  Specifically it states:

FDA’s authority to extend the CBE-0 supplement process for safety-related labeling changes to ANDA holders arises from the same authority under which our regulations relating to NDA holders and BLA holders were issued.  Nothing in the Hatch-Waxman Amendments or subsequent amendments to the FD&C Act limits the Agency’s authority to revise the CBE-0 supplement regulations to apply to ANDA holders to help ensure that generic drugs remain safe and effective under the conditions of use prescribed, recommended, or suggested in the labeling throughout the life cycle of the generic drug product.

78 Fed. Reg. at 67995 (emphasis added).

The trouble is, this global “nothing in Hatch/Waxman” statement is simply not true.  The “sameness” requirement that underlies preemption is in the statute, and is unique to generic drugs.  Thus the “same authority” argument that the FDA makes doesn’t fly – because the statute, which the FDA has no authority to change without congressional action, isn’t the same for “NDA holders and BLA holders” as it is for generics.

We’ll now demonstrate the problem.  Unlike the FDA, we’ll set out precisely what the organic statute says about labeling being the “same” or alternatively not “differing,” from that of the reference innovator drug.

First of all, with respect to the initial approval, §355(j) provides that an ANDA [that is, a generic application for approval] must contain:

information to show that the labeling proposed for the new drug is the same as the labeling approved for the listed drug referred to in clause (i) except for changes required because of differences approved under a petition filed under subparagraph (C) or because the new drug and the listed drug are produced or distributed by different manufacturers.

21 U.S.C. §355(j)(2)(v) (emphasis added).

So we have two exceptions – showing that Congress knew how to write exceptions and didn’t intend sub rosa to include any others that it didn’t specify.  The second exception is facially irrelevant since that is limited to different names, addresses, and the like for identifying other manufacturers of the same drug.  So what does the “subparagraph C” statutory exception to sameness involve?

Prior FDA approval, that’s what:

(C) If a person wants to submit an abbreviated application for a new drug which has a different active ingredient or whose route of administration, dosage form, or strength differ from that of a listed drug, such person shall submit a petition to the Secretary seeking permission to file such an application. . . .  The Secretary shall approve such a petition unless the Secretary finds −

(i) that investigations must be conducted to show the safety and effectiveness of the drug or of any of its active ingredients, the route of administration, the dosage form, or strength which differ from the listed drug; or

(ii) that any drug with a different active ingredient may not be adequately evaluated for approval as safe and effective on the basis of the information required to be submitted in an abbreviated application.

21 U.S.C. §355(j)(2)(C) (emphasis added).

Hatch/Waxman thus does not confer any right upon an ANDA applicant to submit any generic drug with labels that differ/are-not-the-same in ways that affect the safety and effectiveness of that drug.

Section 355(j) also lists the conditions for FDA approval of a generic application.  Specifically,

Subject to paragraph (5) [irrelevant, having to do with patent litigation], the Secretary shall approve an application for a drug unless the Secretary finds. . . .

(G) information submitted in the application is insufficient to show that the labeling proposed for the drug is the same as the labeling approved for the listed drug referred to in the application except for changes required because of differences approved under a petition filed under paragraph (2)(C) or because the drug and the listed drug are produced or distributed by different manufacturers. . . .

21 U.S.C. §355(j)(4)(G) (emphasis added).  Thus the statutory power of the FDA to approve generic drugs parallels the statutory right of applicants to submit ANDAs – the labeling must be the “same” in all relevant ways.  Thus the FDA has no power to allow any generic labels that are not the same as the referenced innovator drug labels without prior approval, as it purports to do in its current proposal.

A third part of §355(j) deals with the FDA’s authority to allow certain “labeling revisions” that result in labels that “differ[] from the listed drug.”  That seems sort of relevant.  Such revisions are only permitted if:  (1) pre-approved by the FDA, and (2) they do not change “warnings”:

(10)(A) If the proposed labeling of a drug that is the subject of an application under this subsection differs from the listed drug due to a labeling revision described under clause (i), the drug that is the subject of such application shall, notwithstanding any other provision of this chapter, be eligible for approval and shall not be considered misbranded under section 352 of this title if −

(i) the application is otherwise eligible for approval under this subsection but for expiration of patent, an exclusivity period, or of a delay in approval described in paragraph (5)(B)(iii), and a revision to the labeling of the listed drug has been approved by the Secretary within 60 days of such expiration;

(ii) the labeling revision described under clause (i) does not include a change to the “Warnings” section of the labeling. . . .

21 U.S.C. §355(j)(10)(A) (emphasis added).

Thus, specifically with respect to revisions to labeling, Congress included an exception to sameness.  But that exception:  (1) explicitly requires agency prior approval, and (2) prohibits revisions that seek to change “warnings” – which is precisely what the CBE provisions involved in the current proposal purport to do.  Several maxims of statutory construction would seem to preclude the FDA from doing what it seeks to do here.

To us it appears that, in three different places in Hatch/Waxman, Congress expressly provided that labeling must be the same, and further that any ANDA revisions that would affect safety or effectiveness must be pre-approved by the Agency.  Finally, in §355(j)(10(A)(ii), Congress essentially stated “this goes double for anything that could change warnings.”

In light of what Hatch/Waxman explicitly provides, we are unimpressed with the proposal’s vague, blanket statement that “Nothing in the Hatch-Waxman Amendments or subsequent amendments to the FD&C Act limits the Agency’s authority to revise the CBE-0 supplement regulations to apply to ANDA holders.”  That’s just not so, and no amount of repetition can change the express language of the statute.  We’d also point out that the FDA has been down this road before – with this same statute – and drew back a nub when it offered regulations that were incompatible with the statutory text.  See Mova Pharmaceutical Corp. v. Shalala, 140 F.3d 1060 (D.C. Cir. 1998).

In Mova, the FDA argued that it could fix perceived flaws in Hatch/Waxman because the statutory language led to “absurd” results.  Id. at 1067.  The FDA added a requirement (relating to patent litigation) for which Hatch/Waxman had “no provision.”  Id.  The court said, no you can’t do that – that the results, while weird and probably unanticipated, weren’t sufficiently absurd to allow the FDA to make up something that wasn’t in Hatch/Waxman at all:

In deciding whether a result is absurd, we consider not only whether that result is contrary to common sense, but also whether it is inconsistent with the clear intentions of the statute’s drafters − that is, whether the result is absurd when considered in the particular statutory context. . . .  But the agency does not thereby obtain a license to rewrite the statute.  When the agency concludes that a literal reading of a statute would thwart the purposes of Congress, it may deviate no further from the statute than is needed to protect congressional intent.

Id. at 1068 (citations omitted).  The record was not “sufficiently clear” that Congress intended what the FDA sought to add.  Id. at 1070.  Specifically, while the situation that the FDA sought to address could produce an “anomaly” or a “strange result,” the “limited” legislative history of Hatch/Waxman did not “show that a literal reading of the statute leads to results manifestly inconsistent with the intent of Congress.”  Id. at 1072.  Likewise, given the overall intent of Hatch/Waxman to ensure a cheap and plentiful supply of generic drugs, and the probable adverse effect of this proposal’s increased litigation and administrative costs on those goals, we don’t think that the “manifestly inconsistent” standard of Mova could be met here.  Indeed, the current proposal, as drafted, doesn’t even try.

The FDA in Mova, like the proposal here, sought to paint with a broad brush, in disregard of the express language of the statute.  Mova held that the FDA was bound by the language Congress had included in Hatch/Waxman and could not add provisions that the statute itself did not support:

We conclude that the FDA’s [regulatory] requirement is inconsistent with the unambiguously expressed intent of Congress.  The rule is gravely inconsistent with the text and structure of the statute.  Nor can the FDA show that the [regulatory] requirement is needed to avoid a result demonstrably at odds with the intentions of [Hatch/Waxman’s] drafters.  The FDA could have adopted a more narrow solution to the problem. . . .  It instead adopted the broad [] rule, which it cannot show is needed to implement congressional intent.  In effect, the FDA has embarked upon an adventurous transplant operation in response to blemishes in the statute that could have been alleviated with more modest corrective surgery.

Id. at 1069 (citations and quotation marks omitted).

Certainly, the FDA could have addressed the issue of updating generic warnings more narrowly.  As we’ve shown, Hatch/Waxman allows modifications to labeling, only subject to FDA pre-approval.  If all that motivated the FDA was to speed up safety-related label changes in light of emerging, new information, the Agency could have established short and strictly enforced time limits for its prior approval within the statute’s “sameness” requirements.  However, the current FDA proposal isn’t primarily concerned with safety, but with preemption.  The FDA had to resort to the broad brush because its goal was to affect civil litigation (in direct contravention of prior statements) rather than to increase the safety of generic products.  We doubt that’s allowed under Mova.

In this respect we think that Mensing itself is illustrative.  The Court found that differences in preemptive effect between generic and branded products “make[] little sense,” 131 S. Ct. 2581, something with we’ve already agreed.  However, just because enforcing a statute as written yields strange results does not give either courts or administrative agencies carte blanche to ignore the statutory language that Congress drafted in Hatch/Waxman (or anywhere else):

We recognize that from the perspective of [plaintiffs], finding pre-emption here but not in [innovator drugs] makes little sense. . . .  But it is not this Court’s task to decide whether the statutory scheme established by Congress is unusual or even bizarre.  It is beyond dispute that the federal statutes and regulations that apply to brand-name drug manufacturers are meaningfully different than those that apply to generic drug manufacturers . . . .  But different federal statutes and regulations may, as here, lead to different pre-emption results.  We will not distort the Supremacy Clause in order to create similar pre-emption across a dissimilar statutory scheme.  As always, Congress and the FDA retain the authority to change the law and regulations if they so desire.

Id. at 2581-82 (citations and quotation marks omitted) (emphasis added).

Changing the FDCA is Congress’ prerogative.  The FDA can change its regulations, but as demonstrated by Mova, only insofar as changes to such regulations are consistent with what the statute provides.  As desirable as it certainly is for there to be “parity” between innovator and generic drugs − and we would add between safety-based FDA approval of drugs and medical devices – the statute here simply doesn’t authorize the avenue that the FDA’s proposal is seeking to create, that is, a safety-related exception to “sameness” that also does away with any requirement of Agency pre-approval.

Only Congress can do that, and it hasn’t.

  • Josh

    I believe your summary of the FDA's authority under §355 contains two omissions that cause me to question your ultimate conclusion about the FDA's lack of authority to require warning changes.

    First the subsections of §355(j)(2) refer to requirements for (and differences in) abbreviated new drug applications. The conclusion of that section of the statute states that:

    "The Secretary may not require that an abbreviated application contain information in addition to that required by clauses (i) through (viii)."

    So, a generic manufacturer must start with the same labeling as the reference drug (except for the provisions under sub-paragraph (C), which you discussed, relating to differences in formulation).

    Likewise, the provisions under §355(j)(4) refer to the fact that applications must be approved if they meet essentially the requirements in (j)(2). Again, this addresses the starting point for the warnings; this fact was brushed under in your analysis.

    The key question is whether the statute prohibits the FDA from authorizing or requiring amendments to the warnings. The act is silent on this point, but there are some provisions that suggest Congress intended the FDA to have this power.

    First, I read paragraph (j)(10) slightly differently than you appear to, particularly as to its purpose. That paragraph addresses the situation where the listed drug makes a change to its label after a generic application has been submitted. The differences created by such a change are OK as long as listed drug label "does not include a change to the 'Warnings' section." What Congress was ensuring here was a minimum requirement that the new generic label is not missing warnings that were added to the reference drug after the application was submitted. Congress wanted to permit other (presumably minor) differences, as long as even those differences are cured within 60 days. (see (j)(10(A)(iii). In other words, a generic may not fail to include a (new) known warning. It is a bit of a leap to read into this provision a requirement that if a manufacturer (even of a generic) obtains new information about a risk that is not accurately described in the label, that Congress has prohibited this information from being provided to doctors and their patients.

    This brings us to §355(j)(4)(K), which was not mentioned in your post. It allows the FDA to not approve a generic application if "the application contains an untrue statement of material fact." Arguably, a misrepresentation or omission about the known risks of a drug would render the warnings as a whole an untrue statement that is material to prescribing doctors and their patients. So, while this paragraph is also silent on the FDA's authority to require a generic manufacturer to change its warnings, it does provide the FDA with authority to reject applications that do not contain appropriate warnings, even if the submitted application has a label that is identical to the reference drug.

    These two provisions indicate a focus on ensuring that the public has information about the risks of prescription medication, even though they all regulate the FDA's authority as to applications.

    Turning to postmarket label changes, under subsection (o)(4), the FDA can require a change in a generic label in light of new safety information. This provision requires notification of a required change to the reference drug and only requires notification to the generic if there is no reference drug on the market. It appears Congress assumed that warnings updated across equivalent drugs and therefore remain the same, but that doesn't mean that the FDA cannot require any manufacturer to initiate a label change. Presumably any such change would then be replicated to ensure that all equivalent drugs have all known information about risks. Such a requirement would appear to be consistent with the purpose of the statute.

  • First, if the FDA thought the provisions you mention could be construed in that fashion, I'm sure it would have trumpeted them to the skies. The Agency didn't even mention them.

    Second, there's something of an inconsistency in your dismissing of explicit sameness requirements as just referring to applications, and then relying on (4)(k), also an application provision that doesn't even mention sameness. While your flip flops aren't as bad as the FDA's, it is noticeable.

    Third, (4)(k) doesn't even deal with sameness, but is a broad "not misleading" section. Specific controls general.

    I'm not interpreting anything as prohibiting information from being provided to physicians. The statute simply says the FDA has to approve any difference first. And the statute has done that in several places.

    (o)(4) gives the FDA the power to mandate a change. That's not in the same ballpark as requiring generic manufacturers to propose changes when the FDA hasn't done anything at all. The FDA can always do that, but (o)(4) doesn't get around the impossibility rationale of Levine/Mensing, so the FDA can't hang it's hat there if it hopes to restrict preemption.

    You tried hard, but there's still nothing in the statute allowing for non-same warnings without prior FDA approval (although in deference to (o)(4), I should also add order).

    – Bexis