IMS Health Inc. v. Sorrell

DEBRA ANN LIVINGSTON, Circuit Judge,

dissenting:

Misconstruing Vermont’s prescription confidentiality law, Vt. Stat. Ann. tit. 18 § 4631 (2007) (hereinafter “section 17”),1 as a direct restriction on pharmaceutical marketing, which is indisputably a form of “commercial speech” for purposes of the First Amendment, the majority extends First Amendment protection to data miners and pharmaceutical companies principally challenging a restriction on access to otherwise private information. In so doing, the majority not only reaches the wrong result in this case, but creates Circuit precedent likely to have pernicious broader effects in a complex and evolving area of First Amendment law. Because I would find that section 17 permissibly restricts access to information that Vermont requires pharmacies to collect and that the statute has very limited, if any, effects on First Amendment activity, I respectfully dissent.

I.

I begin with common ground: there is no dispute that prescriber-identifiable data — i.e., data which documents the prescribing habits of a particular doctor (“PI data”) — is exceptionally valuable to pharmaceutical companies, who make use of it to market their highly profitable brand name drugs through a process known as “detailing.”2 There also is no dispute that the marketing messages “detailers” deliver in meetings with doctors constitute protected First Amendment activity. Finally, there is no dispute that section 17 does not directly regulate those messages or the marketing practices of detailers. Maj. Op. at 277. Instead, Vermont’s law regulates the dissemination of confidential information — specifically, PI data — and the process by which it is collected and sold. Because section 17 targets that process rather than detailing itself, “understanding the sequence of events” section 17 regulates — that is, the process by which PI data travels from the prescription pad to the hands of a pharmaceutical detailer — -“is crucial to understanding the statute’s legal status.” IMS Health Inc. v. Mills, 616 F.3d 7, 40 (1st Cir.2010) (Lipez, J., concurring in part and dissenting in part).

Pursuant to Vermont law, every time a pharmacy fills a prescription within the state, it is required to collect certain information about the doctor, the patient, and *283the medication being prescribed. See, e.g., Vt. Bd. of Pharmacy Admin. Rules §§ 9.1, 9.24, 9.26 (eff.Oct.2009).3 Because that information is so valuable to any number of third parties, including the plaintiffs-appellants in this case, pharmacies, for some time, have made a practice of selling it — often without the knowledge or permission of the doctor, let alone the patient — to various third parties, including data mining vendors such as plaintiffs-appellants IMS Health Inc., Verispan, LLC, and South Healthcare Analytics (collectively, the “data mining appellants”).4 These vendors aggregate and compile the data they acquire from pharmacies and then license it to pharmaceutical companies, represented here by plaintiff-appellant Pharmaceutical Research and Manufacturers of America (“PhRMA”), who use the information to guide some of their marketing and, in particular, their “detailing” efforts. Specifically, pharmaceutical companies use PI data to identify particular doctors for “detailing,” to monitor the success of their detailing efforts, and to compensate individual detailers based on the prescriptions written by the doctors they meet with. Pharmaceutical detailers do not, however, directly reference PI data in their meetings with doctors, and in fact, are prohibited from doing so by the terms of their employers’ licensing agreements with the data mining appellants.

Accordingly, before a detailer ever sets foot in a doctor’s office — that is, before the commercial speech the majority focuses on ever occurs — at least three events take place: first, a pharmacy gathers information from patients seeking to fill prescriptions; second, it collects and sells that data to third parties, principally “data vendors” or “data miners” such as appellants here; and third, these data miners repackage that data and license it to pharmaceutical companies. See generally IMS Health Inc. v. Ayotte, 550 F.3d 42, 48-49 (1st Cir.2008). Only after these three transactions occur does PI data land in the hands of detailers who then use it to facilitate their detailing efforts.

Troubled by this sequence of events whereby otherwise confidential information ends up in the hands of pharmaceutical detailers and in response to concerns about (1) medical privacy, (2) threats to patient health, and (8) rising health care costs attributable to the widespread use of new brand name prescription drugs (which the record indicates are those most likely to be the subject of extensive detailing efforts) Vermont enacted its prescription confidentiality law. In relevant part, the law prohibits any “health insurer, [ ] self-insured employer, [] electronic transmission intermediary, [] pharmacy, or other similar entity” from “selling], licensing], [] exchanging] for value” or otherwise “permitting] the use” of “preseriber-identifiable information for marketing or promoting a prescription drug” absent the *284prescriber’s consent. The law further prohibits “pharmaceutical manufacturers and [] marketers” from “us[ing] prescriberidentifiable information for marketing or promoting a prescription drug” unless the prescriber consents in the manner provided by statute. Vt. Stat. Ann. tit. 18, § 4631(d).

Focusing heavily on that last restriction, the majority begins its analysis at the end of the “sequence of events” — i.e., at the point at which PI data is already in the hands of pharmaceutical companies — and concludes that the law impermissibly “restricts the speech of both the pharmaceutical manufacturers ... who are prohibited from using Vermont PI data for marketing purposes, and the data mining appellants, who are prohibited from selling or transferring Vermont PI data if the data is to be used for marketing purposes.” Maj. Op. at 273. The law, however, starts at the beginning, and seeks to cut off the flow of PI data at its source: section 17 prohibits any pharmacy from “selling] ... prescriber-identifiable information ... [or] permitting its use ... for marketing or promoting a prescription drug.” Vt. Stat. Ann. tit. 18, § 4631(d) (emphasis added).5 Because the restrictions imposed by section 17 begin there, and because that first restriction prevents PI data from ever reaching the hands of plaintiffs-appellants, the principal question to be resolved — and one the majority wholly overlooks — is whether the restriction on pharmacies implicates the First Amendment interests of the data miners and pharmaceutical companies before the Court.6

In considering that restriction, I begin with the undisputed fact that Vermont pharmacies have access to and collect prescription information only under the direction and authority of state law. As noted, Vermont requires pharmacies to collect information such as the name of the prescribing doctor, the name and age of the patient, and the drug and dose prescribed. Having mandated the collection of that otherwise highly confidential information, the state unquestionably has an interest in controlling its further dissemination. It is that interest that section 17 effectuates — with respect to appellants, Vermont’s law operates principally to prevent them from obtaining otherwise private PI data, and as such, does no more than restrict their unfettered access to information. This the First Amendment permits. See Zemel v. Rusk, 381 U.S. 1, 17, 85 S.Ct. 1271, 14 L.Ed.2d 179 (1965) (First Amendment “does not carry with it the unrestrained right to gather information”).

In finding that section 17 operates principally as a permissible regulation on access to information, I am guided by the Supreme Court’s decision in Los Angeles Police Department v. United Reporting Publishing Corporation, 528 U.S. 32, 120 S.Ct. 483, 145 L.Ed.2d 451 (1999). There, a private publishing company challenged a California state law that restricted access *285to information collected by local police departments respecting those arrested within the state. The Court found that, at least with respect to that plaintiff, the law had no First Amendment implications because it did no more than “regulate[ ] access to information in the hands of the police department.” Id. at 40, 120 S.Ct. 483. As the Court further noted, “California could decide not to give out arrestee information at all without violating the First Amendment.” Id.

The majority attempts to distinguish United Reporting on the ground that while the California law amounted to “a government denial of access to information in its possession,” id. (emphasis added), here “the information is not in the government’s possession” but instead “in the hands of pharmacies.” Maj. Op. at 273. As a preliminary matter, the argument completely disregards the fact that the information is only “in the hands of’ pharmacies because the state has directed them to collect it. As such, Vermont’s interest in controlling the further dissemination of that information is not conceptually different from California’s interest in stemming the further dissemination of information in the hands of local police departments. Under the majority’s reasoning, United Reporting hinges on the fact that the City of Los Angeles used its own police officers — rather than the private prison or security contractors it might have relied on — to process and house its arrestees. See Clifford J. Rosky, Force, Inc.: The Privatization of Punishment, Policing, and Milita^ Force in Liberal States, 36 Conn. L.Rev. 879, 903 (2004) (noting the rapid growth of private prisons and them use in more than half the country). I see no basis for reading United Reporting that narrowly.

But second, the majority’s attempt to distinguish United Reporting would lead to the rather startling proposition that the First Amendment rights, if any, of those seeking access to information turn on whom they are requesting it from. Under the majority’s analysis, for example, the Family Educational Rights and Protection Act — which prohibits universities from disseminating information collected about enrolled students, see 20 U.S.C. § 1232g(b)(1) — operates as a permissible restriction on access to information if a request for student records is denied by a public university but implicates the requestor’s First Amendment rights if it leads to a denial by a private school. I find that outcome both illogical and untenable. Cf. United States v. Miami Univ., 294 F.3d 797, 820-24 (6th Cir.2002) (interpreting FERPA and rejecting asserted “First Amendment right of access to student records”). Indeed, for the putative gatherer of information, the difference is of no discernable let alone constitutional significance. Cf. Houchins v. KQED, Inc., 438 U.S. 1, 11, 98 S.Ct. 2588, 57 L.Ed.2d 553 (1978) (“There is an undoubted right to gather news ... but that affords no basis for the claim that the First Amendment compels others — private persons or governments — to supply information.” (plurality opinion) (emphasis added)).

No doubt sensing the tenuous nature of that position, the majority argues that appellants “have not claimed a First Amendment right to obtain information” but instead challenge section 17 insofar as it regulates the “use of information” already “in [their] hands.” Maj. Op. at 273. Cf. United Reporting, 528 U.S. at 40, 120 S.Ct. 483 (“This is not a case in which the government is prohibiting a speaker from conveying information that the speaker already possesses.”). The argument rests on a fundamental misunderstanding of section 17 — of the “sequence of events” that it regulates. Because, as noted, the majority begins at the end of that sequence, it *286ignores the fact that section 17 regulates the flow of PI data well before it ever comes to be “in the hands” of appellants. Indeed, under operation of the law, appellants can only possess PI data if they have obtained it from pharmacies on the condition that it not be used for “marketing or promoting of a prescription drug.” Having thus obtained PI data with conditions clearly attached, appellants cannot subsequently contend those conditions amount to restrictions on information they “already possess.”

I do not question the proposition that different considerations apply where the government is “prohibiting a speaker from conveying information that the speaker already possesses.” I simply conclude that none of the appellants in this case are so affected by operation of section 17. Nor do I pass on the concern — not pressed by appellants here — that selectively restricting access to information may raise First Amendment concerns. United Reporting, 528 U.S. at 42, 120 S.Ct. 483 (Scalia, J., concurring) (allowing selective access may create “restrietion[s] upon speech rather than upon access to government information”); id. at 43, 120 S.Ct. 483 (Ginsburg, J., concurring) (selective restrictions on access could “impermissibly burden[ ] speech” where selection is based upon an “illegitimate criterion”). I simply conclude that, based on the record before this Court, section 17 operates as a permissible restriction on access to information that the government has directed pharmacies to collect, and the majority errs in concluding to the contrary.

II.

Because I thus conclude that section 17 should be upheld as a permissible restriction on access to information, I could end my analysis there. The majority, however, proceeds to the question of whether, as applied to appellants, Vermont’s law regulates conduct or speech. Because I view that issue as one of some importance, and because I am deeply troubled by the majority’s discussion of it, I, too, address the issue in order to express considerable doubt that, as applied to the data mining appellants in particular, section 17 can properly be characterized as a restriction on speech. In considering the law as applied to data miners and pharmaceutical companies, I once again reject the majority’s approach and follow the “sequence of events” the law regulates, beginning, here, with the restriction as applied to the data miners.

As a preliminary matter — overlooked by the majority — the parties dispute whether section 17 actually restricts the data miners at all. Indeed, section 17 makes no mention of data miners or vendors. Accordingly, it is not clear to me that data miners have any interests — First Amendment or otherwise — at stake here. Section 17, would, at most, appear to eliminate a substantial market for data miners’ services by eliminating the desire of pharmaceutical companies to purchase marketing information the statute prohibits them from using. As the First Circuit recently observed, however, “the First Amendment does not safeguard against changes in commercial regulation that render previously profitable information valueless.” Ayotte, 550 F.3d at 53 (quoting Wine & Spirits Retailers, Inc. v. Rhode Island, 418 F.3d 36, 48 (1st Cir.2005)). Nevertheless, because section 17 restricts “other similar entities” from “sell[ing], licensing], or exchanging] for value” PI data if the transfer is made “for marketing or promoting a prescription drug,” and because a data miner could conceivably be deemed a “similar entity” and thus so regulated, I proceed to consider the law as it might be applied to them.

*287The question, thus, is whether that restriction, should it be imposed, infringes data miners’ First Amendment rights. There are significant reasons to conclude that it does not. As the majority concedes, these data miners — who disingenuously style themselves “publishers” for purpose of this litigation- — “do not themselves use PI data” but instead “are in the business of aggregating and selling data.” Maj. Op. at 274. Nevertheless, citing our opinion in Universal City for the proposition that “[t]he First Amendment protects ‘even dry information, devoid of advocacy, political relevance, or artistic expression,’ ” Maj. Op. at 271-72 (citing Universal City Studios, Inc., 273 F.3d at 446) (alteration omitted), the majority concludes that the data miners’ sale of that “dry information” constitutes protected speech, even implying that it may constitute non-commercial speech. Id. at 271, 274.

I do not read Universal City to support such a sweeping proposition. There, we observed in dicta that “even dry information” may be protected “speech” and held, specifically, that “computer programs constructed from code[ ] can merit First Amendment protection,” 273 F.3d at 446, 449 (emphasis added); see also id. at 445 (noting that in the modern age, this Court has taken “an ‘evolutionary’ approach ... favoring ‘narrow holdings that would permit the law to mature on a ‘case-by-case’ basis”) (quoting Name.Space, Inc. v. Network Solutions, Inc., 202 F.3d 573, 584 n. 11 (2d Cir.2000)). On the facts of that case, we concluded that the computer code in question warranted First Amendment protection because it had the capacity to communicate information to human beings and had promoted both “discourse among computer scholars” and the “exchange of ideas and expression.” Id. at 448. However, in so doing, we distinguished Commodity Futures Trading Commission v. Vartuli, 228 F.3d 94, 111 (2d Cir.2000) (Sack, J.), where we found that the computer program in question there did not warrant First Amendment protection on the ground that “the values served by the First Amendment were not advanced by [the Vartuli code].” Id. at 449 (citing Vartuli, 228 F.3d at 111); see also Vartuli 228 F.3d at 111 (noting that those “values” include “the pursuit of truth, the accommodation among interests, the achievement of social stability, the exposure and deterrence of abuses of authority, personal autonomy and personality development, [and] the functioning of democracy”).

Accordingly, the critical question in applying Universal City is not merely whether the appellants are engaged in the sale of “dry information” but rather whether they are engaged in a sale of “dry information” that “advance[s]” the “values served by the First Amendment.” Cf. Vartuli, 228 F.3d at 111 (“Language serves a variety of functions, only some of which are covered by the special reasons for freedom of speech.”) (quoting Kent Greenwalt, Speech and Crime, 4 Am. B. Found. Res. J. 645, 784 (1980)). Here, there are strong reasons to question whether the data mining appellants are engaged in conduct that meets that standard. As the majority characterizes them, the data mining appellants are in the “business of aggregating and selling data” — data which communicates nothing about them nor allows them to express or communicate anything at all. Maj. Op. at 274.

To be clear, the dissemination of dry information can qualify for First Amendment protection. For instance, as we observed in Universal City, “courts have subjected to First Amendment scrutiny restrictions on the dissemination of technical scientific information and scientific research.” Universal City, 273 F.3d at 447 (internal citations omitted); see also Miller v. California, 413 U.S. 15, 34, 93 S.Ct. *2882607, 37 L.Ed.2d 419 (1973) (“The First Amendment protects works which, taken as a whole, have serious literary, artistic, political, or scientific value.” (emphasis added)). But here, data mining appellants do not contend on appeal that section 17 precludes them from distributing data to foster scientific or medical research. To the contrary, to the extent Vermont’s law applies to them at all, it merely prevents them from licensing their data for a single use — the marketing of prescription drugs. Nor do data mining appellants contend the statute prohibits them from fostering public opinion or debate — to the contrary, as noted above, data mining appellants actually prohibit their customers from disclosing the data they license to anyone else, much less the general public. As such, I have some difficulty comparing the data they sell to “discourse” or the “exchange of ideas.”

The First Circuit, in evaluating a similar law, concluded that PI data was just a product, not distinguishable from the data miners’ perspective to widgets, or, as the First Circuit suggested, “beef jerky.” Ayotte, 550 F.3d at 53. As such, the court found that “this is a situation in which information itself has become a commodity”- — -an “informational asset.” Id. at 53; cf Reno v. Condon, 528 U.S. 141, 148, 120 S.Ct. 666, 145 L.Ed.2d 587 (2000) (sale of collected driver information proper subject of federal regulation because the “information is, in this context, an article of commerce”). Under these circumstances, that court was unwilling to conclude that simply because a party’s “product is information” that “any regulation [of that product] constitutes a restriction on speech.” Ayotte, 550 F.3d at 53. Such an interpretation, it concluded, “stretches the fabric of the First Amendment beyond any rational measure.” Id.

The majority rejects, out of hand, the First Circuit’s “beef jerky” analogy and labels “obscure” its distinction between speech and “information asset[s].” I do not necessarily mean to endorse that court’s approach or even its ultimate conclusion. But I am deeply troubled by the fact that the majority opinion — which becomes the first circuit-level opinion to hold that data miners’ sale of PI data constitutes First Amendment activity7 — does not even bother to engage in the fundamental First Amendment analysis our case law requires. The majority offers no cogent reason for why this “dry information” falls into the category the First Amendment protects, nor any discussion of how this “dry information” can be deemed to “advance” the “values served by the First Amendment.” See Vartuli, 228 F.3d at 111.

To reiterate, I do not question that dry information may be of First Amendment importance given the role information frequently plays in forming public opinion or fostering the marketplace of ideas. Indeed, dry information — in the form of a professor’s research or a programmer’s code — may frequently be of core First Amendment value. But in an era where “increasingly, information is sold as a commodity without being embedded in any practice that could reasonably be regarded as an effort to communicate,” Robert Post, Prescribing Records and the First Amendment — New Hampshire’s Data-Mining Statute, New Eng. J. Med., Feb. 19, 2009 at 745, 746,1 am unwilling to presume that *289simply because a business is engaged in the transfer of information rather than widgets that its activities are automatically entitled to the potent shield of the First Amendment. And I cannot join a majority opinion that offers no principled basis for determining when such conduct should and should not be considered protected First Amendment activity.

With respect to the pharmaceutical companies, section 17 primarily prohibits them from accessing and acquiring PI data for a particular purpose — i.e., for use in marketing — and assuming they do acquire it, prohibits them from using it for that purpose. With respect to the first and primary restriction, I would find for the reasons set forth above, that section 17 operates as a perfectly permissible restriction on access to information and thus does not implicate appellants’ First Amendment rights. With respect to the second restriction, I note as I did above that to the extent pharmaceutical companies obtain PI data under the express condition that they cannot use it for marketing purposes, they cannot subsequently be heard to complain that those express conditions-to-receipt operate as restrictions on information already within their possession.

More generally, I question whether First Amendment protection should be afforded to what amounts to a business method or practice, cf. Wine & Spirits Retailers, Inc. v. Rhode Island, 481 F.3d 1, 6-7 (1st Cir.2007) (ban on joint advertising strategies permissible restriction on conduct or business method, not speech), one that itself has no expressive quality, but is instead meant at most to facilitate the delivery of other expressive conduct. There is no dispute that the practice of detailing itself — that is, of delivering a marketing message to doctors — constitutes commercial speech. There is also no dispute, however, that pharmaceutical detailers do not refer to PI data in their conversations with doctors. The data is used, instead, to identify doctors most likely to prescribe particular kinds of drugs so that sales pitches may be effectively directed at them, to monitor the success of these detailing efforts by tracking any changes in the prescribing habits of the doctors thereby targeted, and to compensate detailing personnel based on the success of their efforts.

The majority concludes that section 17 impacts pharmaceutical companies’ “speech” interests because it “affects manufacturers’ ability to promote brand-name drugs to doctors ... by making it harder to identify those physicians for whom the message will be most relevant and to tailor the detailing messages based on individual physicians’ prescribing histories.” Maj. Op. at 274. However, the majority cites no authority for the proposition that the First Amendment provides protection — let alone, the strong protection the majority affords here — for the methods of identifying an audience, and while the process of “tailoring detailing messages” arguably comes closer to First Amendment activity, the record provides little basis for evaluating the extent to which PI data is actually used in that manner. Accordingly, even if section 17 has some minimal and indirect effect on the manner in which detailers “tailor” those messages, that effect is a very thin reed on which to hang a finding that section 17 restricts First Amendment activity rather than conduct. Cf. Rumsfeld v. Forum for Acad. & Inst. Rights, Inc. (FAIR), 547 U.S. 47, 62, 126 S.Ct. 1297, 164 L.Ed.2d 156 (2006) (“ ‘[I]t has never been deemed an abridgment of freedom of speech or press to make a course of conduct illegal merely because the conduct was in part initiated, evidenced, or carried out by means of language, either spoken, written, or printed.’ ”) (quoting Giboney v. *290Empire Storage & Ice Co., 336 U.S. 490, 502, 69 S.Ct. 684, 93 L.Ed. 834 (1949)).

III.

Finally, however, even if I were to conclude that section 17’s total effect on detailing was sufficient to constitute a restriction on commercial speech, I would nonetheless uphold the statute because I would find that it complies with the standard set forth in Central Hudson.

Under Central Hudson, to regulate commercial speech that is “neither misleading nor related to unlawful activity,”8 the government must (1) assert a “substantial interest” to be achieved, and demonstrate that (2) the restriction “directly advances” that interest, and (3) the limitation “is not more extensive than necessary to serve that interest.” Cent. Hudson Gas & Elec. Corp. v. Pub. Serv. Comm’n, 447 U.S. 557, 564-66, 100 S.Ct. 2343, 65 L.Ed.2d 341 (1980); Anderson v. Treadwell, 294 F.3d 453, 460-61 (2d Cir.2002). As we have previously observed, the latter two steps “coalesce to require ‘a reasonable fit between the legislature’s ends and the means chosen to accomplish those ends.’ ” Anderson, 294 F.3d at 462 (quoting Lorillard Tobacco, Co. v. Reilly, 533 U.S. 525, 556, 121 S.Ct. 2404, 150 L.Ed.2d 532 (2001)). Accordingly, while Central Hudson compels more searching review of a restriction on commercial speech than a restriction on pure conduct, it does not require strict scrutiny. See id. at 460 (“[T]he [Supreme] Court has rejected the argument that strict scrutiny should apply to regulations of commercial speech ..., adhering instead to the somewhat less rigorous standards of Central Hudson.” (collecting cases)).

a.

With respect to the first factor, Vermont identifies three “substantial interests” section 17 advances: (1) an interest in “protecting the public health,” (2) an interest in “protecting the privacy of prescribers and prescribing information,” and (3) an interest in “ensuring] costs are contained” in the health care sector. The majority concludes that the first and third constitute “substantial” state interests but that the second is “too speculative” to qualify. Maj. Op. at 275-76. I would conclude that all three constitute “substantial” state interests. With respect to the second, which is the only asserted interest on which the majority and I diverge, I am unable to accept the majority’s conclusion that the state’s interest in medical privacy is “too speculative” to qualify as a substantial interest. The majority’s analysis — which focuses on the evidence, or asserted lack thereof, of section 17’s effect on medical privacy — is relevant only to whether section 17 “directly advances” the state interest.9 It has no bearing on whether that interest is real and substantial, an issue *291which the majority does not directly question. Indeed, neither appellants nor the majority advances any serious argument that the state does not have a legitimate and substantial interest in medical privacy, nor am I aware of any. To the contrary, in an era of increasing and well-founded concern about medical privacy and the rampant dissemination of confidential information, the federal government has repeatedly acted on that interest and legislated to protect the privacy of medical records, see, e.g., 45 C.F.R. §§ 164.501-164.520 (protecting information collected pursuant to the Health Insurance Portability and Accountability Act); 42 U.S.C. § 2000ff et seq. (protecting privacy of genetic information); 42 C.F.R. §§ 431.300, 431.303 (protecting records of Medicaid patients), and thirteen states and the District of Columbia have considered or enacted bills aimed at protecting medical privacy in the very same way Vermont’s statute does. See Br. of Amicus Curiae Elec. Privacy Inf. Ctr. (“EPIC”) at 2 (collecting statutes). Accordingly, I would find that all three of the state’s asserted interests are “substantial” for purposes of Central Hudson and proceed to evaluate whether section 17 “directly advances” those interests.

b.

The second and third prongs of the Central Hudson test require us to consider whether the regulation at issue “directly advances” the asserted state interests as well as whether the restriction “is not more extensive than necessary to serve th[ose] interest[s].” Cent. Hudson, 447 U.S. at 564, 566, 100 S.Ct. 2343. To meet these requirements, the government carries “the burden of establishing a reasonable fit between the [law’s] ends and the means chosen to achieve those ends.” City of Cincinnati v. Discovery Network, Inc., 507 U.S. 410, 414, 113 S.Ct. 1505, 123 L.Ed.2d 99 (1993) (internal quotation marks omitted). However, as we have recently observed, a “reasonable fit” is not a “least restrictive means” test, Clear Channel Outdoor, Inc. v. City of New York, 594 F.3d 94, 104 (2d Cir.2010), and thus we do not ask whether there is “no conceivable alternative” but instead demand “ ‘only that the regulation not burden substantially more speech than is necessary to further the government’s legitimate interests.’ ” Id. (quoting Bd. of Trs. of the State Univ. of N.Y. v. Fox, 492 U.S. 469, 478, 109 S.Ct. 3028, 106 L.Ed.2d 388 (1989)). The critical inquiry, as the district court noted, is therefore whether the restriction on speech is “in reasonable proportion to the substantial state interest^] served.”10 Sorrell, 631 F.Supp.2d at 454 (internal quotation marks omitted).

With respect to these factors, the government cames the burden of showing *292that its law furthers at least one interest “in a direct and material way,” Edenfield v. Fane, 507 U.S. 761, 767, 113 S.Ct. 1792, 123 L.Ed.2d 543 (1993), and accordingly we ask whether the state has demonstrated “that the harms it recites are real and that [the restriction] will alleviate them to a material degree.” Anderson, 294 F.3d at 462 (internal citation and quotations omitted). In evaluating whether the government has met that burden, the parties dispute the level of deference, if any, we owe to the legislature’s determination. Specifically, the parties dispute whether we should apply so-called Turner deference and thereby “accord substantial deference to the predictive judgments” of legislative bodies which, as “institution^ [are] far better equipped than the judiciary to amass and evaluate the vast amounts of data bearing upon legislative questions.” Turner Broad. Sys. Inc. v. Fed. Comm. Comm’n, 520 U.S. 180, 195, 117 S.Ct. 1174, 137 L.Ed.2d 369 (1997) (internal quotation marks and citations omitted). Like the majority, I feel no need to decide the issue, as I would conclude that even without applying Turner deference, Vermont meets its burden. Because I feel the majority overstates that burden, however, I explain briefly what I consider the prevailing standard to be.

As appellants correctly note, Turner did not address a restriction on commercial speech, a context in which the Supreme Court, independent of Turner, has repeatedly urged deference to legislative findings. See Ayotte, 550 F.3d at 93 (Lipez, J., concurring in part and dissenting in part) (“[T]he general principle of legislative deference” articulated in Turner “also is compatible with the Court’s commercial speech precedent.”). Specifically, the Court has found that the commercial speech doctrine allows “some room for the exercise of legislative judgment,” 44 Liquormart, Inc. v. Rhode Island, 517 U.S. 484, 508, 116 S.Ct. 1495, 134 L.Ed.2d 711 (1996) (plurality opinion), and cautioned that, where a legislature has deemed a particular regulation a properly tailored response to a substantial interest, “we have been loath to second-guess the [g]overnment’s judgment to that effect.” Fox, 492 U.S. at 478, 109 S.Ct. 3028. Accordingly, as we recently observed in upholding a commercial speech regulation, “if [a government] determination about how to regulate [commercial speech] is ‘reasonable’ ... then we should defer to that determination.” Clear Channel, 594 F.3d at 104. Such deference is “all the more appropriate” where, as here, the law targets a form of commercial speech that has “traditionally been subject to extensive regulation,” Anderson, 294 F.3d at 463, or where the regulation fits within a broader regulatory or policy framework. Cf. Clear Channel, 594 F.3d at 105 (“[I]t is not this Court’s role to second guess the City’s urban planning decisions.”)

Accordingly, in evaluating legislative findings and conclusions in the context of a commercial speech regulation, we do not necessarily demand hard evidence, particularly, where, as here, the statute had yet to take effect when first challenged, but instead ask “whether the government is able to support its restriction on speech by adducing] either empirical support or at least sound reasoning on behalf of its measure.” Ayotte, 550 F.3d at 93 (Lipez, J., concurring in part and dissenting in part) (internal quotation marks and alterations omitted) (emphasis added); see also id. at 55 (“A state need not go beyond the demands of common sense to show that a statute promises directly to advance an identified government interest.” (citing *293Burson v. Freeman, 504 U.S. 191, 211, 112 S.Ct. 1846, 119 L.Ed.2d 5 (1992))).

The majority, while declining to determine what level of deference is appropriate, contends that Clear Channel should be limited to the context of “commercial billboards.” Maj. Op. at 279-80. There is nothing in the language of that opinion to suggest as much, and indeed, the Clear Channel opinion cites Ward v. Rock Against Racism, 491 U.S. 781, 109 S.Ct. 2746, 105 L.Ed.2d 661 (1989) — a case that did not involve outdoor advertising at all— for the proposition that deference to a government’s determination of reasonableness is appropriate. See Clear Channel, 594 F.3d at 104. Moreover, as noted above, Clear Channel is entirely consistent with a much broader body of our case law making clear that deference to legislative findings in the context of restrictions on commercial speech — and, particularly, commercial speech in a heavily regulated industry — is appropriate.

Accordingly, as I proceed to ask whether section 17 “directly advances” at least one of the three asserted government interests and whether it is “not more extensive than necessary to serve th[ose] interest[s],” I engage in de novo review of the record. Bose Corp. v. Consumers Union of U.S., Inc., 466 U.S. 485, 499, 104 S.Ct. 1949, 80 L.Ed.2d 502 (1984). But in so doing, I do not substitute my judgment for that of the legislature and instead defer to that body’s determinations where “reasonable.” Clear Channel, 594 F.3d at 94; see also Ayotte, 550 F.3d at 93 (Lipez, J., concurring in part and dissenting in part) (“If the government makes the requisite showing, we defer to the legislative judgment to adopt the challenged measure.”). Moreover, I am cognizant of the context in which the restriction was passed and examine section 17 “in relation ‘to the overall problem the government seeks to correct.’” Clear Channel, 594 F.3d at 94 (quoting Ward, 491 U.S. at 801, 109 S.Ct. 2746). Engaging in such review, I would conclude that the statute directly advances each of the asserted interests in a material manner, and that it is “reasonably proportional” which is to say that it does not burden “substantially more speech than is necessary to further the government’s legitimate interests.” Id. at 104.

c.

First, I would find on this record that section 17 “directly advances” all three of Vermont’s asserted substantial interests. With respect to cost containment and the public health, the district court found, and the record supports the finding that, section 17 materially advances both. The record establishes that pharmaceutical companies spend billions to “detail” new brand name prescription drugs that are more expensive, although not necessarily more effective, than generic class equivalents and whose effects and potential risks are less well known than those associated with generic class equivalents. Sorrell, 631 F.Supp.2d at 451-54. The record further establishes that detailing works — doctors who are “detailed” are more likely to prescribe new brand name drugs, despite the fact that generic class equivalents are more cost-effective and their risks are better known. Id. Finally, the record establishes that PI data is a critical tool for increasing the effectiveness of detailing. IMS Health, for example, promises “big returns” for its PI data clients, noting that a sample client “increased its market share 86% with PI data.” Id. at 451.

Vermont thus took the reasonable course of restricting use of that critical tool. By preventing pharmaceutical companies from using PI data, section 17 makes detailing less effective, which in turn, makes it less likely that doctors will *294prescribe less cost-effective, and potentially riskier brand name drugs over generic class equivalents. That “sound reasoning,” which is amply supported by the testimony of expert witnesses — including some of appellants’ witnesses — and other evidence adduced by the state, is sufficient to satisfy the second prong of the Central Hudson standard. Ayotte, 550 F.3d at 93 (Lipez, J., concurring in part and dissenting in part).

The majority, in concluding otherwise, does not dispute any of the state’s evidence or contest the district court’s findings. Instead, it argues the “route” by which section 17 furthers the state’s interests is “too indirect to survive intermediate scrutiny.” Maj. Op. at 279. However, it is that very same “route” that the majority travels in order to find a First Amendment implication — and thus a need to apply Central Hudson — in the first place. As the majority argues, section 17 implicates First Amendment interests because it restricts access to PI data which in turn “affects manufacturers’ ability to [detail] ... by making it harder to identify those physicians for whom the message will be most relevant and to tailor the detailing messages based on individual physicians’ prescribing habits.” Maj. Op. at 274. In other words, the majority’s First Amendment holding is premised on the understanding that section 17 not only travels that route but travels that route successfully — it achieves its purpose of making detailing more difficult and less effective, which in turn promotes the state’s asserted interests in controlling costs and protecting the public health. Cf. Sorrell, 631 F.Supp.2d. at 451 (“strongest evidence” that section 17 advances state interests is the fact that “if PI data did not help sell new drugs, pharmaceutical companies would not buy it.”) Having found section 17’s route sufficiently direct to establish the First Amendment violation in the first place, the majority’s conclusion that the statute is too indirect to survive Central Hudson is nothing short of bewildering.

No doubt, there are more direct ways Vermont could contain costs or promote health, many of them, I note, far more restrictive of detailers’ activities and First Amendment conduct than the regulation actually passed. But that is not what the second prong of the Central Hudson test requires. Instead, all that standard demands is that the “harms” the state identifies “are real and that [the] restriction will in fact alleviate them to a material degree.” Anderson, 294 F.3d at 462. I would find, on this record, that Vermont meets that standard. The evidence developed below and unchallenged by the majority here establishes that the harms— i.e., exorbitant health care costs and threats to patient safety — are real, and that section 17, by restricting access to PI data, makes detailing more difficult and less effective, which, in turn, reduces the pressure on doctors to prescribe more expensive, less proven drugs. Indeed, as discussed above, the majority agrees that section 17 is likely to be effective in this regard.

Moreover, I note that I would also find that section 17 “directly advances” the state’s third interest — i.e., in “protecting the privacy of prescribers and prescribing information.” Without question, the law restricts the flow of otherwise private information about doctors’ prescribing habits and the care they provide to their patients. No party seriously disputes that. Appellants contend that the interest cannot be deemed “directly advanced” because section 17 still permits the sale and use of PI data for other purposes. As a preliminary matter, I note that the record supports the conclusion that section 17 does not just reduce but dramatically reduces the spread of PI data. As the district court *295found, with respect to PI data, pharmaceutical companies are the data mining appellants’ “only paying customers.” Sorrell, 631 F.Supp.2d at 451. More important, what amounts to an “underinclusiveness” argument is not availing in the context of Central Hudson, which does not require strict scrutiny. See Posadas de Puerto Rico Assocs., 478 U.S. 328, 342, 106 S.Ct. 2968, 92 L.Ed.2d 266 (1986) (statute’s “underinclusive[ness]” not controlling of determination as to whether it “directly advances” state interests); Clear Channel, 594 F.3d at 110 (“[T]he Supreme Court has made clear that underinclusiveness will not necessarily defeat a claim that a stat interest has been materially advanced.”). All that Central Hudson demands is that a regulation materially advance a real harm, which section 17 plainly does.

Accordingly, I would find that section 17 meets the second Central Hudson factor.

d.

The third Central Hudson factor requires consideration of whether the statute is “not more extensive than necessary to serve” the asserted state interests. Because, as noted, this “narrow tailoring” requirement is not a “least restrictive means” test, we look only for a fit “that is not necessarily perfect, but reasonable” and ask whether the restriction is one “whose scope is in proportion to the interest served.” Greater New Orleans Broad. Ass’n, 527 U.S. at 188, 119 S.Ct. 1923.

Because we thus look for “proportion[ality],” the inquiry inherently requires us not simply to evaluate the extent to which the statute furthers the state interests, but also to quantify and then balance the actual burden imposed on speech. It is this latter inquiry that the majority wholly sidesteps in its analysis but that I begin with, because to the extent section 17 restricts commercial speech — a finding that, as set forth above, I doubt — the restriction imposed is both minimal and indirect. At most, section 17 indirectly limits the message detailers convey by preventing them from “tailoring” their message based on a particular doctor’s past prescribing habits. The law does not otherwise affect the message they deliver, nor does it directly restrict detailing in any way. Indeed, as the majority notes, section 17 “does not ... directly restrict the marketing practices of detailers.” Maj. Op. at 277.

Given that minimal and indirect burden on speech, section 17 is inherently distinct from the sorts of “categorical” and direct bans on commercial speech the Supreme Court has previously struck down. See Ayotte, 550 F.3d at 97 (Lipez, J., concurring in part and dissenting in part) (“[T]he restriction on speech imposed by the Prescription Act is significantly more limited than similar restrictions on commercial speech that have been considered by the Supreme Court. It is neither a complete ban on the marketing or advertising of a product ... nor a blanket prohibition on in-person solicitation.”) (internal citations omitted). It is with that limited burden imposed by section 17 in mind, that I consider the “proportion[ality]” of the law.

I would find that the minimal and indirect burden section 17 imposes on speech is not “more than is necessary to further” the government’s three asserted interests. Clear Channel, 594 F.3d at 104 (quoting Fox, 492 U.S. at 478, 109 S.Ct. 3028). The statute directly advances three substantial state interests in material ways, and it does so by imposing exceedingly limited burdens on commercial speech. As such, I find a “reasonable fit” between the burdens imposed and the interests furthered. In so finding, I would note that many of the alternatives proposed by appellants and the majority are actually far more *296restrictive of appellants’ activities. For example, the data mining appellants suggest the state could instead “limit advertising of drugs that it concluded were unnecessarily expensive,” while the majority suggests, “mandatfing] the use of generic drugs as a first course of treatment ... for all those patients receiving Medicare Part D funds.” Maj. Op. at 280. The state instead adopted a regulation that promotes all three interests without directly regulating speech or the content of detailers’ messages, and without unduly interfering in the prescribing habits of doctors. As such, I would find it to be a “reasonable” regulatory choice, one that deserves deference from this Court. See Clear Channel, 594 F.3d at 104.

The majority contends that section 17 cannot be deemed “narrowly tailored” because it is overinclusive in several respects. First, the majority contends that section 17 is over-inclusive because it applies “without regard to whether the data pertains to a prescription drug that is efficacious.” Maj. Op. at 281. However, the very harm section 17 seeks to avoid is aggressive marketing of drugs whose efficacy is not yet known because the drug has not been subject to much actual use or patient experience. Alternatively, the majority contends that section 17 is over-inclusive because it applies even where no generic alternative exists or where a new drug is “unique.” The majority’s analysis, however, overlooks the state’s third asserted interest — that in protecting medical privacy. Because I do not overlook that interest, I would reject both overinclusiveness arguments on the ground that section 17 furthers the state interest in protecting medical privacy by prohibiting the transfer of PI data for marketing purposes irrespective of whether the brand-name drug being detailed is effective or has a generic equivalent.

Alternatively, the majority contends that section 17 is not “narrowly tailored” because Vermont failed to consider “less speech-restrictive means available.” Maj. Op. at 280. As noted, among those “less speech-restrictive” measures the majority posits are mandating the use of generic drugs. Alternatively, the majority suggests that, among other things, Vermont could await the results of a “counter-speech” measure already adopted by the state. First, none of these “less restrictive” means would address all three state interests because none would further the state’s substantial interest in protecting medical privacy. That alone is grounds for accepting the state’s decision not to seriously pursue those alternatives. Cf. Thompson v. Western States Med. Ctr., 535 U.S. 357, 371, 122 S.Ct. 1497, 152 L.Ed.2d 563 (2002) (“[I]f the Government could achieve its interests in a manner that does not restrict speech ... the Government must do so.”).

But second, as noted above, many of the less speecA-restrictive alternatives the majority considers to be “available” are, in fact, far more intrusive restrictions on appellants’ business practices or doctors’ prescribing habits. And while Central Hudson and its progeny make clear that a state may not default to speech restrictions where other, equally effective remedies are available, I do not read that body of law to require a state to adopt far more restrictive and intrusive measures simply because the less restrictive measure imposes an incidental burden on speech.

Finally, where, as here, the state is legislating within an already heavily regulated field, we owe particular deference to the specific regulatory choice the state makes. See Anderson, 294 F.3d at 463. Especially in that context, it is not the role of this Court to “second guess” a legislature’s decision as to which regulatory approach is *297best. See Fox, 492 U.S. at 478, 109 S.Ct. 3028; Clear Channel, 594 F.3d at 105. It is, instead, our role to ensure that the restriction chosen is “reasonably proportional” to the interests it furthers. Section 17 meets that standard. Indeed, the majority offers no significant argument to the contrary — it does not engage in proportionality analysis at all — and instead converts the “reasonable proportionality” standard into a far more aggressive form of inquiry which in effect, if not form, bears striking resemblance to strict scrutiny.

I am unwilling to proceed down that road, particularly where, as here, the law restricts the sale and use of an informational product — PI data — and does not directly limit commercial speech. Because I would find that section 17 constitutes a reasonable restriction that satisfies Central Hudson, I would defer to the state’s conclusion that this particular method of furthering its substantial interests is best. See Clear Channel, 594 F.3d at 105. I would thus conclude that to the extent section 17 can be construed as a restriction on commercial speech, it satisfies Central Hudson and should therefore be affirmed.

IV.

Because I would find that appellants’ First Amendment challenge fails, I briefly address the data mining appellants’ additional dormant Commerce Clause challenge. I would reject that challenge as well, substantially for the reasons cogently set forth by the district court. See Sorrell, 631 F.Supp.2d at 457-59.

The so-called “dormant Commerce Clause,” which refers to the “negative implication” the Supreme Court has long drawn against state interference in Congress’ constitutional authority to regulate interstate commerce, Dep’t of Revenue of Ky. v. Davis, 553 U.S. 328, 337, 128 S.Ct. 1801, 170 L.Ed.2d 685 (2008), prohibits states from regulating “commerce occurring wholly outside [a] State’s borders.” Healy v. Beer Inst., 491 U.S. 324, 332, 109 S.Ct. 2491, 105 L.Ed.2d 275 (1989). In evaluating whether a state law violates the dormant Commerce Clause, the Supreme Court has articulated two primary concerns: first, a concern about “economic protectionism — that is, regulatory measures designed to benefit in-state interests by burdening out-of-state [interests],” Davis, 553 U.S. at 337-38, 128 S.Ct. 1801 (internal quotation marks omitted); and, second, a concern about “inconsistent legislation” or incompatible cross-state regulatory regimes “arising from the projection of one state regulatory regime into the jurisdiction of another State,” Healy, 491 U.S. at 337, 109 S.Ct. 2491.

Section 17 implicates neither concern. Section 17 does not discriminate against out-of-state entities in favor of in-state competitors nor does it risk imposing regulatory obligations inconsistent with those of other states. Instead it restricts the sale of data collected within the state and the use of that data within the state. That data mining appellants seek to take that data out of state to compile it does not relieve them of restrictions on their instate purchase of that data and in-state resale of that data. Cf. Mills, 616 F.3d at 28 (finding similar Maine statute “implicates none” of the “concerns [] central to the way the Supreme Court has framed the dormant Commerce Clause in its recent opinions”).

Accordingly, I would find no basis in dormant Commerce Clause jurisprudence to disturb Vermont’s statute.

V.

In striking down section 17, the majority not only misconstrues a statutory ban on *298access to private information as a speech restriction, but it then breaks from the law of this Court, first, in labeling data miners’ sale of “dry information” protected First Amendment activity, and, second, in applying an aggressive form of Central Hudson that affords insufficient deference to legislative findings and determinations. As a result, I cannot and do not sign on either to the majority’s outcome or the manner by which it arrives thereto.

As noted above, the transfer of data has become a burgeoning business, with those engaged in such transfers frequently having no intention of engaging in expressive or communicative conduct. For the reasons set forth above, I am unwilling to accept the majority’s conclusion that such business operations have an inherent right to invoke the First Amendment as a shield against reasonable regulation simply because their business deals in “dry information” rather than dry goods. Moreover, I express serious concern that the majority’s discussion not only of the First Amendment interests at issue here but also of the standard imposed by Central Hudson will make it unduly and inappropriately difficult for states to properly and constitutionally regulate in furtherance of substantial interests, including a state’s very serious interest in the protection of private information.

I would thus affirm section 17 as a legitimate restriction on access to information and commercial conduct with few, if any, attenuated effects on First Amendment activity. Alternatively, even were I to conclude that section 17 restricts First Amendment activity, in applying Central Hudson, I would afford far greater deference to the eminently reasonable legislative judgments the state has made here in furtherance of several substantial state interests and the reasonably proportional response its statute effects. Accordingly, I respectfully dissent.

. While the Vermont law is captioned "Confidentiality of prescription information,” it is disingenuously referred to as a "Prescription Restraint Law” by plaintiffs-appellants IMS Health Inc., Verispan LLC, and Source Healthcare Analytics, Inc. (collectively, the "data mining appellants”). Data Mining Appellants’ Br. at 2.

. As discussed further below, "detailing” involves the face-to-face promotion of a particular brand name drug by sales representatives — known as "detailers” — who are employed by the pharmaceutical company that manufactures and distributes that drug and make in-person visits to physicians for the purpose of such promotion.

. The state rules are available at http:// vtprofessionals.org/oprl/pharmacists/rules/ Pharmacy% 20Adopted% 20Rules% 20Effec-tive% 20October% 201,% 202009% 20PDF% 20Version.pdf (last visited Nov. 18, 2010).

. The information commonly sold includes the prescriber’s name and address; the name, dosage, and quantity of the drug prescribed; the date and location at which the prescription was filled; and the patient's age and gender. The patient’s name is encrypted, but this "de-identified” personal data still permits the data miners to track the patient's use of a drug or drugs over time and to associate this use with a given prescriber, payment source, and pharmacy. Accordingly, even as "deidentified,” the data is such that a purchaser would know that “a 50-year-old woman who lives in Central Vermont; has prescriptions filled in Montpelier; [and] is a patient of Dr. Jones in Montpelier ... regularly takes an antidepressant and a cholesterol-lowering drug.” Respondents' Br. at 7.

. As noted above, the law also prohibits such sales by health insurers, self-insured employers, and electronic transmission intermediaries. See Vt. Stat. Ann. tit. 18, § 4631(d). The record, however, is clear that pharmacies are the principal, if not sole, source of the PI data aggregated and then licensed by data mining appellants in this case.

. The rules of professional conduct applicable to pharmacies in Vermont place strict limits on the unauthorized release of "patient or practitioner information,” defining it as "unprofessional conduct” subject to discipline. See Vt. Bd. of Pharmacy Admin. Rules § 20.1(1). Because no pharmacy is a party to this action, neither the First Amendment rights, if any, of pharmacies to sell PI data, nor the impact of these restrictions on the assessment of any such rights need be addressed.

. While Judge Lipez, concurring in part and dissenting in part in Ayotte, argued that the New Hampshire law, as applied to pharmaceutical companies’ use of PI data, restricted commercial speech, he found it "self-evident” that the data miners' "acquisition, aggregation, and sale of prescriber-identifiable data” is "not speech within the purview of the First Amendment.” Ayotte, 550 F.3d at 64 (Lipez, J., concurring in part and dissenting in part).

. While Vermont conceded below that the speech at issue here is not "misleading,” the record provides some evidence to the contrary. For example, one former sales representative testified that PI data was used to create sales presentations that are "very skewed” and “distorted.” Another expert testified that PI data was used to tailor detailing messages such that "information [is] provided in ... a selective manner.” The state does not raise the issue on appeal and thus I do not consider it here but note only in passing that, if construed as a law meant to restrict misleading speech or advertising, section 17 would be subject to far less searching review and would unquestionably be within the bounds of the state's regulatory authority. See, e.g., Edenfield v. Fane, 507 U.S. 761, 768, 113 S.Ct. 1792, 123 L.Ed.2d 543 (1993) ("[0]ur cases make clear that the State may ban commercial expression that is ... deceptive without further justification.” (collecting cases)).

. For similar reasons, I reject the majority's suggestion that Vermont has no legitimate interest in medical privacy because the state allows the dissemination of PI data for certain *291non-marketing purposes. The argument, which also bears on the effectiveness of section 17 in furthering the interest in medical privacy rather than on the legitimacy of that interest, suggests, at most, that section 17 may be “underinclusive.” However, as noted below, underinclusiveness, even if established, is not a basis for voiding a statute under Central Hudson analysis.

. As the majority correctly notes, in Thompson v. Western States Medical Center, 535 U.S. 357, 371, 122 S.Ct. 1497, 152 L.Ed.2d 563 (2002), the Supreme Court observed that "if the Government could achieve its interests in a manner that does not restrict speech, or that restricts less speech, the Government must do so." However, there is no indication that the Court’s observation was meant to displace the entirely consistent principle that Central Hudson does not require consideration of every “conceivable alternative” or amount to a "least restrictive means” test. Instead, the Thompson Court was reacting to the government's failure, there, to "even consider ... any other alternatives” — i.e., to the fact that a restriction on speech “seems to have been the first strategy the Government thought to try.” Id. at 373, 122 S.Ct. 1497; cf. id. at 368, 122 S.Ct. 1497 (affirming that *292Central Hudson controls and finding "no need in this case to break new ground”).