Under Department of Transportation regulations, employees in the aviation, rail, motor carrier, mass transit, maritime and pipeline industries who either fail or refuse to take a drug test must successfully complete a drug treatment program and pass a series of urine tests as a condition of performing any safety-sensitive duties. To prevent cheating, the Department modified its regulations in 2008 to require that such tests be conducted under direct observation. Petitioners, a railway company and several transportation unions, challenge the revised regulation, arguing that it violates both the Administrative Procedure Act and the Fourth Amendment. For the reasons set forth in this opinion, we find the Department’s considered justification for its policy neither arbitrary nor capricious, and although we recognize the highly intrusive nature of direct observation testing, we conclude that the regulation complies with the Fourth Amendment.
I.
Acting pursuant to the Omnibus Transportation Employee Testing Act of 1991, Pub.L. No. 102-143, tit. V, 105 Stat. 917, the Department of Transportation promulgated regulations requiring pre-employment, random, and post-accident drug and alcohol tests for employees throughout the transportation industry. 49 C.F.R. pt. 40. Employees who fail or refuse to take drug tests are barred from performing safety-sensitive duties until they complete a treatment program under the supervision of a substance abuse professional. 49 C.F.R. § 40.285. Employees who successfully complete the program must then pass a “return-to-duty” urine test before resuming safety-sensitive duties. 49 C.F.R. §§ 40.285, .305. During the next twelve months, the employees must also pass at least six unannounced “follow-up” urine tests. 49 C.F.R. §§ 40.307(d), .309.
Prior to the rulemaking at issue in this case, employers had the option of conducting return-to-duty and follow-up tests using so-called “direct observation,” a procedure that requires a same-gender observer to “watch the urine go from the employee’s body into the collection container.” 49 C.F.R. § 40.67(i) (2007). Concerned that employers were underutilizing this option, especially in light of evidence of a growing proliferation of products that facilitate cheating on drug tests, the Department solicited comment on additional procedures to strengthen testing integrity. In 2008, the Department promulgated a regulation requiring transportation industry employers to use direct observation for all return-to-duty and follow-up testing. Procedures for Transportation Workplace Drug and Alcohol Testing Programs, 73 Fed.Reg. 62,910, 62,918 (Oct. 22, 2008) (“Direct Observation Rule”). The regulation also requires that immediately prior to all direct observation tests, employees must raise *203their shirts above the waist and lower their lower clothing so as to expose their genitals and allow the observers to verify the absence of any cheating devices. 49 C.F.R. § 40.67© (2008).
Several transportation industry unions and the BNSF Railway Company, sup1 ported by amicus Association of American Railroads, petition for review. Although the partial disrobing requirement became effective on August 27, 2008, we stayed the direct observation requirement pending our resolution of these consolidated petitions. BNSF Ry. Co. v. DOT, No. 08-1264 (D.C.Cir. Nov. 12, 2008). Petitioners argue that the Department’s decision to impose these requirements violates the Administrative Procedure Act’s (APA) prohibition on arbitrary and capricious agency action and the Fourth Amendment’s protection against unreasonable searches. We consider each argument in turn.
II.
Under the Hobbs Administrative Orders Review Act, we evaluate Department of Transportation orders using the familiar standards set forth in the APA. ICC v. Bhd. of Locomotive Eng’rs, 482 U.S. 270, 282, 107 S.Ct. 2360, 96 L.Ed.2d 222 (1987); 28 U.S.C. § 2342(3)(A). Under that framework, agencies must provide a “rational connection between the facts found and the choice made.” Motor Vehicle Mfrs. Ass’n of the United States, Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983) (internal quotation marks omitted). Petitioners argue that the Department’s promulgation of the revised regulation was arbitrary and capricious under this standard. We disagree.
The Department marshaled and carefully considered voluminous evidence of the increasing availability of a variety of products designed to defeat drug tests. It cited congressional testimony describing the ready availability, through Internet sales, of hundreds of different cheating products, Direct Observation Rule, 73 Fed. Reg. at 62,912, the most elaborate of which is a “prosthetic device that looks like real human anatomy, color-matched,” that can be used to deliver synthetic or drug-free urine, id. at 62,911. The Department also relied on a Government Accountability Office (GAO) report indicating that existing drug testing protocols were inadequate to prevent cheating. According to the report, GAO undercover investigators were able to adulterate their urine specimens even at testing sites that followed then-existing procedures. Id. at 62,912. Based on this and similar evidence, the Department determined it was “not practicable” to ignore the cheating problem. Id. at 62,916.
Petitioners dispute none of this evidence. Instead, they fault the Department for failing to provide direct evidence that employees are actually using cheating devices. Acknowledging that it had no statistics on the rates of actual use of such devices, the Department inferred their use from the anecdotal evidence of their availability. Id. at 62,913. As any successful use of cheating devices would not show up in statistics, the Department reasoned, it was “illogical” to require statistical evidence of cheating. Id. Given that people presumably buy cheating devices to use them, we think this approach quite reasonable. As the Supreme Court said just over two weeks ago, “It is one thing to set aside agency action under the Administrative Procedure Act because of failure to adduce empirical data that can readily be obtained. It is something else to insist upon obtaining the unobtainable.” FCC v. Fox Television Stations, Inc., — U.S. *204-, 129 S.Ct. 1800, 1813-14, 173 L.Ed.2d 738 (2009) (citation omitted).
Petitioners devote most of their effort to a separate argument — that whether or not cheating is a problem generally, the Department acted arbitrarily and capriciously in concluding that returning employees are more likely to cheat than employees not subject to direct observation testing. But the Department’s approach was sound. Acknowledging the intrusiveness of direct observation testing, the Department sought to limit it to situations posing a high risk of cheating, id. at 62,911, and then concluded — reasonably in our view — that returning employees have a heightened incentive to cheat, and that this incentive, coupled with the increased availability of cheating devices, creates such a high risk, id. at 62,916.
The Department’s conclusion that returning employees have a heightened incentive to cheat rested in part upon the heavy sanctions reserved for repeat violations. The Department noted that many employers have adopted “two strikes and out” policies that require termination upon a second drug violation, id. at 62,914, and that in the aviation industry second offenders are subject to a statutory permanent bar on aviation-related employment, id.; see 49 U.S.C. § 45103(c). Petitioners object that the Department’s reasoning is inconsistent with its treatment of post-accident testing. As petitioners point out, although employees involved in accidents are subject to mandatory testing immediately after the event, see, e.g., 49 C.F.R. §§ 199.105(b); 219.201(a); 382.303(b); 655.44(a), that testing is not directly observed, § 40.67(a)-(c). According to petitioners, treating post-accident and returning employees differently is irrational because the former, subject as they are to civil or criminal liability, have just as great an incentive to cheat as the latter.
Petitioners’ argument might have had some force had the Department relied solely on this theory. But it didn’t. Substantial additional evidence supports the Department’s conclusion that returning employees are particularly likely to cheat. Specifically, several substance abuse professionals submitted comments supporting the direct observation requirement, and the Department reasonably put “a great deal of weight” on their assessments, stressing their expertise and first-hand experience in administering the treatment programs and planning the follow-up testing. Id. at 62,914. To be sure, several substance abuse professionals spoke only generally about the cheating problem, but others expressly stated that returning employees in particular have a heightened motive to cheat. One said that “[pjersons who have broken trust with the traveling public by testing positive for a prohibited substance, although they knew they would be drug tested, are high risk for using that substance again and motivated to conceal their conduct.” Comments of Evan M. Peterson, Dep’t of Transp. Docket No. OST-2003-15245 (Sept. 9, 2008). [J.A. 272.] Another said that “those who have tested positive in the past, and who continue to abuse drugs, are motivated by their addiction to adulterate, substitute, or use prosthetic-type devices to provide a ‘clean’ specimen at the collection site.” Comments of Susan L. Clark, Dep’t of Transp. Docket No. OST-2003-15245 (Sept. 26, 2008). [J.A. 323.] Given the experience possessed by these substance abuse professionals, such assessments provide substantial evidence supporting the Department’s conclusion that returning employees are particularly likely to cheat on drug tests.
Moreover, the Department supplemented its conclusion about returning employees’ motivations with evidence of their ac*205tual behavior. To rebut the argument— offered by several commenters and echoed here by petitioners — that returning employees are lower risk because they have successfully completed drug treatment programs, the Department emphasized data showing that “the violation rate for return-to-duty and follow-up testing is two to four times higher than that of random testing.” Direct Observation Rule, 73 Fed.Reg. at 62,916. Petitioners point out that these statistics measure only failure, not cheating. Indeed, petitioners claim that data showing returning employees’ higher recidivism rates may simply indicate that they are less likely to cheat on drug tests. Theoretically we suppose it might. But the Department was surely entitled to take the contrary view. We can hardly fault the Department for inferring that the reason for higher failure rates is not that returning employees are more honest, but that they are more likely to use drugs. And given that employees who never use drugs are — to say the least— much less likely to cheat on drug tests than those who do, we think it quite reasonable for the Department to see a higher underlying rate of drug use as evidence of a higher risk of cheating.
Finally, petitioners complain that the Department failed to consider less intrusive alternatives. They point out that some commenters suggested that the Department test hair and saliva rather than urine. As the Department explained, however, the Omnibus Testing Act required it to use only testing methods approved by the Department of Health and Human Services, which “ha[d] not approved any specimen testing except urine.” Id. at 62,-917; see also 105 Stat. 917, 955, 957, 959, 963. And although commenters suggested other safeguards such as further training of collection personnel and pursuit of additional legislative authority, the Department responded — again reasonably in our view — that it was pursuing these approaches as well but that they could not substitute for the efficacy of direct observation. Direct Observation Rule, 73 Fed. Reg. at 62,916-17.
In their brief, petitioners suggest some additional less intrusive alternatives, pointing out for example that the government has successfully prosecuted makers of one prominent prosthetic device, the “Whizzinator,” for conspiring to defeat federal drug tests. Petrs.’ Reply Br. 11. But the mere fact that the government can occasionally prosecute makers of some cheating devices hardly renders irrational the Department’s decision to address the risks posed by the host of similar devices still on the market. Petitioners also suggest that the existing regulations, permitting but not requiring direct observation for returning employees, represent an alternative means of adequately ensuring transportation safety. But the Department found that employers, concerned about the effects on “labor-management agreements” and fearing “upsetting employees,” rarely exercise this option. Direct Observation Rule, 73 Fed.Reg. at 62,917. Indeed, amicus Association of American Railroads confirms that direct observation tests “generate resentment and ill will towards management,” Amicus Br. 8, further supporting the Department’s conclusion that the status quo was untenable.
Thus, the Department acted neither arbitrarily nor capriciously in concluding that the growth of an industry devoted to circumventing drug tests, coupled with returning employees’ higher rate of drug use and heightened motivation to cheat, presented an elevated risk of cheating on return-to-duty and follow-up tests that justified the mandatory use of direct observation. We thus turn to petitioners’ argument that the Department’s suspicionless *206use of direct observation for returning employees, as well as the partial disrobing requirement, runs afoul of the Fourth Amendment.
III.
Compelled urine tests are searches for the purposes of the Fourth Amendment’s prohibition on “unreasonable searches and seizures,” U.S. Const, amend. IV. See Skinner v. Ry. Labor Executives Ass’n, 489 U.S. 602, 617, 109 S.Ct. 1402, 103 L.Ed.2d 639 (1989). Although warrantless searches are, “subject only to a few specifically established and well-delineated exceptions,” Arizona v. Gant, — U.S. —, 129 S.Ct. 1710, 1716-17, 173 L.Ed.2d 485 (2009), generally unreasonable, drug tests for transportation safety fall into the “special needs” exception to the warrant requirement. Skinner, 489 U.S. at 619-20, 109 S.Ct. 1402. Under this framework, we may uphold a warrantless search serving “special needs, beyond the normal need for law enforcement,” Vernonia Sch. Dist. 47J v. Acton, 515 U.S. 646, 653, 115 S.Ct. 2386, 132 L.Ed.2d 564 (1995) (internal quotation marks omitted), if, upon conducting a balancing test, we find that the government’s interest in conducting the search outweighs the individual’s privacy interest, id. at 652-53, 115 S.Ct. 2386.
The government’s interest in transportation safety is “compelling,” to say the least. Skinner, 489 U.S. at 628, 109 S.Ct. 1402. “Employees subject to the tests discharge duties fraught with such risks of injury to others that even a momentary lapse of attention can have disastrous consequences.” Id. Petitioners dispute the extent of the cheating problem, but as discussed above, the Department permissibly found it to be great indeed. Cf. Vernonia, 515 U.S. at 662-63, 115 S.Ct. 2386 (reviewing for clear error district court findings of fact regarding the extent of a school’s drug problem). And although the effectiveness of a search compared to available alternatives may be relevant to the government’s interest in conducting the search, see Delaware v. Prouse, 440 U.S. 648, 659, 99 S.Ct. 1391, 59 L.Ed.2d 660 (1979), there is no per se requirement that the government use the least intrusive practicable means, Vernonia, 515 U.S. at 663, 115 S.Ct. 2386. Given the proliferation of cheating devices, we have little difficulty concluding that direct observation furthers the government’s interest in effective drug testing.
Petitioners argue that the unannounced nature of follow-up tests diminishes the need for direct observation testing. We think the Department’s contrary assessment was reasonable. See Skinner, 489 U.S. at 629 n. 9, 109 S.Ct. 1402 (deferring to agency’s rejection of less intrusive alternatives). Though the precise dates of follow-up tests are unannounced, returning employees know they will have to face at least six such tests over the first year of their return to work. § 40.307(d). Armed with such foreknowledge, returning employees can easily obtain and conceal cheating devices, keeping them handy even for unannounced follow-up tests. See Direct Observation Rule, 73 Fed.Reg. at 62,-912 (noting concealability of cheating products). The government thus has a strong interest in conducting direct observation testing to ensure transportation safety.
The other side of the balance is trickier. Individuals ordinarily have extremely strong interests in freedom from searches as intrusive as direct observation urine testing. In this case, however, those interests are diminished because the airline, railroad, and other transportation employees subject to direct observation perform safety-sensitive duties in an industry that is “regulated pervasively to ensure safety.” *207 Skinner, 489 U.S. at 627, 109 S.Ct. 1402. That said, when the Supreme Court recognized the diminished nature of transportation employees’ privacy interests and found suspicionless drug testing permissible, it stressed that the tests at issue in that case required no direct observation. Id. at 626, 109 S.Ct. 1402; see also NTEU v. Von Raab, 489 U.S. 656, 672 n. 2, 109 S.Ct. 1384, 103 L.Ed.2d 685 (1989) (similar regarding testing of armed customs officers). The Court thus had no occasion to decide whether merely performing safety-sensitive duties in an industry pervasively regulated for safety diminishes employee privacy interests so drastically as to allow direct observation urine testing.
According to the Department, returning employees have diminished privacy interests for reasons over and above their performance of safety-sensitive duties in a pervasively regulated industry. It claims that them privacy interests are diminished by the existing drag testing regulations, which currently permit employers to use direct observation on return-to-duty and follow-up examinations. See supra at 202. To avoid circularity, of course, one’s privacy interests can only be diminished by a valid regulation. Trae, as the Department points out, petitioners don’t challenge the existing regulations, but petitioners contend that under those regulations discretionary direct observation is employed only in cases of reasonable suspicion, a claim the Department never rebuts. Petrs.’ Opening Br. 9; Petrs.’ Reply Br. 17. For our purposes, then, the existing regulations are of limited relevance to the employees’ interests in freedom from the suspicionless direct observation searches required by the challenged regulation.
We see more merit in the Department’s second reason for suggesting that returning employees’ privacy interests are diminished, namely that all have violated the Department’s drug regulations by either refusing to take a test or testing positive. As petitioners make no claim that the drag tests suffer from a false positive problem, the violations were, for the purposes of this case, actual and intentional, and in this sense the Department is correct. By choosing to violate the Department’s perfectly legitimate — and hardly onerous — drag regulations, returning employees have placed themselves in a very different position from their coworkers. Of course, this does not mean, as the Department claims, that returning employees are akin to convicted offenders on probation or parole; after all, the latter are subject to penal sanctions imposed after criminal process. Cf. Griffin v. Wisconsin, 483 U.S. 868, 874, 107 S.Ct. 3164, 97 L.Ed.2d 709 (1987) (“Probation, like incarceration, is a form of criminal sanction imposed by a court upon an offender after verdict, finding, or plea of guilty.” (internal quotation marks omitted)). Nor is the privacy diminution occasioned by the intentional violation of a drug regulation either everlasting or dispositive — even following a fully informed violation, some searches might still be so disproportionate to government interests as to be unreasonable. That said, we have little trouble concluding that employees who have intentionally violated a valid drug regulation, at least in the relatively recent past, see § 40.307 (providing a five-year time limit on followup tests), have less of a legitimate interest in resisting a search intended to prevent future violations of that regulation than do employees who never violated the rule.
We turn, then, to balancing the individuals’ interest with the government’s. Although weighing the strength of each is necessarily imprecise, we think that the employees’ prior misconduct is particularly salient, especially compared to their choice to work in a pervasively regulated industry. It’s one thing to ask individuals seek*208ing to avoid intrusive testing to forgo a certain career entirely; it’s a rather lesser thing to ask them to comply with regulations forbidding drug use. True, direct observation is extremely invasive, but that intrusion is mitigated by the fact that employees can avoid it altogether by simply complying with the drug regulations. On the other side of the balance, the Department has reasonably concluded that the proliferation of cheating devices makes direct observation necessary to render these drug tests — needed to protect the traveling public from lethal hazards — effective. Weighing these factors, we strike the balance in favor of permitting direct observation testing in these circumstances.
Petitioners insist that we reached a different result in NTEU v. Yeutter, 918 F.2d 968, 976 (D.C.Cir.1990). There we held unconstitutional a regulation requiring direct observation for drug tests that rested on reasonable suspicion of drug use but no suspicion of cheating. Id. at 976; see also Piroglu v. Coleman, 25 F.3d 1098, 1101 (D.C.Cir.1994) (citing to Yeutter’s holding without further analysis). In Yeutter, however, we expressly left open the question whether direct observation could ever be permissible, instead relying solely on our conclusion that the direct observation procedures at issue “d[id] not significantly improve testing accuracy.” Yeutter, 918 F.2d at 976. That conclusion in turn rested largely on the premise, supported by the record in that case, that standard monitoring procedures — “collecting outer-garments, dying toilet water, listening for urination” — were adequate to detect cheating. Id. But that was before the Whizzinator and its like. Given the proliferation of such cheating devices, here we have a very different record, one that fully supports the Department’s finding that standard monitoring procedures are inadequate. We thus conclude that here, unlike in Yeutter, direct observation testing will “significantly improve testing accuracy,” id.
Petitioners also claim that the partial disrobing requirement amounts to a strip search. As they acknowledge, however, the balancing inquiry remains the same regardless of how one characterizes the search. See Bell v. Wolfish, 441 U.S. 520, 559-60, 99 S.Ct. 1861, 60 L.Ed.2d 447 (1979) (analyzing cavity search by balancing interests); Stanley v. Henson, 337 F.3d 961, 964 (7th Cir.2003) (“Whether we ... label the process a ‘strip search’ or merely a ‘search’ is unimportant, as the analysis remains the same.”). Applying that analysis, we recognize the intrusiveness of the partial disrobing requirement, but find it only somewhat more invasive than direct observation, which already requires employees to expose their genitals to some degree. Because of this, and because the Department has permissibly found the requirement necessary to detect certain widely-available prosthetic devices, we conclude that it represents a reasonable procedure for situations posing such a heightened risk of cheating as to justify direct observation in the first place.
At oral argument petitioners claimed that no court has ever upheld suspicionless direct observation testing of non-incarcerated civilians. Maybe so, but they cite no ease presenting facts similar to those we face here. Given the combination of the vital importance of transportation safety, the employees’ participation in a pervasively regulated industry, their prior violations of the drug regulations, and the ease of obtaining cheating devices capable of defeating standard testing procedures, we find the challenged regulations facially valid under the Fourth Amendment.
We emphasize the limited nature of our holding. Because petitioners bring a facial challenge, we consider only “whether the *209tests contemplated by the regulations can ever be conducted.” Skinner, 489 U.S. at 632 n. 10, 109 S.Ct. 1402. We thus express no view on either the merits of any as-applied challenge to this rule or the constitutionality of any other rule.
IV.
For the reasons stated above, we deny the petitions for review.
So ordered.