This case presents this court for the first time with the issue of the constitutionality of random and return-to-work drug and alcohol testing of public employees holding "safety sensitive" positions within the public transportation authority serving the Philadelphia metropolitan area.1 The district court upheld the constitutionality of the random drug testing policy, but found the addition of return-to-work testing to be unconstitutional. The court enjoined commencement of random testing of those employees covered by the Railway Labor Act, pending settlement of the dispute through the conciliation procedures specified in the Act. Transport Workers' Union Local 234 v. Southeastern Pennsylvania Transportation Authority, 678 F. Supp. 543 (E.D.Pa.1988).

The Unions argue that indiscriminate use of random drug tests to ferret out illegal drug use by public employees is irreconcilable with the guarantee against unreasonable government intrusion on individual privacy of the Fourth Amendment. We do not underestimate their concern that the public clamor about the nation's drug problem creates the risk that judgments may be distorted in evaluating the constitutionality of government measures taken to curb this crisis. Nonetheless, in light of this court's precedent sustaining drug testing in other occupations and in light of the record presented in this case where the public transportation authority has documented a serious drug problem among its employees, to which it has traced accidents causing injury to the public, we conclude that the random drug testing at issue here is not facially invalid under the Fourth Amendment.

We also uphold the district court's finding that the transportation authority failed to justify the reasonableness of its return-to-work drug testing policy. Finally, we uphold the injunction imposed on commencement of random testing of employees covered by the Railway Labor Act.

Appellants/cross-appellees in this case ("the Unions") are the Transport Workers Union Local 234 ("TWU"), which represents approximately 5,700 of SEPTA's 6,500 employees, and five unions which represent SEPTA employees under section one of the Railway Labor Act ("RLA"), 45 U.S.C. Sec. 151, Sixth (1982): the Brotherhood of Locomotive Engineers, Division 71; the Brotherhood of Locomotive Engineers; the Brotherhood of Railway, Airline and Steamship Clerks; the Brotherhood of Maintenance of Way Employees; and the Brotherhood of Railroad Signalmen ("the Railway Unions").

Appellees/cross-appellants, the Southeastern Pennsylvania Transportation Authority and some of its officers and Board members (referred to collectively as "SEPTA" or "the Authority"), operate mass transportation facilities in the five-county Philadelphia metropolitan area. SEPTA is an agency of the Commonwealth of Pennsylvania, currently organized under the authority of the Pennsylvania Urban Mass Transportation Act of January 22, 1968, P.L. 42, No. 8, as amended, 55 Pa.Stat.Ann. Secs. 600.101-600.407 (Purdon Supp.1988). SEPTA operates subways, railroads, buses, streetcars, and trackless trolleys, and maintains stations, depots, platforms, tracks, and other installations. In 1986, SEPTA accommodated more than 1.2 million passengers on its transit systems on an average week day.

On January 16, 1987, SEPTA publicly announced, without prior notice or consultation with the Unions, that it planned to begin administering random urinalysis tests to detect the presence of psychotropic drug metabolites and alcohol in employees' body systems. On January 21 and 26, 1987, the Unions filed complaints against SEPTA's plan in the United States District Court for the Eastern District of Pennsylvania, alleging that SEPTA's program violated their members' constitutional rights under the Fourth and Fourteenth Amendments.

The Railway Unions also alleged that SEPTA had violated section 6 of the RLA, 45 U.S.C. Sec. 156 (1982), by unilaterally instituting a random drug testing policy without serving notice and exhausting the mediation and conciliation procedures mandated by the RLA before implementation of changes in working conditions or rules under a labor contract.

In early February 1987, SEPTA posted an additional order at its offices announcing that all employees returning to work after an absence from employment of over thirty days for any reason other than vacation would be required to submit to urinalysis testing. The Unions filed amended complaints alleging that SEPTA's return-to-work testing policy also violated the constitutional rights of their members.

After four days of hearings, the district court entered a preliminary injunction enjoining SEPTA from implementing its random drug and alcohol testing program. The court expressed various concerns about, inter alia, the plan's failure to include reliable confirmatory testing procedures, failure to use a less intrusive breathalyzer test to detect the presence of alcohol in employees' body systems, and the lack of justification of the scope of the job positions to be tested. It directed SEPTA to submit any further plan for random testing to it.

In July 1987, after SEPTA published the first draft of a revised drug and alcohol testing and education program which it had designed to respond to the district court's concerns, it moved to vacate the court's preliminary injunction. The case went to trial in September 1987, at which time both SEPTA and the Unions introduced extensive expert testimony about testing technology and the effects of drug use. SEPTA also presented evidence of the scope and severity of the drug problem affecting its workforce.

On September 25, 1987, the district court again delivered bench remarks criticizing various aspects of SEPTA's proposed program. SEPTA again made revisions in its proposal, and resubmitted it to the Unions and the court.

On January 19, 1988, the district court entered its final order and issued a memorandum opinion. The order dissolved the preliminary injunction against SEPTA and denied the Unions' request for a permanent injunction against the random testing component of SEPTA's drug and alcohol testing program, but granted a permanent injunction against implementation of SEPTA's return-to-work testing policy. The court also enjoined SEPTA from commencing random drug testing of workers represented by the Railway Unions, holding that unilateral implementation of such a policy gave rise to a "major dispute" within the meaning of section 6 of the RLA and thus required SEPTA to exhaust the mediation and conciliation procedures specified in that Act. Finally, the court ordered that any final revisions of SEPTA's drug and alcohol testing program be submitted for review prior to implementation. SEPTA filed a motion asking the court to amend its order to allow return-to-work testing, which the district court denied. This appeal and cross-appeal followed.

The final random testing proposal approved by the district court subjects approximately 2,600 workers to random urinalysis testing over a one-year period, and approximately 5,400 workers per year to random breathalyzer testing. Testing is limited to "operating engineers" in "safety sensitive" positions.2

Under the breathalyzer testing program for alcohol, an employee will be considered to test positive if two successive tests yield blood alcohol levels of .04 percent or higher. Under the urinalysis test designed to detect drug metabolites, initial screening will be conducted by an EMIT immunoassay. Positive results obtained by this screening method will be subject to confirmation testing by gas chromatography/mass spectrometry (GC/MS). The parties have stipulated that this two-step procedure is reliable. Employees may also request an additional confirmatory test at their own expense if they wish.

Detailed specifications assure chain of custody and proper laboratory controls. SEPTA will not voluntarily disclose test results to law enforcement officials, and results will be kept confidential within the medical department, being disclosed only on a need-to-know basis. Employees who are taking prescription medication must inform SEPTA's medical department of this fact. The department will make "fitness for duty" determinations with respect to these employees in consultation with the physician prescribing the medication.

Random testing is scheduled to begin ninety days after SEPTA institutes an educational campaign and thirty days after it begins to offer an Employee Assistance Program ("EAP") administered by an outside provider. All employees whose urine or breathalyzer tests produce positive results will be given a "not fit for duty" classification, removed from their jobs, and required as a condition of retaining their employment with SEPTA to enter an EAP for counseling and treatment. After successful completion of treatment, employees will be assigned to an alternative job for an observation period of up to ninety days during which they may be tested at any time. 678 F. Supp. at 547. After employees are returned to their prior positions, they are subject to follow-up tests on an unannounced basis for twelve months and on a scheduled basis for up to two years thereafter.

As an alternative to mandatory treatment following a positive test result, employees may voluntarily refer themselves to the EAP before being tested. Employees who voluntarily enter treatment will be removed from duty until completion of treatment and assigned to alternative duty for ninety days thereafter, for which they will be paid at that job's wage rate. They will be subject to unannounced follow-up testing for six months and scheduled testing for one year thereafter. Employees who voluntarily refer themselves to the EAP are given one additional chance to obtain treatment if their urine produces a positive result after treatment.

Employees who refuse to be tested, receive positive test results after treatment, or refuse to undergo treatment and counseling after a positive result will be discharged.

There is no dispute that, as a government agency, SEPTA is bound by the strictures of the Fourth Amendment. Contrary to the argument put forth by SEPTA in this appeal, it is by now well settled that government drug testing of employees constitutes a search or seizure for purposes of the Fourth Amendment. Railway Labor Executives' Ass'n v. Burnley, 839 F.2d 575, 580 (9th Cir.), cert. granted, --- U.S. ----, 108 S. Ct. 2033, 100 L. Ed. 2d 618 (1988); Everett v. Napper, 833 F.2d 1507, 1511 (11th Cir.1987); Jones v. McKenzie, 833 F.2d 335, 338 (D.C.Cir.1987); National Treasury Employees Union v. Von Raab, 816 F.2d 170, 176 (5th Cir.1987), cert. granted, --- U.S. ----, 108 S. Ct. 1072, 99 L. Ed. 2d 232 (1988); McDonell v. Hunter, 809 F.2d 1302, 1307 (8th Cir.1987); Shoemaker v. Handel, 795 F.2d 1136, 1141 (3d Cir.), cert. denied, 479 U.S. 986, 107 S. Ct. 577, 93 L. Ed. 2d 580 (1986). The same principle is applicable to a compelled breathalyzer test for alcohol. See, e.g., Burnley, 839 F.2d at 580; Shoemaker, 795 F.2d at 1141-42.

The Fourth Amendment guarantees the right of the people against "unreasonable search and seizure." Except in certain carefully defined classes of cases, searches without consent or a valid search warrant are unreasonable. O'Connor v. Ortega, 480 U.S. 709, 720, 107 S. Ct. 1492, 1499, 94 L. Ed. 2d 714 (1987) (plurality opinion) (quoting Mancusi v. DeForte, 392 U.S. 364, 370, 88 S. Ct. 2120, 2124, 20 L. Ed. 2d 1154 (1968)). In a certain narrow class of exceptions, however, where "special needs, beyond the normal need for law enforcement, make the warrant and probable cause requirement impracticable,"New Jersey v. T.L.O., 469 U.S. 325, 351, 105 S. Ct. 733, 747, 83 L. Ed. 2d 720 (1985) (Blackmun, J., concurring), a search may be reasonable even though it is not conducted pursuant to a probable cause determination. Id. at 340, 105 S.Ct. at 742; see also O'Connor v. Ortega, 480 U.S. at 725, 107 S.Ct. at 1502.

In such circumstances, intrusions on employees' constitutionally protected privacy interests must be evaluated by "the standard of reasonableness under all the circumstances." Id. at 725-26, 107 S.Ct. at 1502. Such an evaluation involves " 'balanc[ing] the nature and quality of the intrusion on the individual's Fourth Amendment interests against the importance of the governmental interests alleged to justify the intrusion.' " Id. at 719, 107 S.Ct. at 1499 (quoting U.S. v. Place, 462 U.S. 696, 703, 103 S. Ct. 2637, 2642, 77 L. Ed. 2d 110 (1983)).

The Unions argue that under the constitutional equation used to determine if body fluid testing is reasonable, there must always be particularized or individualized suspicion. In essence, they argue that random testing for drugs or alcohol is never permissible under the Fourth Amendment. The difficulty with their argument is that this circuit has already rejected precisely such a contention, and we are bound to those holdings.

In two prior cases, we applied "the administrative search exception to the Fourth Amendment warrant requirement" to uphold drug testing without individualized suspicion of employees in "highly regulated industries." Policeman's Benevolent Ass'n, Local 318 v. Township of Washington, 850 F.2d 133, 135 (3d Cir.1988); Shoemaker v. Handel, 795 F.2d 1136, 1142-43 (3d Cir.), cert. denied, 479 U.S. 986, 107 S. Ct. 577, 93 L. Ed. 2d 580 (1986).

The administrative search exception has been used to permit legislatively authorized warrantless inspections of commercial property in cases in which the government interest in conducting the search would be frustrated by requiring prior notice. See Donovan v. Dewey, 452 U.S. 594, 598-603, 101 S. Ct. 2534, 2537-40, 69 L. Ed. 2d 262 (1981). Cases applying the administrative search exception generally involve statutory schemes aimed at ensuring safety or other "substantial" government interests. See, e.g., New York v. Burger, 482 U.S. 691, 107 S. Ct. 2636, 2646-47, 96 L. Ed. 2d 601 (1987) (upholding warrantless inspection of automobile junkyards where state had substantial interest in reducing growing problem of motor vehicle theft associated with this industry); Donovan v. Dewey, 452 U.S. at 603, 101 S.Ct. at 2540 (permitting warrantless safety inspections of mines); United States v. Biswell, 406 U.S. 311, 316-17, 92 S. Ct. 1593, 1596-97, 32 L. Ed. 2d 87 (1972) (approving warrantless searches pursuant to gun control legislation); see also Colonnade Catering Corp. v. United States, 397 U.S. 72, 76-77, 90 S. Ct. 774, 776-77, 25 L. Ed. 2d 60 (1970) (approving congressional scheme of warrantless searches and seizures in liquor industry in light of long history of regulation of this industry).

In Shoemaker, this court made a substantial leap beyond the application of the administrative search exception to commercial property by applying that exception to uphold random urine testing of jockeys mandated by the New Jersey Racing Commission under its statutory power to regulate horse racing in that state. We stated that "while there are distinctions between searches of premises and searches of persons, in the intensely-regulated field of horse racing, where the persons engaged in the regulated activity are the principal regulatory concern, the distinctions are not so significant that warrantless testing for alcohol and drug use can be said to be constitutionally unreasonable." 795 F.2d at 1142. Thereafter, in Policeman's Benevolent Association, 850 F.2d at 141, we categorized a township's random drug testing program for police officers as an administrative search and upheld the program against a facial challenge. We explained that a drug testing program will come within the administrative search exception if it is (1) based on a strong state interest in determining whether employees are using illegal substances; and (2) directed at an industry that is pervasively regulated in such a manner that employees' justifiable privacy expectations are diminished. Id. at 136.

The Unions do not argue that SEPTA's drug testing program fails to meet the first prerequisite for administrative searches since they concede that protection of the public safety is an important government interest. The Unions argue, however, that the administrative search exception is inapplicable because there is no basis for finding a diminished expectation of personal privacy from the prior government regulation of the transportation industry.

We agree that the regulations to which SEPTA points differ from those governing jockeys in New Jersey and police conduct in Washington Township. Because the police officers in Policeman's Benevolent Association, 850 F.2d at 138-41, were part of a "quasi-military" organization, they were subject to extensive personal regulations, including conduct rules for their private lives, a rule requiring off-duty carrying of a firearm, and the requirement of constant readiness for duty. Although the jockeys in Shoemaker, 795 F.2d at 1141-42, did not have the "awesome ... power" to arrest and detain citizens that distinguishes police officers, Policeman's Benevolent Ass'n, 850 F.2d at 141, there is intense regulation of horse racing in New Jersey, including the licensing of all employees in the industry, which is designed to instill confidence in the integrity of the horse racing industry. This led us to reason that because the Racing Commission "historically ... exercised its rulemaking authority in ways that have reduced the justifiable privacy expectations of persons engaged in the horse-racing industry," jockeys who accepted a state license did so with knowledge that the Commission would exercise its power to conduct warrantless searches. Shoemaker, 795 F.2d at 1142.

Although the enabling legislation cited by SEPTA establishes that public transportation in Pennsylvania is a highly regulated industry, see 55 Pa.Stat.Ann. Secs. 600.101--600.407 (Purdon Supp.1988), SEPTA does not argue that its regulations, like those involved in our prior cases, implicated the private conduct of the workers themselves. Thus, contrary to SEPTA's argument, the regulatory scheme on which SEPTA relies is not equivalent to the extensive personal regulation involved in Policeman's Benevolent Association or even the lesser regulation in Shoemaker. Hence, the Commonwealth's regulation of the transportation industry in and of itself has not been shown to effect a comparable diminution of employees' expectations of personal privacy.3 Therefore, the testing in this case cannot be upheld as an administrative search.

Contrary to the way the parties have read our cases, the standard for sustaining administrative searches does not represent an independent line of Fourth Amendment analysis. Our reference to administrative searches as an "exception" was as an exception to the warrant requirement. As the plurality opinion in O'Connor v. Ortega makes clear, although "the appropriate standard for administrative searches is not probable cause in its traditional meaning," administrative searches are merely one illustration of the application of the reasonableness standard arising out of the "careful balancing of governmental and private interests." 480 U.S. at 722-23, 107 S.Ct. at 1501. Thus, while we cannot uphold SEPTA's random drug and alcohol testing policy as an administrative search because of the absence of evidence of regulation directed to the employees affected, we proceed to consider it under the more inclusive analysis of "reasonableness under all the circumstances" articulated by the Supreme Court in recent cases.

When justification for a warrantless search conducted without probable cause is sought on the basis of the balancing of the government interest and the individual's privacy expectations, we must apply a two-part test. Courts must look, first, to whether the search was "justified at its inception," and, second, to whether the search as actually conducted "was reasonably related in scope to the circumstances which justified the interference in the first place." New Jersey v. T.L.O., 469 U.S. at 341, 105 S.Ct. at 742 (quoting Terry v. Ohio, 392 U.S. 1, 20, 88 S. Ct. 1868, 1879, 20 L. Ed. 2d 889 (1968)).

Obviously, any erosion of the probable cause standard of the Fourth Amendment must be scrutinized with great care because the " 'overriding function of the Fourth Amendment is to protect personal privacy and dignity against unwarranted intrusion by the State.' " Winston v. Lee, 470 U.S. 753, 760, 105 S. Ct. 1611, 1616, 84 L. Ed. 2d 662 (1985) (quoting Schmerber v. California, 384 U.S. 757, 767, 86 S. Ct. 1826, 1834, 16 L. Ed. 2d 908 (1966)). Individuals cannot be said to relinquish all expectations of privacy simply because they participate in spheres that are regulated by government. See Delaware v. Prouse, 440 U.S. 648, 663, 99 S. Ct. 1391, 1401, 59 L. Ed. 2d 660 (1979) ("[P]eople are not shorn of all Fourth Amendment protection ... when they step from the sidewalks into their automobiles."). "The reasonableness of a warrantless search ... will depend upon the specific enforcement needs and privacy guarantees of each statute." Marshall v. Barlow's, Inc., 436 U.S. 307, 321, 98 S. Ct. 1816, 1825, 56 L. Ed. 2d 305 (1978).

We note at the outset that the constitutional issue presented for our review in this case is a narrow one. We are presently called upon to rule on the constitutionality of SEPTA's random and return-to-work drug and alcohol testing policies as described on the face of its final revised proposal; the application of these policies in practice is not before us for review at this time.4 Our review of the district court's application of constitutional law under the Fourth Amendment is plenary.5

1. The Constitutionality of SEPTA's Random Testing Program

a. The "Initial Justification" for Random Testing

Even within the narrow category of cases in which the Court has found that "special needs" justify relaxing the probable cause requirement for searches, it has generally required "individualized suspicion." See, e.g., Terry v. Ohio, 392 U.S. 1, 21 & n. 18, 88 S. Ct. 1868, 1879 & n. 18, 20 L. Ed. 2d 889 (1968) (police "pat down" searches for concealed weapons do not require probable cause but do require "specific articulable facts, ... together with rational inferences from those facts, [that] reasonably warrant that intrusion"). However, the Court has held that the suspicion prerequisite is not an "irreducible requirement" and that under certain limited circumstances, the balance of interests between the government's need to conduct the search and the degree of intrusion on the individual may justify dispensing with the individualized suspicion requirement. See, e.g., United States v. Martinez-Fuerte, 428 U.S. 543, 561-62, 96 S. Ct. 3074, 3084-85, 49 L. Ed. 2d 1116 (1976) (upholding constitutionality of brief vehicle stops at fixed checkpoints to question occupants without individualized suspicion).

We agree with the Unions that the intrusion on individuals' privacy caused by drug testing is more than minimal, involving an order to perform the highly private activity of urination, an order considered by some to be "offensive to personal dignity," Patchogue-Medford Congress of Teachers v. Board of Educ., 70 N.Y.2d 57, 68, 517 N.Y.S.2d 456, 461, 510 N.E.2d 325, 330 (1987); see also Capua v. City of Plainfield, 643 F. Supp. 1507, 1511 (D.N.J.1986).

We do not believe that this ipso facto renders the search unreasonable. Although in its most recent cases discussing the reasonableness standard for Fourth Amendment search purposes, the Supreme Court did not decide whether individualized suspicion is an essential element, see O'Connor v. Ortega, 480 U.S. at 726, 107 S.Ct. at 1503; New Jersey v. T.L.O., 469 U.S. at 342 n. 8, 105 S. Ct. at 743 n. 8, this court held it was not when upholding drug testing in Shoemaker and Policeman's Benevolent Association as an administrative search. For comparable reasons, we conclude that under the balancing required by the reasonableness analysis, if the state's interest is substantial enough it may override even the intrusive searches required for urinalysis. A fortiori, the less intrusive search by breathalyzer testing would be sustained.

Of course, the more intrusive the search, the more substantial the governmental interest that must be shown to make the requisite balance. The state interest asserted here, that of guaranteeing the safety of the public, has been accepted in other contexts as justifying drug testing without particularized suspicion. See, e.g., Jones v. McKenzie, 833 F.2d at 340 (routine drug testing of bus drivers and attendants entrusted with safety of handicapped school children justified by "significant and compelling governmental interest"); Rushton v. Nebraska Pub. Power Dist., 844 F.2d 562, 567 (8th Cir.1988) (individualized suspicion not required for drug testing in nuclear power industry). Contra Burnley, 839 F.2d at 587-88 (invalidating post-accident drug testing of railroad employees).

As stated in O'Connor v. Ortega, 480 U.S. at 724, 107 S.Ct. at 1502, "in many cases, public employees are entrusted with tremendous responsibility, and the consequences of their misconduct ... [to] the public interest can be severe." The government interest in protecting public safety, often of large groups of persons traveling by mass transit, is patently of a higher order than simply maintaining the "efficient and proper operation of the workplace," found sufficient by the plurality to justify limited searches of the desk and personal files of an employee in O'Connor. See 480 U.S. at 723, 107 S.Ct. at 1501.

We view with amazement the opening sentence in the Railway Unions' brief that "[t]his case involves an overreaction to an exaggerated hysteria conserning the use of drugs and alcohol in our society." Brief of Certain Plaintiffs-Appellants-Cross-Appellees at 1. The enormity of the problem facing this country because of drug use is sufficiently well established to justify judicial notice.

Of course, it was SEPTA's burden to show drug use by its operating employees that poses a danger to the safe operation of its vehicles. The district court made findings that drug use by SEPTA's operating employees presents a significant safety problem and that the Authority's prior suspicion-based testing program proved insufficient "to satisfy the needs of public safety." 678 F. Supp. at 549. Compare Lovvorn v. City of Chattanooga, 846 F.2d 1539, 1541, 1547 (6th Cir.1988) (random drug testing of fire fighters unconstitutional where "[t]here has been no statistical or objective evidence that the performance of any member of the department ... has been affected by the use of drugs") with Jones v. McKenzie, 833 F.2d at 336 (upholding constitutionality of routine drug tests of school bus drivers and attendants following "repeated incidents of bizarre or dangerous drug-related behavior by drivers and attendants while on duty").

There is adequate support for the district court's findings. SEPTA established that in six major accidents occurring in 1986 and 1987, collectively involving at least 89 injuries to persons, the operators of vehicles at fault tested positive for alcohol or drug metabolites. App. at 1017. The operator at fault in another accident refused to submit to a test. Further, of 262 employees tested "for cause" between January 1985 and July 1987, 82 employees (thirty-one percent) tested positive for drug metabolites or alcohol. App. at 1166. Twelve percent of the "new hires" given drug tests within the same period produced positive results (116 of 1,004 tested). App. at 1165. The Unions do not argue that these are statistically insignificant figures. Instead, they argue that the presence of metabolites, particularly of marijuana metabolites, does not demonstrate present impairment and, in the case of employees "at fault," does not demonstrate causation of the accident.

Although there is no evidence of direct correlation between the presence of drug metabolites and impairment at the time of testing, users are frequently able to mask their impairment so that it is not apparent. App. at 347. Also, SEPTA produced testimony establishing that use of drugs in itself results in significant impairment. There was testimony that long after all traces of drugs have left the system, there are significant delayed and chronic impairment effects.

As Union witness Dr. John P. Morgan acknowledged, individuals whose drug use does not render them obviously impaired in performing routine work tasks can display a lack of flexibility and inability to respond to sudden changes or emergency situations. App. at 771. The acute effects of marijuana include distortion of judgment and loss of good judgment, a decrease in alertness and caution, impairment of visual, auditory and space perception, and loss of propriety of responses, such as ability to steer. App. at 373. Cocaine ingestion typically causes aggressiveness, overreaction, and misperception. Id. at 378. According to SEPTA expert witness Robert J. Pandina, residual impairment effects of marijuana and cocaine include irritability, agitation, lack of alertness and awareness, and memory deficits. App. at 356. After the stimulative effects of cocaine have worn off, users' functioning becomes depressed and slower, and judgment remains impaired. App. at 378. Other delayed impairment effects occur during the period in which individuals are experiencing withdrawal from a drug on which they have become physically dependent. App. at 782, Morgan Testimony. Chronic users will experience long-term effects including a desire to increase the dose of drug taken, which will in turn increase the probability of impairment. App. at 360, Pandina Testimony.

While the Unions downplay the statistical evidence of marijuana use, a SEPTA witness testified that the impairment effects of marijuana as measured by the ability to perform certain tasks, such as simulated driving, include statistically significant effects on performance lasting for as long as twelve hours after use in some individuals. App. at 523.

In light of the evidence connecting impairment with drug use, it was appropriate for SEPTA to design its program in an effort to detect drug users. It was not required to limit its detection efforts to those employees whose then-current impairment could be detected, particularly in light of the testimony by experts offered by both the Unions and SEPTA that current urine testing technologies do not permit the detection or measurement of current impairment. See App. at 540, 548, 747, 780, 354-55.

The Unions next argue that SEPTA's program should be rejected because there are less restrictive means to deal with the drug problem of transit employees. The Court has made clear that the least restrictive means analysis applied in cases implicating the First Amendment is not applicable to cases arising under the Fourth Amendment. Colorado v. Bertine, 479 U.S. 367, 373-74, 107 S. Ct. 738, 742-43, 93 L. Ed. 2d 739 (1987) (quoting Illinois v. Lafayette, 462 U.S. 640, 647, 103 S. Ct. 2605, 2610, 77 L. Ed. 2d 65 (1983)); Bell v. Wolfish, 441 U.S. 520, 559 n. 40, 99 S. Ct. 1861, 1884 n. 40, 60 L. Ed. 2d 447 (1979) (quoting United States v. Martinez-Fuerte, 428 U.S. 543, 556-57 n. 12, 96 S. Ct. 3074, 3082-83 n. 12, 49 L. Ed. 2d 1116 (1976)). Of course, a factor relevant to the inquiry into the reasonableness of the search at its inception would be whether less intrusive but equally effective means to the same goal are available. There has been no such showing here. In light of the district court's finding that "[m]any drugs have delayed, residual or chronic effects not ascertainable through behavioral observation," 678 F. Supp. at 550, and that SEPTA's operating employees often work in circumstances which preclude close supervision, making it infeasible to monitor performance or detect signs of drug-induced impairment by direct observation, the Unions have not shown that there is any less intrusive but equally effective means of detecting employee drug use. Id.

Finally, the Unions' argument that SEPTA can rely on its already existing policy of testing for drug use after accidents "chargeable" to driver fault6 does not effectively address the public interest in detecting employees in safety sensitive positions with drug problems before accidents occur.

The safety justification for SEPTA's drug testing policy, SEPTA's documentation of a drug use problem among its workforce, the evidence of the deleterious effect of drug use, and SEPTA's showing of positive tests for drugs or alcohol by operating personnel at fault in accidents are sufficient bases for us to conclude that the random testing policy is constitutionally justified at its inception.

b. The Relationship of the Search to the Need

The Supreme Court has held that even when the search program is designed with important public safety considerations in mind, to survive the required Fourth Amendment balancing there must be sufficient safeguards to protect against abuse of official discretion in deciding whom and how to search and the search must be sufficiently productive to qualify as a reasonable law enforcement practice. Delaware v. Prouse, 440 U.S. 648, 655, 660, 662, 99 S. Ct. 1391, 1397, 1399, 1400, 59 L. Ed. 2d 660 (1979).

The testing policy that developed as a consequence of the district court's supervision of this case is limited to employees whose job performance implicates public safety. Even the non-operating personnel included, such as signal operators and repair persons, perform functions closely connected to the public safety and their classification as "safety sensitive" is justified.

SEPTA's random drug testing program is also reasonable with respect to the type of testing conducted. Following the district court's instructions, SEPTA substituted less intrusive testing where possible, such as breathalyzer tests to detect blood alcohol levels. Although urine testing can be used to reveal private information other than drug use, see National Treasury Employees Union v. Von Raab, 816 F.2d 170, 175-76 (5th Cir.1987), cert. granted, --- U.S. ----, 108 S. Ct. 1072, 99 L. Ed. 2d 232 (1988) (testing may disclose pregnancy and certain psychological and physical illnesses), no allegation has been made that SEPTA's plan contemplates use of body fluid samples to detect such information. We are satisfied that, as described, the plan contains sufficient safeguards, in the form of confidentiality, chain of custody, verification, and random selection procedures, to protect against abuse of discretion by implementing officials. Any improper use of the personal information obtainable through SEPTA's testing program must await a challenge to the program as applied.

As noted above, one of the Unions' principal challenges to the reasonableness of SEPTA's random drug testing policy is based on their contention that the mere presence of drug metabolites in an employee's urine is not indicative of on-the-job impairment. However, as the district court explained, the expert psychologist who designed SEPTA's revised program testified that SEPTA's drug testing was aimed at detecting chronic abusers rather than on-the-job impairment. 678 F. Supp. at 548. The district court was satisfied that a sufficient nexus existed between this rationale for testing and the information gleaned through the use of urine tests. As we concluded above, since the evidence shows that chronic users present a threat to the safety of the transit system, we do not regard it as fatal that the tests are designed to uncover drug users rather than impairment.

It follows that SEPTA has satisfied the second prong of the "reasonableness under all the circumstances" standard, and that its program for random testing for drugs and alcohol satisfies the rigorous inquiry required under that standard.

As we noted above, the district court found that the component of SEPTA's proposed drug testing program which imposed body fluid testing on all employees who returned to work following absences for 30 days for any cause such as illness, on the job injury, leave of absence, reinstatement from discipline, or other absence (other than vacation), failed constitutional muster. The court found such testing was not justified at its inception because the required physical examination would identify employees who were clearly not fit for duty. See 678 F. Supp. at 551. Since those employees were also subject to the extensive random testing program and SEPTA had not shown that return-to-work testing was an "integral part" of a carefully considered program designed to meet a substantial need, it had failed to meet its burden to justify the need for such testing. Id.

Return-to-work testing must be justified either as part of a truly random testing program which does not single out a specified group or on the basis of particularized suspicion. Since the random testing which we are sustaining will include employees returning to work, SEPTA must justify its return-to-work testing on the basis of some particularized suspicion. It has, however, failed to present any evidence that the employees returning to work present some unique risk directly related to drug or alcohol use. Thus, SEPTA has not shown that this aspect of its program is initially justified or that testing of all employees returning after an absence for whatever cause has any relationship to the articulated need for the program. We will therefore not disturb the court's injunction against return-to-work testing.

SEPTA's challenge to the district court's injunction of the testing policy as applied to those employees covered by the RLA must be rejected on the basis of our recent decision in Railway Labor Executives' Association v. Conrail, 845 F.2d 1187 (3d Cir.), cert. granted, --- U.S. ----, 109 S. Ct. 52, 102 L. Ed. 2d 31 (1988). This case does not lose its status as controlling precedent because the Supreme Court has agreed to review the decision.

In Conrail, we held that the unilateral addition of drug testing to employee medical examinations gave rise to a "major" dispute under the RLA, entitling the parties to an injunction maintaining the status quo while they pursued the notice, mediation, and conciliation procedures mandated under sections 5 and 6 of the RLA, 45 U.S.C. Secs. 155, 156. 845 F.2d at 1194.

A labor dispute over implementation of a drug testing program that "can 'arguably' be justified by the existing agreement," or "from habit and custom," is classifiable as a minor one and the proposed change may be implemented unilaterally. Id. at 1190-91 (citations omitted). In Conrail, we rejected the company's argument that no significant departure from past practices was involved in its addition of drug screening, which it had previously administered only "for cause," to urine tests for other substances routinely administered during periodic and return-to-work medical exams. Id. at 1194. We found that there were significant differences between the company's new and old practice, both because the company had switched from testing on the basis of particularized suspicion to testing without cause, and because it had substituted its previous disciplinary rule directed against employees discovered to be drug-impaired on the job with a new rule mandating discharge of employees with positive drug test results regardless of a finding of impairment. Id. at 1193-94.

Here, SEPTA's institution of random drug testing differs significantly from prior practice. For several decades, SEPTA's collective bargaining agreements have provided that an employee would be discharged for reporting to work under the influence of intoxicating liquors or drugs. In late 1983, SEPTA began to require urinalysis testing of workers suspected of violating this rule. When the first four employees with positive results for marijuana metabolites were discharged, arbitrators who heard the resulting grievances reinstated the employees on the ground that the presence of marijuana metabolites in the employees' urine did not establish that they were impaired at the time of the urinalysis test.

To meet this objection, SEPTA announced in September 1985 that henceforth it would consider the presence of traces of certain intoxicants or controlled substances in the body fluids of employees on duty, reporting to duty, or off-duty but subject to duty to be a dischargeable offense per se. Any employee suspected of being in violation of this rule would be subject to drug testing.

In 1985, SEPTA began testing employees for drug and alcohol use on promotion or transfer to certain positions. SEPTA also began administering drug tests to employees committing "chargeable accidents," i.e., accidents with respect to which the employee had been determined to be solely at fault. In July 1986, SEPTA extended this policy to require body fluid testing for any signal violation.

SEPTA argues that the Railway Unions' acquiescence, without requesting section 6 notice, in SEPTA's implementation of post-accident, suspicion-based, and transfer and promotion drug testing constitutes a waiver of their right to object to the "incremental" addition of random and return-to-work drug testing. Reply Brief of Appellees-Cross-Appellants at 11. This argument ignores the significant difference between drug testing predicated on requests for transfer or promotion, which carries an element of voluntariness, and random testing, which does not. See National Treasury Employees Union v. Von Raab, 816 F.2d 170, 177 (5th Cir.1987), cert. granted, --- U.S. ----, 108 S. Ct. 1072, 99 L. Ed. 2d 232 (1988) (employees tested on transfer "only as a result of a process that they choose to set in motion"). Moreover, testing on the basis of reasonable suspicion of drug use bears an element of "cause" not present in the case of random testing. See Division 241, Amalgamated Transit Union v. Suscy, 538 F.2d 1264, 1267 (7th Cir.), cert. denied, 429 U.S. 1029, 97 S. Ct. 653, 50 L. Ed. 2d 632 (1976) ("probable cause" exists for post-accident and suspicion-based testing). Thus, SEPTA's addition of random testing to its previous policies constitutes at least as significant a change in working conditions and rules as did the addition of drug screening to routine medical examinations in Conrail.

SEPTA argues that because the Unions agreed to some types of workplace testing, they waived their right to object to any drug testing policy. We reject such a rule because it would deter unions from acquiescing or compromising on the drug testing question at pain of being found to have agreed to a yet broader program. The district court found "no evidence that the unions agreed to or acquiesced in widespread testing or to any significant testing without individualized suspicion." 678 F. Supp. at 551. This finding is amply supported by the record.

SEPTA also argues that it reserved its right to modify its drug testing policies through a "management functions" clause, contained in all of its collective bargaining agreements, which provides that SEPTA retains "[a]ll management functions and responsibilities which SEPTA has not expressly modified or restricted by a specific provision of this Agreement." App. at 1191, 1204, 1218, 1228, 1287. We agree with the court's ruling in International Brotherhood of Teamsters v. Southwest Airlines Co., 842 F.2d 794, 804 (5th Cir.), reh'g en banc granted, 853 F.2d 283 (5th Cir.1988), that such broad management rights clauses cannot serve to exempt management from the RLA prohibition against unilateral institution of changes in rules and working conditions.

SEPTA's addition of broad random testing to its previous policies thus constitutes a major dispute for the purposes of the RLA, and the district court rightfully enjoined SEPTA from implementing random testing of employees covered by the RLA pending exhaustion of the mediation and conciliation procedures required under that Act.

We are aware that our decision sustaining a random drug and alcohol testing program under the reasonableness standard recently enunciated by the Supreme Court can be viewed as a further erosion of government workers' important Fourth Amendment rights. We believe, however, that implicit in the balancing that underlies application of the reasonableness test is the recognition that there are some public interests that carry sufficient weight to justify even a more than minimal privacy intrusion. Public health and public safety concerns have traditionally been considered to be of the highest magnitude of public interest considerations. The effects of drugs on transit workers in safety sensitive positions as established in this record lead us to conclude that random testing by SEPTA is a reasonable response. We venture no opinion on the validity of such testing in other situations, and emphasize that our decision is confined to the case presently before us.

For the reasons set forth in more detail above, we will uphold the random drug and breathalyzer testing in SEPTA's program. We agree with the district court's conclusion that the return-to-work testing does not satisfy the first prong of the reasonableness test, since it has not been shown to be reasonable at its inception. We also have concluded that under the precedent of our Conrail opinion, random drug and alcohol testing cannot be unilaterally imposed on those workers covered by the RLA because that statute requires different procedures for major disputes.

We will therefore affirm the district court's order in all respects. Each party to bear its own costs.