Defendants Fibreboard Corporation and Pittsburgh Corning Corporation, joined by other defendants, petition for writ of mandamus, asking that we vacate pretrial orders consolidating 3,031 asbestos cases for trial entered by Judge Robert Parker, Eastern District of Texas.
In 1986 there were at least 5,000 asbestos-related cases pending in this circuit. We then observed that “because asbestos-related diseases will continue to manifest themselves for the next fifteen years, ings will continue at a steady rate until the year 2000.” 1 Id. at 470. That observation is proving to be accurate. In Jenkins v. Raymark, we affirmed Judge Parker’s certification of a class of some 900 asbestos claimants, persuaded that the requirements of Rule 23(b)(3) were met for the trial of certain common questions including the “state of the art” defense. After that order and certain settlements, approximately 3,031 asbestos personal injury cases accumulated in the Eastern District of Texas.
The petitions for mandamus attack the district court’s effort to try these cases in a common trial. In summary, and we will explain later in more detail, the district court has set these 3,031 cases for trial commencing February 5, 1990. The trial will proceed in three phases. Phase I is similar to the procedure approved in Ray-mark in which common defenses and punitive damages will be tried. In Phase II, and before the same jury, certain representative cases will be fully tried and the jury will decide the total, or “omnibus” liability to the class. In Phase III, any awarded damages will be distributed utilizing various techniques. Petitioners grumble over Phase I, conceding that it is no more than we have approved in Raymark, and focus their fire upon Phase II. Petitioners also attack limits placed on discovery from class members as well as the intense schedule for their oral depositions.
The standard of review is familiar. We are to issue a writ of mandamus only “to remedy a clear usurpation of power or abuse of discretion” when “no other adequate means of obtaining relief is available.” In re Paradyne Corp., 803 F.2d 604, 612 (11th Cir.1986) (quoting In re Extradition of Ghandtchi, 697 F.2d 1037, 1038 (11th Cir.1983) and United States v. Fernandez-Toledo, 737 F.2d 912, 919 (11th Cir.1984)). As we stated in In re Willy, 831 F.2d 545, 549 (5th Cir.1987):
Mandamus cannot be used as a substitute for appeal even when hardship may result from delay or from an unnecessary trial. Schlagenhauf v. Holder, 379 *708U.S. 104, 110, 85 S.Ct. 234, 238, 13 L.Ed.2d 152 (1964). Mandamus is an extraordinary remedy that should be granted only in the clearest and most compelling cases. Kerr v. United States District Court, 426 U.S. 394, 402-03, 96 S.Ct. 2119, 2123-24, 48 L.Ed.2d 725 (1976); In re Davis, 730 F.2d 176, 181 (5th Cir.1984). A party seeking mandamus must show that no other adequate means exist to attain the requested relief and that his right to the issuance of the writ is ‘clear and indisputable.’ Kerr v. United States, 426 U.S. at 403, 96 S.Ct. at 2124 (quoting Banker’s Life & Cas. Co. v. Holland, 346 U.S. 379, 384, 74 S.Ct. 145, 148, 98 L.Ed. 106 (1953)).
Finally, mandamus relief is ordinarily inappropriate when review is obtainable on direct appeal. After a brief look at the background of these cases, we will return to the question of whether petitioners have met this extraordinary burden.
I
On September 20, 1989, Professor Jack Ratliff of the University of Texas Law School filed his special master’s report in Cimino v. Raymark. The special master concluded that it was “self-evident that the use of one-by-one individual trials is not an option in the asbestos cases.” The master recommended four trial phases: I (class-wide liability, class representatives' cases), II (classwide damages), III (apportionment) and IV (distribution). On October 26, the district court entered the first of the orders now at issue. The district court concluded that the trial of these cases in groups of 10 would take all of the Eastern District’s trial time for the next three years, explaining that it was persuaded that “to apply traditional methodology to these cases is to admit failure of the federal court system to perform one of its vital roles in our society ... an efficient, cost-effective dispute resolution process that is fair to the parties.” The district court then consolidated 3,031 cases under Fed.R.Civ.P. 42(a) “for a single trial on the issues of state of the art and punitive damages and certified a class action under rule 23(b)(3) for the remaining issues of exposure and actual damages.” The consolidation and certification included all pending suits in the Beaumont Division of the Eastern District of Texas filed as of February 1, 1989, by insulation workers and construction workers, survivors of deceased workers, and household members of asbestos workers who were seeking money damages for asbestos-related injury, disease, or death.
Phase I is to be a single consolidated trial proceeding under Rule 42(a). It will decide the state of the art and punitive damages issues. The district court explained that:
“the jury will be asked to decide issues such as (a) which products, if any, were asbestos-containing insulation products capable of producing dust that contained asbestos fibers sufficient to cause harm in its application, use, or removal; (b) which of the Defendants’ products, if any, were defective as marketed and unreasonably dangerous; (c) when each Defendant knew or should have known that insulators or construction workers and their household members were at risk of contracting an asbestos-related injury or disease from the application, use, or removal of asbestos-containing insulation products; and (d) whether each Defendant’s marketing of a defective and unreasonably dangerous product constituted gross negligence. In answering issue (d), the Jury will hear evidence of punitive conduct including any conspiracy among the Defendants to conceal the dangers (if any) of asbestos. The wording of issues (c) and (d) will depend on the applicability of the 1987 Texas Tort Reform legislation to a particular class member’s individual case.
By its order of December 29, 1989, the district court explained that “the jury may be allowed to formulate a multiplier for each defendant for which the jury returns an affirmative finding on the issue of gross negligence.”
The district court also described the proceedings for Phase II in its October 26 order. In Phase II the jury is to decide the percentage of plaintiffs exposed to each defendant’s products, the percentage *709of claims barred by statutes of limitation, adequate warnings, and other affirmative defenses. The jury is to determine actual damages in a lump sum for each disease category for all plaintiffs in the class. Phase II will include a full trial of liability and damages for 11 class representatives and such evidence as the parties wish to offer from 30 illustrative plaintiffs. Defendants will choose 15 and plaintiffs will choose 15 illustrative plaintiffs, for a total of 41 plaintiffs. The jury will hear opinions of experts from plaintiffs and defendants regarding the total damage award. The basis for the jury’s judgment is said to be the 41 cases plus the data supporting the calculation of the experts. regarding total damages suffered by the remaining 2,990 class members.
Class members have answered questionnaires and are testifying in scheduled oral depositions now in progress. Petitioners attack the limits of discovery from the class members, but we will not reach this issue. It is sufficient to explain that defendants are allowed a total of 45 minutes to interrogate each class member in an oral deposition. These depositions will not be directly used at the trial in Phase II. Rather, the oral depositions, with the other discovery from class members, provide information for experts engaged to measure the damages suffered by the class.
II
Defendants find numerous flaws in the procedures set for Phase II of the trial. They argue with considerable force that such a trial would effectively deny defendants' rights to a jury under the seventh amendment, would work an impermissible change in the controlling substantive law of Texas, would deny procedural due process under the fifth amendment of the United States Constitution, and would effectively amend the rules of civil procedure contrary to the strictures of the enabling acts.
Plaintiffs deny that Phase II would deny defendants any right. Plaintiffs argue that every plaintiff is effectively before the court; that the evidence to be offered by experts more use evidence under Rule 1006 of the Federal Rules of Evidence than the use of math models to extrapolate total damages from sample plaintiffs. Plaintiffs concede that the contemplated trial is extraordinary, but argue that extraordinary measures are necessary if these cases are to be tried at all. While extraordinary, the measures are no more than a change in the mode of proof, plaintiffs say. The argument continues that Rule 23 is not the necessary vehicle for the ordered trial, but will sustain it, if the “consolidation” is viewed as a class. We turn to these arguments.
-A-
The contentions that due process would be denied, the purposes of Erie would be frustrated, and the seventh amendment circumvented are variations of a common concern of defendants. Defendants insist that one-to-one adversarial engagement or its proximate, the traditional trial, is secured by the seventh amendment and certainly contemplated by Article III of the Constitution itself. Defendants point out, and plaintiffs quickly concede, that under Phase II there will inevitably be individual class members whose recovery will be greater or lesser than it would have been if tried alone. Indeed, with the focus in Phase II upon the “total picture”, with arrays of data that will attend the statistical presentation, persons who would have had their claims rejected may recover. Plaintiffs say that “such discontinuities” would be reflected in the overall omnibus figure. Stated another way, plaintiffs say that so long as their mode of proof enables the jury to decide the total liability of defendants with reasonable accuracy, the loss of one-to-one engagement infringes no right of defendants. Such unevenness, plaintiffs say, will be visited upon them, not the defendants.
With the procedures described at such a level of abstraction, it is difficult to describe concretely any deprivation of defendants’ rights. Of course, there will be a jury, and each plaintiff will be present in a theoretical, if not practical, sense. Having *710said this, however, we are left with a profound disquiet. First, the assumption of plaintiffs’ argument is that its proof of omnibus damages is in fact achievable; that statistical measures of representativeness and commonality will be sufficient for the jury to make informed judgments concerning damages. We are pointed to our experience in the trial of Title VII cases and securities cases involving use of fraud on the market concepts and mathematical constructs for examples of workable trials of large numbers of claims. We find little comfort in such cases. It is true that there is considerable judicial experience with such techniques, but it is also true we have remained cautious in their use. Indeed, as the district court stated in one massive Title VII case resting on math models:
[I]t has to judicial eyes a surrealistic cast, mirroring the techniques used in its trial. Excursions into the new and sometimes arcane corners of different disciplines is a familiar task of American trial lawyers and its generalist judges. But more is afoot here, and this court is uncomfortable with its implications. This concern has grown with the realization that the esoterics of econometrics and statistics which both parties have required this court to judge have a centripetal dynamic of their own. They push from the outside roles of tools for “judicial” decisions toward the core of decision making itself. Stated more concretely: the precision-like mesh of numbers tends to make fits of social problems when I intuitively doubt such fits. I remain wary of the siren call of the numerical display ... 2
This concern is particularly strong in this ease, where there are such disparities among “class” members.
The plaintiffs’ answers to interrogatories and the depositions already conducted have provided enough information to show that if, as plaintiffs contend, the representative plaintiffs accurately reflect the class, it is a diverse group. The plaintiffs’ “class” consists of persons claiming different diseases, different exposure periods, and different occupations. The depositions of ten tentative class representatives indicate that their diseases break down into three categories: asbestosis (plural and pulmonary) — eight representatives; lung cancer — three representatives; and Mesothelioma — one representative. The class breaks down as follows:
[[Image here]]
In addition, plaintiffs’ admissions of fact show the following disparities among class members.
a. The class includes persons who do not have legal claims against Defendant ACandS, Inc.
b. One or more members of the class may be barred from prosecuting claims against ACandS by virtue of their prior employment with ACandS.
The severity and type of physical or mental injuries varies among class members.
The nature and type of damage varies among class members. d.
Not all of the Plaintiffs have been injured by the acts, omissions, conduct or fault of all of the Defendants. e.
f. The dates of exposure to asbestos-containing products varies among class members.
g. The types of products to which class members were exposed varies among class members.
h. The dates that class members knew or should have known of their exposure to asbestos-containing products is not identical among class members.
We are also uncomfortable with the suggestion that a move from one-on-one “traditional” modes is little more than a move to modernity. Such traditional ways of proceeding reflect far more than habit. They reflect the very culture of the jury trial and the case and controversy requirement of *711Article III. It is suggested that the litigating unit is the class and, hence, we have the adversarial engagement or that all are present in a “consolidated” proceeding. But, this begs the very question of whether these 3,031 claimants are sufficiently situated for class treatment; it equally begs the question of whether they are actually before the court under Fed.R.Civ.Proc. Rules 23 and 42(b) in any more than a fictional sense. Ultimately, these concerns find expression in defendants’ right to due process.
-B-
These concerns are little more than different ways of looking at a core problem. The core problem is that Phase II, while offering an innovative answer to an admitted crisis in the judicial system, is unfortunately beyond the scope of federal judicial authority. It infringes upon the dictates of Erie that we remain faithful to the law of Texas, and upon the separation of powers between the judicial and legislative branches.
Texas has made its policy choices in defining the duty owed by manufacturers and suppliers of products to consumers. These choices are reflected in the requirement that a plaintiff proVe both causation and damage.3 In Texas, it is a “fundamental principle of traditional products liability law ... that the plaintiffs must prove that the defendant supplied the product which caused the injury.”4 These elements focus upon individuals, not groups. The same may be said, and with even greater confidence, of wage losses, pain and suffering, and other elements of compensation. These requirements of proof define the duty of the manufacturers.
Plaintiffs say, of course, that these requirements will be met by the proposed procedures. This proof for 2,990 class members will be supplied by expert opinion regarding their tive plaintiffs. Plaintiffs deny that they will be extrapolating a total universe from a sample. While we are skeptical of this assertion, plaintiffs’ characterization is of little moment. The inescapable fact is that the individual claims of 2,990 persons will not be presented. Rather, the claim of a unit of 2,990 persons will be presented. Given the unevenness of the individual claims, this Phase II process inevitably restates the dimensions of tort liability. Under the proposed procedure, manufacturers and suppliers are exposed to liability not only in 41 cases actually tried with success to the jury, but in 2,990 additional cases whose claims are indexed to those tried.5
Texas has made its policy choices in its substantive tort rules against the backdrop of a trial. Trials can vary greatly in their procedures, such as numbers of jurors, the method of jury instruction, and a large number of other ways. There is a point, however, where cumulative changes in procedure work a change in the very character of a trial. Significantly, changes in “procedure” involving the mode of proof may alter the liability of the defendants in fundamental ways. We do not suggest that procedure becomes substance whenever outcomes are changed. Rather, we suggest that changes in substantive duty can come dressed as a change in procedure. We are persuaded that Phase II would work such a change.
The basic changes in the dynamics of trial caused by the rules of evidence and procedure have been particularly noted with respect to the use of expert testimony.6 A contemplated “trial” of the 2,990 class members without discrete focus can be no more than the testimony of experts regarding their claims, as a group, compared to the claims actually tried to the jury. That procedure cannot focus upon such issues as individual causation, but ultimately must accept general causation as *712sufficient, contrary to Texas law. It is evident that these statistical estimates deal only with general causation, for “population-based probability estimates do not speak to a probability of causation in any one case; the estimate of relative risk is a property of the studied population, not of an individual’s case.” This type of procedure does not allow proof that a particular defendant’s asbestos “really” caused a particular plaintiff’s disease; the only “fact” that can be proved is that in most cases the defendant’s asbestos would have been the cause.7 This is the inevitable consequence of treating discrete claims as fungible claims. Commonality among class members on issues of causation and damages can be achieved only by lifting the description of the claims to a level of generality that tears them from their substantively required moorings to actual causation and discrete injury. Procedures can be devised to implement such generalizations, but not without alteration of substantive principle.
We are told that Phase II is the only realistic way of trying these cases; that the difficulties faced by the courts as well as the rights of the class members to have their cases tried cry powerfully for innovation and judicial creativity. The arguments are compelling, but they are better addressed to the representative branches— Congress and the State Legislature. The Judicial Branch can offer the trial of lawsuits. It has no power or competence to do more. We are persuaded on reflection that the procedures here called for comprise something other than a trial within our authority. It is called a trial, but it is not.
The 2,990 class members cannot be certified for trial as proposed under Rule 23(b)(3), Fed.R.Civ.Pro. Rule 23(b)(3) requires that “the questions of law or fact common to the members of the class predominate over any questions affecting individual members.” There are too many disparities among the various plaintiffs for their common concerns to predominate. The plaintiffs suffer from different diseases, some of which are more likely to have been caused by asbestos than others. The plaintiffs were exposed to asbestos in various manners and to varying degrees. The plaintiffs’ lifestyles differed in material respects. To create the requisite commonality for trial, the discrete components of the class members’ claims and the asbestos manufacturers’ defenses must be submerged. The procedures for Phase II do precisely that, but, as we have explained, do so only by reworking the substantive duty owed by the manufacturers. At the least, the enabling acts prevent that reading.
Finally, it is questionable whether defendants’ right to trial by jury is being faithfully honored, but we need not explore this issue. It is sufficient now to conclude that Phase II cannot go forward without changing Texas law and usurping legislative prerogatives, a step federal courts lack authority to take.
Ill
We admire the work of our colleague, Judge Robert Parker, and are sympathetic with the difficulties he faces. This grant of the petition for writ of mandamus should not be taken as a rebuke of an able judge, but rather as another chapter in an ongoing struggle with the problems presented by the phenomenon of mass torts. The petitions for writ of mandamus are granted. The order for Phase II trial is vacated and the cases are remanded to the district court for further proceedings. We find no impediment to the trial of Phase I should the district court wish to proceed with that trial. We encourage the district court to continue its imaginative and innovative efforts to confront these cases. We also caution that defendants are obligated to cooperate in the common enterprise of obtaining a fair trial.
. See Jenkins v. Raymark Industries, Inc., 782 F.2d 468, 470 (5th Cir.1986).
. Vuyanich v. Republic Nat. Bank of Dallas, 505 F.Supp. 224, 394 (N.D.Tex.1980).
. Bell Helicopter Co. v. Bradshaw, 594 S.W.2d 519, 532 (Tex.Civ.App.-Corpus Christi (1979)).
. Gaulding v. Celotex Corp., 772 S.W.2d 66, 77 (Tex.1989).
. Actually, only eleven cases will be fully tried. The parties may offer any evidence they wish about the thirty illustrative plaintiffs.
. In re Air Crash Disaster at New Orleans, LA, 795 F.2d 1230, 1233-34 (5th Cir.1986).
. Gold, Causation in Toxic Torts: Burdens of Proof, Standards of Persuasion, and Statistical Evidence, 96 Y.L.J 376, 384, 390 (1986).