OPINION
Virginia Hernandez Frias, individually, and as next friend of Nicholas Brian Frias and Thomas Isaac Frias, George Anthony Frias, Jesse Valentin Frias, Steven Simon Frias, Carlos Jesus Frias, Christina Mejia, and Jose Manuel Galvan, as representative of the Estate of Jesus Valentin Frias (collectively, the “Friases”) appeal a no-evidence summary judgment granted in favor of Atlantic Richfield Company, Lyondell Petrochemical Company, and Lyondell-Citgo Refining Company Ltd. We affirm.
*927Background
The Friases sued appellees alleging that Jesus Frias (“Jesus”) died of aplastic anemia caused by exposure to benzene containing products at the refinery (the “refinery”) where he was employed by ap-pellees successively from 1974 to 1994. Appellees filed a motion for summary judgment asserting, among other things, that there was no evidence that: (1) the levels of benzene to which Jesus was allegedly exposed were sufficient to cause his death from aplastic anemia; or (2) exposure to benzene under circumstances similar to those allegedly experienced by Jesus causes aplastic anemia.1 The Frias-es filed a summary judgment response containing affidavits of expert witnesses who concluded that Jesus was exposed to enough benzene at the refinery to cause his aplastic anemia. Appellees then filed a reply to the response contending that the Friases’ expert affidavits were no evidence of causation because they did not contain scientifically reliable and legally sufficient expert evidence. The trial court granted the summary judgment without specifying the ground(s) upon which its decision was based.
Standards of Review
Summary Judgment
A no-evidence motion for summary judgment must be granted if: (1) the moving party asserts that there is no evidence of one or more specified elements of a claim or defense on which the adverse party would have the burden of proof at trial; and (2) the respondent produces no summary judgment evidence raising a genuine issue of material fact on those elements. See Tex.R. Civ. P. 166a(i). In reviewing a no-evidence summary judgment, we review the record in the light most favorable to the nonmovant to determine whether more than a scintilla of evidence was presented on the challenged elements of the nonmovant’s claim. See Wal-Mart Stores, Inc. v. Rodriguez, 92 S.W.3d 502, 506 (Tex.2002). When a trial court’s order granting a no evidence summary judgment does not specify the ground(s) relied upon for its ruling, the summary judgment will be affirmed if any of the theories advanced are meritorious. Dow Chem. Co. v. Francis, 46 S.W.3d 237, 242 (Tex.2001).
Expert Testimony
In order to be admissible into evidence, an expert witness’s testimony must, among other things, be reliable. Exxon Pipeline Co. v. Zwahr, 88 S.W.3d 623, 628-29 (Tex.2002). Expert testimony is unreliable if: (1) it is not grounded in the methods and procedures of science and is thus no more than subjective belief or unsupported speculation; or (2) there is too great an analytical gap between the data upon which the expert relies and the opinion he offers. Id. at 629. The purpose of the reliability determination is not to decide whether the expert’s conclusions are correct, but only whether the analysis used to reach them is reliable. Id.
A party may object to the reliability of expert testimony either before trial or when it is offered. See Guadalupe-Bianco River Auth. v. Kraft, 77 S.W.3d 805, 807 (Tex.2002). Once such an objection is made, the burden is on the proponent of the evidence to establish its reliability. See id. A trial court’s decision whether to admit expert testimony is reviewed for abuse of discretion. Id. If the ground for excluding such evidence is not specified by the trial court, the ruling will *928be upheld if any ground is meritorious. See K-Mart Corp. v. Honeycutt, 24 S.W.3d 357, 360 (Tex.2000). In addition to being a determinant of the admissibility of such evidence, the reliability of expert testimony is also a prerequisite to its legal sufficiency. See Merrell Dow Pharms. v. Havner, 953 S.W.2d 706, 714 (Tex.1997).2
Evidence of Causation
The Friases challenge the summary judgment on the ground that their expert witness testimony was reliable to establish: (1) the range and length of exposure to benzene necessary to cause aplastic anemia; and (2) that Jesus’s exposure to benzene was within that range and length.3
Causation in toxic tort cases requires both general and specific causation. See Havner, 953 S.W.2d at 714. General causation is whether a substance is capable of causing a particular injury or condition, while specific causation is whether a substance actually caused a particular individual’s injury. Id.4 In this regard, an unsupported expert opinion, based only on credentials and subjective opinion, will not suffice to prove causation. Id. at 711-12, 720. Rather, the reliability of the data underlying the opinion must be independently evaluated in determining whether the opinion is reliable. See id. at 713-14.
At a minimum, to be considered reliable scientific evidence of general causation, epidemiological studies must: (1) reflect that the risk of an injury or condition in the exposed population is more than double that in the unexposed or control population; and (2) have a “confidence level” of 95%.5 Id. at 717-24. Moreover, an isolated study finding a statistically significant association between a substance and a disease or condition is not legally sufficient evidence of causation because any conclusion about causation can be reached only after an association is observed in studies among different groups and the association continues to hold when the effects of other variables are taken into account. Id. at 727.
In this case, it is undisputed that, at some level and length of exposure, benzene *929causes aplastic anemia.6 Therefore, the controlling issues are: (1) whether there is scientifically rehable evidence of what that level and period of exposure are; and, if so, (2) whether there is legally sufficient evidence in this case that Jesus was exposed at that level for that period. To provide evidence of these two matters, appellants rely on the affidavits of two expert witnesses: (1) Frank Gardner, a licensed physician, board certified internist, and medical school professor of hematology and oncology; and (2) Vernon E. Rose, an industrial hygienist.7
General Causation
Of the two experts’ affidavits, only that of Gardner addresses the level and period of exposure to benzene that can cause aplastic anemia:
Ironically, because the relationship between benzene exposure and aplastic anemia is so well accepted, little work has been done to specifically delineate the levels of exposure at which aplasia can occur. We do, however, have published literature that informs our view of the exposure picture necessary to cause aplastic anemia.
8. Early American studies suggested that benzene exposures in excess of 350 ppm could cause aplasia. More recent studies support the conclusion that repeated exposures to much lower levels can sufficiently damage the marrow to cause aplastic anemia. In 1987, Yin published a study that investigated over 500,000 Chinese workers exposed to benzene. The study concluded that individuals could develop aplastic anemia after being regularly exposed for k months (118 days) to )) ppm of benzene. The study showed nearly a six-fold increase (5.8) in aplastic anemia at these levels of exposure. In 1989, Paci published a similar study that examined Italian shoe workers exposed to benzene. Unlike the Yin study, Pad’s workers were not exposed to pure benzene but to solvents containing a high percentage of benzene (approximately 70%). The Paci study did not report any quantitative exposure data, but descriptions of work practices are similar to those described by Yin. The study showed more than a 15-fold increase in aplastic anemia. The Paci study support[s] the conclusion that benzene exposure even at lower levels than those reported in the Yin study can cause aplastic anemia.
(emphasis added) (footnote citations omitted).
In that Gardner’s expert opinion is based on studies conducted by others, we must determine whether those underlying studies are scientifically rehable to support his conclusion. See Havner, 953 S.W.2d at 713-14. Although Gardner refers to unspecified early American studies, his affidavit does not indicate whether those studies addressed the time period of exposure, no such studies are part of our record, and we lack any other information with which to determine whether those studies met the more than double relative risk or 95% confidence level prescribed by Hamer. See id. at 717-24.
The Yin study, a copy of which is in our record, concludes that “aplastic anemia could develop in about 118 days in workers exposed to benzene concentrations at a level of 1035.6 mg/m.”8 This *930conclusion was based on an occurrence of four cases of aplastic anemia reported among 211 workers employed in a shoe-making factory during an eight month period. Although Gardner relies on the Paci study (a copy of which is also in our record) to corroborate the findings of the Yin study, the Paci study merely concludes “Our results confirm the very high SMR [standardized mortality ratio] values for ... aplastic anemia among Italian shoe workers and are fully consistent with an association with exposure to benzene in glues.” As Gardner acknowledges, the Paci study does not report any quantitative exposure data.
It is not apparent from the Yin study or Gardner’s affidavit that the relative risk figure calculated in the Yin study had a confidence level of 95%, or how the exposure level and period identified for the narrow group of shoe factory workers surveyed by Yin and Paci could equate to an association that has been observed in studies among different groups such that it would continue to hold when the effects of other variables are taken into account. See Havner, 953 S.W.2d at 727. Without a 95% confidence level and an association observed in workers outside the shoe industry, the Friases’ expert testimony does not satisfy the Havner requirements for scientifically reliable evidence of general causation regarding the level and period of exposure to benzene necessary to cause aplastic anemia. Therefore, we overrule the Friases’ first issue.
Specific Causation
As an alternative ground for our holding, we also address the legal sufficiency of the Friases’ evidence of the level and length of Jesus’s benzene exposure while working at the refinery. Apart from the abstract characterizations regarding conditions at the refinery, such as “frequent contact” with and “regular use” of benzene containing products, “significant vapor hazard,” and “dangerously high” levels of benzene exposure, the only evidence attempting to quantify Jesus’s exposure is the concluding paragraph of Rose’s affidavit:
The available information supports the conclusion that [Jesus], during the course of his work as a No. 2 dockman9 [from 1976 to 1994] ... was consistently exposed to benzene levels in the 10 to 20 ppm range ... and that he had regular exposures above 100 ppm (sampling, disconnecting, and draining lines). Occasional peak exposures of hundreds of ppm and, in some cases, approaching 1000 ppm were experienced during washing of the docks and tools with benzene, benzene spills, product sampling, and unusual weather conditions (e.g., no wind).
(emphases added). Even assuming this to be the most precise and comprehensive conclusion that can be reached based on the limited data available, such indefinite terms as “consistently,” “regular,” and “occasional” leave the frequency (ie., hourly, daily, weekly, monthly, etc.) and duration (minutes, hours) of exposure, at any single exposure level and in total, subject to wide variance and thus largely open to speculation. Therefore, even if the Friases’ evidence of general causation was scientifically reliable to show, for example, that the level and period of exposure to benzene necessary to cause aplastic anemia was as reflected in the Yin study, their evidence *931of specific causation would be legally insufficient to establish that Jesus had actually been exposed to that level for that period. Therefore, we overrule the Friases’ second issue and affirm the judgment of the trial court,