OPINION ON REHEARING
Opinion by
Justice Whitehill
We deny appellees’ motion for rehearing. On the Court’s own motion, we withdraw our opinion dated December 31, 2014, and vacate the judgment of that date. This is now the Court’s opinion.
Appellees’ daughter Daniella was born prematurely. Her treating physicians included appellants Enrique N. Ponte Jr., M.D., and Jorge Fabio Llamas-Soforo, M.D. Despite the doctors’ efforts, Daniella eventually went blind in her right eye and lost most of the vision in her left eye. Appellees, acting on Daniella’s behalf, sued Ponte, Llamas, and their employers. The case was tried to a jury, which made findings in appellees’ favor. The trial judge rendered judgment against appellants based on the jury’s findings, reduced by settlement credits. Ponte, Llamas, and their employers appealed. Because appel-lees adduced no non-conclusory evidence of causation, we reverse and render judgment that appellees take nothing.
I. Background
A. Facts
The evidence adduced at trial supported the following facts. Appellee Marcela Bustamante gave birth to Daniella at Del Sol Medical Center in El Paso on May 19, 2005. The admission note describes Dan-iella as weighing 600 grams at birth and as having a “23 week 1 day gestational age.” She was admitted to Del Sol’s neonatal intensive care unit. Appellant Ponte, a neuroneonatologist and an employee of appellant Pediatrix, was medical director of Del Sol’s neonatal ICU. Ponte was Daniel-la’s attending physician while she was in Del Sol’s neonatal ICU.
Daniella suffered from several medical problems related to her premature birth, including seizures, bleeding and inflammation in the brain, and patent ductus arter-iosus, which is a condition involving the blood vessels near her heart that required [74]*74surgery to repair. Ponte was concerned that Daniella would eventually develop a damaging eye condition called retinopathy of prematurity (ROP), which involves abnormal blood-vessel growth in the retina and can cause scarring, detachment of the retina, and blindness.
Ponte contacted appellant Llamas, an ophthalmologist, and asked him to examine Daniella. Llamas examined Daniella’s eyes on July 4, 2005, and he observed no sign of ROP. Llamas’s note from the examination reflects that there was to be a follow-up examination in four weeks. Llamas examined Daniella again on August 1. During that examination he determined that Daniella had developed ROP, and he recommended surgical treatment of the ROP as soon as possible. He performed the surgery on August 4. That procedure involved using a laser to burn parts of Daniella’s retinas. The surgery sacrifices the patient’s peripheral vision to some extent to conserve his or her “central vision.”
At some point after the August 4 surgery, Daniella’s right retina became detached. As a result, she is blind in her right eye, and eventually the eye may have to be removed. She has some vision in her left eye, but it is significantly impaired. There was evidence that Daniella must wear glasses and must hold symbols a few inches from her left eye in order to see them. There was also evidence that she suffers from other conditions, such as cerebral palsy, and that she is developmentally delayed to an undefined extent.
B. Procedural history and issues on appeal
In 2008, appellees, as next friends of Daniella, sued appellants and Del Sol Medical Center’s owners for negligence and gross negligence that allegedly caused Daniella’s vision impairment. The owners settled with appellees before trial. The remaining claims were tried to a jury in 2011. The trial judge submitted jury questions regarding the negligence of Ponte, Llamas, and Del Sol Medical Center. He did not submit any questions regarding any independent negligence by Pediatrix or by Llamas’s professional association.
The jury found Ponte, Llamas, and Del Sol Medical Center negligent. It apportioned 45% of the responsibility for Daniel-la’s injury to Ponte, 45% to Llamas, and 10% to Del Sol. For damages, the jury found that Daniella would incur future medical expenses of $962,000 after she turned 18 and future attendant care expenses of $988,000 after she turned 18. The jury also found damages for Daniella’s pain and mental anguish, disfigurement, and physical impairment totaling $174,000. Because the verdict was not unanimous, the jury did not answer the question about Ponte’s and Llamas’s gross negligence.
The parties engaged in post-verdict motions practice. The trial judge signed a final judgment, a corrected final judgment, and finally a second corrected final judgment. In the second corrected final judgment, the judge rendered judgment against appellants based on the jury verdict, adjusted to account for the settlement credit. In that judgment, Llamas and his professional association were held jointly and severally liable for about $873,000, and Ponte and Pediatrix were separately held jointly and severally liable for the same amount.
II. Sufficiency of the Proximate Cause Evidence
Appellants argue that the evidence is legally insufficient to support the jury’s findings that Ponte’s and Llamas’s negligence proximately caused any injury to Daniella’s vision. Appellants preserved their legal sufficiency challenge by means [75]*75of a motion for judgment notwithstanding the verdict attacking the legal sufficiency of the evidence of proximate cause.4
A. Standard of review
When an appellant attacks the legal sufficiency of the evidence to support an adverse finding on an issue on which it did not have the burden of proof, it must demonstrate that no evidence supports the finding. If evidence is so weak that it does no more than create a surmise or suspicion of the matter to be proved, the evidence is no more than a scintilla and, in legal effect, is no evidence. The evidence is legally sufficient if it is sufficient to enable reasonable and fair-minded people to reach the verdict under review. In conducting our review, we view the evidence in the light most favorable to the verdict and indulge every reasonable inference that would support it. We must credit evidence favorable to the verdict if a reasonable person could, and we must disregard contrary evidence unless a reasonable person could not.5
B. Applicable law
In a medical-malpractice case, the plaintiff must prove the existence of a legal duty, a breach- of that duty by the defendant, proximate causation, and damages.6 Proximate cause has two sub-elements of foreseeability and cause in fact.7 Foreseeability requires proof that a person of ordinary intelligence should have anticipated the danger created by the negligent act or omission.8 Cause in fact requires proof (1) that the defendant’s negligence was a substantial factor in bringing about the injuries and (2) that absent the negligence — that is, but for the negligent act or omission — the harm would not have occurred.9 The plaintiff must introduce evidence of a “reasonable medical probability” or a “reasonable probability” that her injury was caused by the defendant’s negligence, “meaning simply that it is ‘more likely than not’ that the ultimate harm or condition” resulted from the defendant’s negligence.10 Proof of a mere possibility of causation, however, is insufficient.11
If the causation issue is not one within the experience of a lay person, the plaintiff must ordinarily produce expert testimony of proximate causation.12 The parties did not dispute that expert testimony of proximate causation was required in this case. An expert cannot opine simply that the defendant’s negligence probably caused the injury; the expert must also [76]*76explain how and why the negligence probably caused the injury.13
If no factual basis is offered for the causation opinion, or the factual basis that is offered provides no support, the causation opinion is conclusory, even if it is not objected to.14 Expert testimony that is conclusory, speculative, or based on assumed facts contrary to the evidence is legally insufficient to prove the facts testified to.15
C. Review of the evidence
1. Background facts and standard-of-care evidence
The jury heard evidence of these background facts. Normally, retinal blood vessels develop flat against the back wall of the retina. But in premature babies, the retina and its blood vessels are not fully developed at birth. Such babies often develop ROP, which is characterized by the development of weak, abnormal blood vessels that can actually grow out into the center cavity of the eye. ROP is referred to as “plus disease” if the abnormal blood vessels are thick, tortuous, and located in the back of the eye near the optic nerve. Scientists believe that ROP is caused when the parts of the retina that lack blood vessels produce a growth factor that triggers the development of the abnormal blood vessels. Evidence was presented that ROP eventually goes away by itself, and that not all babies who develop ROP need treatment. In some cases, however, ROP can cause scar tissue that can in turn cause retinal detachment and vision loss. Thus, in serious ROP cases, surgical treatment is appropriate. The current form of treatment is laser surgery, which uses a laser to burn some of the parts of the retina that produce the growth factor. The evidence showed that it was extremely likely that Daniella would eventually develop ROP. Ponte himself testified that 100% of babies born at 23 weeks develop ROP.
There was evidence that in 2001, the American Academy of Pediatrics and the American Academy of Ophthalmology promulgated guidelines recommending that physicians start examining premature babies for ROP at the later of four to six weeks after birth or 31 to 33 weeks gestational age. The 2001 guidelines were not changed until 2006, and Ponte testified that he followed the 2001 guidelines until 2006 (the year after Daniella’s birth and treatment).
Appellees’ expert witness William Good, M.D., a pediatric ophthalmologist, however, testified that a study called the Early Treatment for Retinopathy of Prematurity (ETROP) was conducted before 2003. He opined that publishing of the study’s results in 2003 changed the standard of care very quickly. Good testified that the ETROP study showed that babies should be screened and treated for ROP earlier than was previously believed.
Good identified three standard of care breaches in this case, the first two he ascribed to both Ponte and Llamas and the last he ascribed to Llamas alone. First, [77]*77he testified that the four-week delay between examinations, from July 4 to August 1, violated the standard of care because Daniella should have been examined more than once in the interim. Specifically, Daniella should have been examined on July 11, July 18, and, depending on the results of the July 18 exam, on or before July 25. Good testified that if Daniella had been examined on a proper schedule, in reasonable medical probability Daniella would have received laser treatment on or about July 25.
Second, Good testified that the three-day delay in treatment between August 1 and August 4 violated the standard of care because Daniella’s ROP should have been treated earlier than that.
Third, he testified that the manner in which Llamas performed the laser surgery was inadequate.
Appellees also introduced testimony from expert witness and neonatologist Dale Phelps, M.D. She testified that Ponte breached the standard of care by failing to “work[ ] out a set of written guidelines for his staff to communicate with the ophthalmologist.” She indicated that this criticism related to Daniella’s follow-up examinations for ROP. She also testified that Llamas breached the standard of care by scheduling Daniella’s follow-up eye examination four weeks after the July 4 exam and by failing to describe his observations in his notes from the July 4 exam in a way that could be understood by the staff that he was working with. She testified that, depending on Llamas’s observations on July 4, a follow-up eye examination should have been done somewhere from one to three weeks after the July 4 examination. She did not, however, provide any testimony regarding the surgery itself.
2. Proximate-cause evidence
Appellees do not argue that they proved cause in fact without expert testimony. They instead rely on Good and Phelps for that evidence. Good testified about the effect of delaying treatment after Llamas examined Daniella on August 1: “[Ejvery day that went by [without treatment] put [Daniella] at further risk of suffering an adverse outcome.” He then criticized the manner in which Llamas performed the laser surgery as follows:
Thirdly, as I have looked at the pictures of the treatment administered by Dr. Llamas to Daniella, it looks to me like it was inadequate. There are skip lesions. There are areas where the retinal burns either didn’t take or were not administered. And this also contributed to, in a proximate way, to her loss of vision in the right eye and some detrimental loss of vision in the left eye.
[T]he treatment pattern is not as confluent as it should have been, even where the laser burns have taken. But there are also areas called skip areas. These are areas where there is no treatment at all, and these contribute to an adverse outcome in a manageable ROP.
(Emphases added.) Later, Good addressed causation more globally and at greater length:
Q. What injury or change in Daniella resulted from the failures to examine appropriately, timely, and to treat properly and timely, more likely than not?
A; Well, if you look at the various places where negligence occurred, in an incremental fashion, each of those contributed to the poor visual outcome that Daniella experienced. The delay in screening examinations for four weeks to a probability prevented Dr. Llamas from identifying ROP when it could have been treat[78]*78ed earlier. That would have improved the chance of a good visual outcome for her. The delay in laser treatment for three days also in my opinion incrementally increased the chances of a bad outcome for her. And then, finally, the inadequate laser treatment I think really was the coup de grace here, and it basically placed these eyes at extremely high risk and was causally or proximately responsible for blindness in the right eye and poor vision in the left eye.
Q. All right. Let me tell you that in Texas, the definition of proximate cause is that cause which, in a natural and continuous sequence, produces an event, and without which cause the event would not have occurred. And there may be more than one proximate cause of an event. Can you accept that definition?
A. Yes.
Q. Is Daniella’s blindness in your opinion proximately caused by the negligence that you described for us?
A. Yes, it was.
Q. Can you tell us more likely than not what her vision would be like had these defendants acted properly?
A. More likely than not, she would have what I would call a sighted life. In other words, she would be able to use her vision to function in her environment.
Good also opined that Daniella’s vision would be better than 20/200 if appellants had given her appropriate treatment. Instead, her right eye is blind, and her left eye is roughly 20/1200 or 20/1300.
Phelps, who focused on the four week gap between the July 4 and the August 1 examinations as the principal negligence committed by Ponte and Llamas, provided the following causation testimony:
Q. ... So what more likely than not would have resulted had Dr. Llamas and Dr. Ponte followed the followup examination schedule in the guidelines?
A. He would have seen the ROP as it started up or as it became established and before it became advanced. If it was looking very threatening, they might have moved their schedules — exam schedules closer together to make sure they caught it quickly when it got to the point where it needed surgery.
Q. Okay.
A. That was — that would be my expectation.
Q. Do you think more likely than not Daniella would have functional vision had they done it correctly?
A. Yes.
Shortly thereafter, • appellees’ counsel again asked Phelps, “Would Daniella Bus-tamante more likely than not in your opinion have functional vision had they done it correctly — these doctors done this correctly?” Phelps answered, “Yes.”
a. Ponte
After reviewing both experts’ testimony, we conclude that appellees adduced legally insufficient evidence of cause in fact against Ponte.
(1) Good’s causation testimony
Good did not testify that Ponte’s negligent acts and omissions, individually or collectively, were more likely than not a but-for cause of Daniella’s injuries. Rather, he testified that each instance of negligence by both doctors contributed “in an incremental fashion” to Daniella’s poor visual outcome. Ponte’s first act of negli[79]*79gence was failing to ensure that Daniella received additional eye examinations between July 4 and August 1. As to that negligence, Good testified as follows: “The delay in screening examinations for four weeks to a probability prevented Dr. Llamas from identifying ROP when it could have been treated earlier. That would have improved the chance of a good visual outcome for her.” This testimony is insufficient to support a finding of but-for causation regarding this delay because it does not show that the lack of interim examinations was, more likely than not, a but-for cause of Daniella’s injuries. Unquantified evidence that a different course of action would have created a lower risk of harm is legally insufficient to show but-for causation.16
Good also failed to establish but-for causation as to Ponte’s other act of negligence, permitting a three day delay in the performance of the laser surgery after the August 1 examination. According to Good, this delay “incrementally increased the chances of a bad outcome for” Daniella. Again, his testimony that swifter surgery would have had some unquantified positive effect on Daniella’s chance of a better outcome is legally insufficient to show but-for causation.17
Good gave additional causation testimony after he added to the equation Llamas’s negligence in performing the laser surgery. Good described the inadequate laser surgery as “the coup de grace” and said that the surgery “was causally or proximately responsible for blindness in the right eye and poor vision in the left eye.” Appellees’ counsel then gave Good a definition of proximate cause that included the cause-in-fact requirement, and he asked Good whether all of the negligence he had described, committed by both Ponte and Llamas, proximately caused Daniella’s injuries: “Is Daniella’s blindness in your opinion proximately caused by the negligence that you described for us? ” (Emphasis added.) Good answered, ‘Tes, it was.” Counsel then asked, “Can you tell us more likely than not what her vision would be like had these defendants acted properly?” (Emphasis added.) Good answered, “More likely than not, she would have what I would call a sighted life. In other words, she would be able to use her vision to function in her environment.”
When Good expressed a causation opinion in terms of but-for causation, he predicated his opinion on the combined negligence of both Ponte and Llamas as to all three events — the negligent examination schedule, the three day delay in performing the surgery, and Llamas’s negligent performance of the laser surgery. But Ponte was not involved in performing the laser surgery, so he was not responsible for the consequences of any negligence that necessarily included negligent performance of the laser surgery. As emphasized above, Good declined to opine that the four week delay in examining Daniel-la or the three day delay in performing the surgery, either separately or considered together, was more likely than not a but-for cause of Daniella’s injuries. We conclude that Good’s testimony constitutes no evidence that Ponte’s negligence was a but-for cause of the injuries to Daniella’s eyes.
(2) Phelps’s causation testimony
This leaves the causation testimony of Phelps, who identified Ponte’s negli[80]*80gence as failing to “work[] out a set of written guidelines for his staff to communicate with the ophthalmologist.” She later explained that her Ponte criticisms concerned Daniella’s “follow-up for ROP.” As to causation, Phelps said that had Llamas and Ponte used the follow-up schedule in the guidelines, they might have moved the exams closer together to make sure they caught Daniella’s ROP quickly. And Phelps testified that, more likely than not, Daniella would have functional vision if Ponte and Llamas had “done it correctly.”
We conclude that Phelps’s testimony about the causal connection between Ponte’s negligence and Daniella’s injuries was conclusory and therefore constituted no evidence of but-for causation against Ponte.18 The record contains no factual explanation to substantiate Phelps’s opinion that earlier examinations and earlier laser treatment more likely than not would have prevented Daniella’s injuries.
Phelps testified about the ETROP study results and said that favorable outcomes were more common among babies who received early laser treatment as compared to babies who received laser treatment later. She also testified that bad results occurred after conventionally timed laser treatment about 14% to 16% of the time and that babies in the early treatment group suffered bad results only about 9% of the time. But she did not say that the ETROP study showed that earlier treatment of Daniella probably would have prevented her injuries in her particular case. (In fact, Good specifically denied that the ETROP study could be used to determine whether a delay in treatment actually affected a baby’s vision.)
When the evidence shows that a particular treatment helps some patients and not others, the expert must explain the facts justifying a conclusion that a particular patient probably would have been helped by the treatment.19 In this case, there was no factual substantiation or explanation to support Phelps’s testimony that earlier treatment, more likely than not, would have improved Daniella’s outcome. Her causation testimony was conclusory and constituted no evidence of causation against Ponte.
b. Llamas
Next, we examine the cause in fact evidence as to Llamas. We hold that all of the evidence that Llamas’s negligence caused Daniella’s injuries was con-clusory and therefore is no evidence of proximate cause.
(1) Phelps’s causation testimony
We have already held that Phelps’s causation testimony was conclusory and non-probative as to Ponte. Phelps did not offer any different causation testimony as to Llamas. Indeed, she lumped Ponte and Llamas together when she opined that Daniella would have functional vision if Ponte and Llamas had not been negligent regarding Daniella’s follow-up exam schedule. Accordingly, we conclude that Phelps’s conclusory causation testimony was no evidence of but-for causation against Llamas.
(2) Good’s causation testimony
This leaves the testimony of Good, who identified Llamas’s three negligent acts as failing to examine Daniella between July 4 and August 1, failing to perform the laser [81]*81surgery promptly enough after the August 1 examination, and negligently performing the laser surgery itself. As discussed above, Good did not testify that the first two negligent acts were, more likely than not, a but-for cause of Daniella’s vision injuries. He said only that those acts increased the risk of harm to some unknown and unquantified degree. This is no evidence of but-for causation against Llamas regarding these allegedly negligent acts.20
We next consider Good’s testimony regarding Llamas’s third act of negligence, his negligent performance of the laser surgery. When Good considered the surgery by itself, he did not opine clearly that the surgery alone was probably a but-for cause of Daniella’s injuries. He testified that the inadequate surgery “contributed to, in a proximate way, to her loss of vision in the right eye and some detrimental loss of vision in the left eye.” He also testified that the flaws in the surgery “eontribute[d] to an adverse outcome in a manageable ROP.” But Good did not opine that the inadequate surgery alone was, more likely than not, a but-for cause of Daniella’s vision loss. His opinions that the negligence “contributed to” the vision loss are legally insufficient evidence of proximate cause.21
But Good did say that all three negligent acts by Llamas collectively were, more likely than not, a but-for cause of Daniel-la’s injuries. After testifying that Llamas’s first two negligent acts increased Daniella’s risk, Good testified that Llamas’s negligent performance of the laser surgery “basically placed these eyes at extremely high risk and was causally or proximately responsible for blindness in the right eye and poor vision in the left eye.” (Emphasis added.) Then appellees’ lawyer defined proximate cause for Good, and Good testified that Daniella’s extremely limited vision was proximately caused by “the negligence” that Good had previously described — meaning, in context, all the negligence by both appellants. Good also testified that if appellants had acted properly, “[m]ore likely than not, [Daniel-la] would have ... a sighted life. In other words, she would be able to use her vision to function in her environment.” Good’s testimony, taken at face value, was sufficient to show that Llamas’s negligence, considered collectively, was more likely than not a but-for cause of Daniella’s injuries. But, as discussed below, that does not answer the question of whether the causation evidence was legally sufficient.
We next consider Llamas’s argument that Good’s cause-in-fact testimony against Llamas was conclusory. The question is whether Good explained how and why Llamas’s three acts of negligence caused the injury.22 Good testified that the delays in treatment increased the likelihood of injury to an undefined extent and, when combined with the negligent surgery, became a but-for cause of Daniella’s injuries. But he did not explain how or why the delays in treatment contributed to Daniella’s injuries or worsened her chances of a better outcome. The only basis for Good’s opinion about the effect of the delays in treat[82]*82ment we can find in the testimony is the ETROP study, which he testified showed that earlier treatment of ROP led to better outcomes. But Good did not explain how the ETROP study showed that Daniella’s particular outcome would have been affected by earlier treatment. He also agreed that the ETROP study was not designed “to determine whether a delay in screening or a delay in treatment actually affected a baby’s vision,” and that it would be “an incorrect use of that study” to “claim that it was designed to detect or to look at the effect of delay.” So Good’s testimony about the causal connection between the delays in examination and treatment in this ease was conclusory — it was not supported by facts and explanations as to how and why the delays caused Daniella’s injuries in this case.
This leaves Good’s testimony that Llamas’s negligent performance of the laser surgery was causally connected to Daniel-la’s injuries. Good testified that laser therapy has an overall success rate of over 75%. He criticized Llamas’s surgery because, Good claimed, he could see from photographs that there were areas where Llamas failed to make the laser burns sufficiently close together, and there were other places, which he called “skip lesions” or “skip areas,” where the laser burns “didn’t take” or where Llamas applied no laser treatment at all. Good testified that in a proper laser surgery, the burn areas should be confluent or nearly confluent, meaning no more than one burn width apart. He also said that leaving skip areas can allow the disease to progress further.
On the other hand, Good also testified that there was a significant chance that Daniella would have suffered the same bad outcome even with proper treatment. On cross-examination, Good agreed that 89 of the babies involved in the ETROP study, or about 22%,23 suffered retinal detachment despite receiving early surgical treatment. Good admitted he did not review the ETROP results to see how close together the laser burns were in the, 89 babies who suffered retinal detachments; rather, he assumed that the laser therapy was appropriate. He did not explain why he concluded Daniella would not have been among the 22% of babies who suffered blindness even after receiving proper laser treatment. Good also testified that there are several factors that make a bad outcome more likely even if proper treatment is given, and that Daniella had some of those factors, including “plus disease,” extremely low birth weight, and gestational age (apparently meaning Daniella’s extreme prematurity). He did not explain why or how different performance of the laser surgery, more likely than not, would have overcome Daniella’s preexisting adverse risk factors.
The key legal principle is stated in the Jelinek case: “[W]hen the facts support several possible conclusions, only some of which establish that the defendant’s negligence caused the plaintiffs injury, the expert must explain to the fact finder why those conclusions are superior based on verifiable medical evidence, not simply the expert’s opinion.”24 In this case, the evidence showed that there was a significant chance that even a properly performed surgery would not have changed Daniella’s outcome, but Good testified that Daniella probably would have had a better outcome if Llamas had performed the surgery properly. Thus, it was incumbent on Good or some other expert to explain why his conclusion that Daniella [83]*83would have, within reasonable medical probability, had a better outcome was medically preferable to the inference that Daniella would have suffered the same bad outcome even with a better-performed surgery.
Appellees rely heavily on the following testimony by Good in their attempt to show that Good sufficiently explained the basis for his causation opinion:
[A]s I have looked at the pictures of the treatment administered by Dr. Llamas to Daniella, it looks to me like it was inadequate. There are skip lesions. There are areas where the retinal burns either didn’t take or were not administered. And this also contributed to, in a proximate way, to her loss of vision in the right eye and some detrimental loss of vision in the left eye.
[[Image here]]
When a laser burn takes, it makes a white mark. And where there is no laser treatment or the laser burn does not take, there is no mark.
And so there are — the treatment pattern is not as confluent as it should have been, even where the laser burns have taken. But there are also areas called skip areas. These are areas where there is no treatment at all, and these contribute to an adverse outcome in a manageable ROP.
[[Image here]]
We think the physiology is that the cells that exist outside of where the retina is vascularized, meaning the area where there are no blood vessels, we think that these cells after a while begin to send a signal. The signal is a protein, which is a growth factor. It is called vascular endothelial growth factor. And this signal incites the development of abnormal, more fragile blood vessels to grow.
So literally what happens is, blood vessels begin to grow instead of normally along — flat against the back wall of the retina, they start to grow out into the center cavity of the eye.
Eventually, these blood vessels will go away, but in the process they bring in cells that are precursors to the formation of scar tissue. And so the retina can detach, and it can detach partially or it can detach completely.
And then, when asked whether skip areas “keep the disease progressing instead of stopping,” Good answered, “Well, it can, yes.” In his testimony, Good gave a general explanation of the physiology underlying ROP, and in Phelps’s testimony, she explained that using a laser to burn part of the retina that lacks blood vessels prevents the release of the growth factors that cause ROP. The record thus contains evidence explaining how laser treatment effectively treats ROP when it does effectively treat ROP. But no one explained why laser treatment failed roughly 22% of the time in the ETROP study, and no one explained why proper laser treatment in Daniella’s specific case probably would have resulted in a better outcome, despite Daniella’s specific risk factors that made her prognosis worse than it otherwise would have been.
This case is similar to the Jelinek case, in which a hospital negligently failed to treat its patient with antibiotics as prescribed, and the patient suffered a significant amount of pain thereafter.25 The patient later died, and her survivors sued the hospital for negligence.26 The plaintiffs won a judgment based on a jury verdict, but the supreme court reversed because [84]*84the plaintiffs’ expert causation testimony was conclusory. The plaintiffs’ expert testified that the patient’s pain was probably caused by the failure to give her the prescribed antibiotics, but he acknowledged that the pain could have been caused by two other factors that were present and that the antibiotics would not have treated.27 Although the expert testified that the hospital’s negligence probably caused increased pain and suffering to the patient, he did not explain why this possible cause was more likely than the other possible causes that would not have been affected by the hospital’s negligence.28 The supreme court held that the expert’s causation testimony was conclusory and constituted no evidence of causation.29
Here, like Jelinek, the evidence showed that there was a significant chance that Daniella would have suffered the same injuries even if Llamas had performed the surgery as Good said he should have. The evidence also showed that Daniella had specific risk factors that made a bad outcome in her case more likely. But Good did not explain why performing the surgery differently probably would have prevented Daniella’s injuries in her particular case, despite the possibility of a bad outcome even with proper treatment and despite her particular risk factors that increased the likelihood of a bad outcome. After reviewing the record, we conclude that Good’s causation testimony regarding Llamas’s negligent surgery was conclusory because Good did not explain how and why performing the surgery differently probably would have resulted in a different outcome for Daniella.
(3) Statistical evidence
Appellees’ motion for rehearing does not argue that Good or Phelps offered enough factual and explanatory testimony to make their ultimate causation opinions nonconclusory. Instead they argue that the evidence about the ETROP study results was sufficient evidence of but-for causation, citing Merrell Dow Pharmaceuticals, Inc. v. Havner ex rel. Havner,30 Havner involved the sufficiency of causation proof in the context of toxic substance exposure, not medical malpractice. But to the extent it is relevant to the instant case, it does not aid appellees.
In Havner, the Texas Supreme Court specifically declined to hold “that a single epidemiological test is legally sufficient evidence of causation.”31 The court also held that:
To raise a fact issue on causation and thus to survive legal sufficiency review, a claimant must do more than simply introduce into evidence epidemiological studies that show a substantially elevated risk. A claimant must show that he or she is similar to those in the studies. This would include proof that the injured person was exposed to the same substance, that the exposure or dose levels were comparable to or greater than those in the studies, that the exposure occurred before the onset of injury, and that the timing of the onset of injury was consistent with that experienced by those in the study. See generally Thompson, supra, 71 N.C. L.Rev. at 286-88. Further, if there are other plausible causes of the injury or condi[85]*85tion that could be negated, the plaintiff must offer evidence excluding those causes with reasonable certainty. See generally E.I. du Pont de Nemours & Co. v. Robinson, 923 S.W.2d 549, 559 (Tex.1995) (finding that the failure of the expert to rule out other causes of the damage rendered his opinion little more than speculation); Parker v. Employers Mut. Liab. Ins. Co., 440 S.W.2d 43, 47 (Tex.1969) (holding that a cause becomes “probable” only when “in the absence of other reasonable causal explanations it becomes more likely than not that the injury was a result”).32
We conclude for several reasons that the ETROP study evidence constitutes no evidence of but-for causation of Daniella’s injuries.
First, there is no evidence showing that Daniella’s particular characteristics were substantially similar to the ETROP study group’s characteristics. No summary or report of the ETROP study’s results was admitted into evidence, so we know very little about it other than its concluding results. Accordingly, we have no way to determine how the ETROP results might apply to Daniella’s particular facts, and there is no reasonable basis for concluding that the ETROP study’s results would also apply to her. Good admitted that Daniella had risk factors that made a bad outcome more likely in her case regardless of the course of treatment. Nothing in the record indicates whether the babies involved in the ETROP study suffered from any of the same risk factors that Daniella had. The record does not disclose whether the ETROP babies who experienced bad outcomes despite early treatment were more similar to Daniella than were the ETROP babies who experienced good outcomes. Without evidence showing that Daniella was similar to the babies involved in the ETROP study, the ETROP study cannot support a causation finding.33
. Second, this excerpt from the supreme court’s Jelinek decision represents that opinion’s core principles:
It is not enough for an expert simply to opine that the defendant’s negligence caused the plaintiffs injury. The expert must also, to a reasonable degree of medical probability, explain how and why the negligence caused the injury. We have rejected expert opinions not grounded in a sound evidentiary basis: “[I]f no basis for the opinion is offered, or the basis offered provides no support, the opinion is merely a conclusory statement and cannot be considered probative evidence, regardless of whether there is no objection. ‘[A] claim will not stand or fall on the mere ipse dixit of a credentialed witness.’ ” City of San Antonio v. Pollock, 284 S.W.3d 809, 818 (Tex.2009) (quoting Burrow v. Arce, 997 S.W.2d 229, 235 (Tex.1999)); see also Whirlpool Corp. v. Camacho, 298 S.W.3d 631, 637 (Tex.2009) (“Conclusory or speculative opinion testimony is not relevant evidence because it does not tend to make the existence of material facts more probable or less probable.”). When the only evidence of a vital fact is circumstantial, the expert cannot merely draw possible inferences from the evidence and state that “in medical probability” the injury was caused by the defendant’s negligence. The expert must explain why the inferences drawn are medically preferable to competing inferences that are equally consistent with the known facts. Thus, when the facts support several possible conclusions, only [86]*86some of which establish that the defendant’s negligence caused the plaintiffs injury, the expert must explain to the fact finder why those conclusions are superior based on verifiable medical evidence, not simply the expert’s opinion. See Lenger [v. Physician’s General Hospital, Inc. ], 455 S.W.2d [703] at 707 [ (Tex.1970) ] (“[E]xpert testimony that the event is a possible cause of the condition cannot ordinarily be treated as evidence of reasonable medical probability except when, in the absence of other reasonable causal explanations, it becomes more likely than not that the condition did result from the event.”); Hart v. Van Zandtl 399 S.W.2d [791] at 792 [ (Tex.1965) ] (“The burden of proof is on the plaintiff to show that the injury was negligently caused by the defendant and it is not enough to show the injury together with the expert opinion that it might have occurred from the doctor’s negligence and from other causes not the fault of the doctor. Such evidence has no tendency to show that negligence did cause the injury.”).34
Jelinek thus instructs that, where an expert attempts to explain his or her con-elusory opinion, that explanation must have a sound evidentiary basis or the expert’s opinion is still no evidence.35 In Jelinek, like here, the plaintiffs attempted to prove causation with circumstantial evidence, but they lacked legally sufficient evidence because they did not exclude other potential causes for the patient’s injuries.36
Third, the ETROP study evidence furnishes no causation evidence regarding the theory that Llamas ⅛ negligent performance of the surgery caused Daniella’s blindness. Good testified that Llamas’s surgery performance was negligent because the laser burns were not as confluent as they should have been and because there were skip areas. But he did not testify that the ETROP study surgeries conformed to the standard of care as Good defined it. He acknowledged that he did not look at all of the babies’ eyes that were operated on in the ETROP study to see how those surgeries were performed:
Q. In fairness, in the ETROP, did you ever go back and look at the amount of laser that was done by the doctors in your study to see if there was near confluent treatment in every eye that was treated?
A. Not in every eye, but in many eyes.
Q. All right. Did you go back and look at the 89 eyes that had retinal detachments and see if those doctors had done near confluent laser?
A. Did I go back and do it? No.
Q. And so as you sit here today, are you assuming that those 89 eyes went on to retinal detachment even though they got appropriate laser therapy?
A. Yes, I am.
Good’s testimony does not show that the ETROP study has any probative force regarding the causal connection between the manner of performing laser surgery and the outcome. We can infer from his testimony that he looked at some of the surgical results, and in “many” eyes the laser was applied in a confluent manner, but it is unknown whether the laser was not applied confluently to other eyes in the majority cohort. And he testified that he did not examine the eyes that suffered retinal detachment, so it is possible that those [87]*87eyes were treated the way he testified they should have been treated and still suffered a bad outcome. Thus, the ETROP study has no probative value regarding the manner in which Llamas performed the surgery and the cause of Daniella’s injuries.
Fourth, Good admitted that the ETROP study is not proper causation evidence as to negligent treatment delays — which means that the study is useless as to both acts of negligence alleged against Ponte and two of the three acts of negligence alleged against Llamas.
Finally, the ETROP study is just one study. The Havner court specifically declined to hold that a single epidemiological test is legally sufficient causation evidence.37 Absent additional evidence demonstrating the probative value of the single ETROP study on the facts of this case, we follow the path marked by Havner and conclude the ETROP study is not legally sufficient evidence of causation.
Jelinek and Havner required appellees to provide a factual basis for comparing Daniella and the ETROP study participants. To say that the ETROP study shows that Daniella — with her specific additional risk factors — had about a 78% chance of a sighted life assumes that her situation is sufficiently similar to the study participants. However, the record provides no evidence supporting that assumption. Because the required evidence is missing, appellees did not provide the circumstantial causation evidence that Jeli-nek requires.38
D. Conclusion
Appellees’ causation evidence showed only that a different treatment plan could have resulted in a different outcome for Daniella, which does not suffice to show but-for causation. To the limited extent appellees’ experts purported to establish but-for causation, their testimony lacked factual explanation and support and was thus non-probative. Accordingly, appel-lees adduced legally insufficient proximate cause evidence against appellees Ponte and Llamas. The trial court erred by failing to grant appellants’ motion for a take nothing judgment notwithstanding the verdict.
III. Disposition
For the foregoing reasons, we reverse the judgment of the trial court and render judgment that appellees take nothing from appellants.
Schenck, J., dissenting