OPINION OF THE COURT
Smith, J.Defendants challenge a large medical malpractice judgment against them on various grounds. We reject most of their arguments, but modify the Appellate Division’s order to require a new trial on damages for pain and suffering, because the courts below erred in holding that causation issues could not be litigated at a damages trial.
I
On July 18, 1998 Daniel Oakes suffered a sudden, severe headache that was accompanied by vomiting and sensitivity to light. The headache persisted over the next three weeks. Evidence at trial showed that it resulted from an aneurysm (i.e., a balloon-like bulge) in a blood vessel near the brain; the aneurysm went undetected until it ruptured on August 7, 1998, causing a severe stroke that left Mr. Oakes permanently disabled. During the 20-day interim, Mr. Oakes consulted several doctors, including two who now remain as defendants: his primary care doctor, Rajnikant Patel, and a neurologist, Satish Mongia. He also had a CT scan at Millard Fillmore Suburban Hospital; Fillmore Suburban was owned by Millard Fillmore Hospitals, a corporation to which defendant Kaleida Health is the successor in interest. There was evidence at trial supporting a finding that the CT scan was either misread or not read at all, *641and that if it had been read properly the aneurysm could have been detected and the stroke prevented.
Mr. Oakes and his wife brought this action in 2000.* In 2008, after a trial on liability and damages, a jury found that Dr. Patel, Dr. Mongia and Fillmore Suburban were negligent and that their negligence was a substantial factor in causing the injuries complained of. The jury assigned 5% of the fault to Dr. Patel, 1% to Dr. Mongia, and 75% to Fillmore Suburban. It assigned the remaining 19% of the fault to Dent Neurologic Group, a nonparty with which Fillmore Suburban had an arrangement for the reading of CT scans. The jury also found that Millard Fillmore Hospitals was vicariously liable for Dent’s negligence. The jury awarded plaintiffs damages in various categories totaling approximately $5.1 million. The main issues on this appeal arise from three motions that were made after this verdict was rendered.
First, plaintiffs moved to set aside as inadequate the damages portion of the verdict. As to the most significant categories of damages, on which the jury had awarded approximately $4 million, the trial court granted the motion, ordering a new trial as to these items unless defendants agreed to increase the amounts to a total of $17.4 million. Defendants rejected the additur, and a new trial on damages took place.
Secondly, between the first and second trials, Kaleida (which was sued for Millard Fillmore Hospitals’ liability) moved to amend its answer to assert a defense of release, and to dismiss conditionally, on the ground of release, most of plaintiffs’ claims against Kaleida. The release defense was based on proofs of claim that plaintiffs had submitted in 2003 in liquidation proceedings relating to Kaleida’s liability insurer, which we will explain in more detail below. The trial court denied the motion, holding that plaintiffs would be prejudiced by the belated amendment.
Thirdly, shortly before the new trial on damages, plaintiffs moved in limine to preclude, among other things, “any testimony or evidence contesting . . . causation.” The trial court granted this relief, concluding that the issue of causation had been decided by the first jury, and could not be relitigated.
In 2009, the jury at the second trial awarded damages totaling approximately $16.7 million, in addition to the undisturbed *642$1.1 million portion of the first verdict. The trial court entered judgment accordingly, and the Appellate Division affirmed, with two Justices dissenting in separate opinions (Oakes v Patel, 87 AD3d 816 [4th Dept 2011]).
The Appellate Division majority held, among other things, that Kaleida’s motion to amend was properly denied because the releases it sought to rely on were “null and void” (87 AD3d at 819); and that defendants had failed to preserve their argument that the trial court’s additur was excessive {id. at 819-820). The majority did not discuss the preclusion of defendants’ causation testimony at the second trial. Justice Smith dissented on the ground that the damages awarded by the second jury were excessive (a claim defendants do not pursue in this Court). Justice Peradotto dissented from the majority’s ruling on the additur: she would have found that defendants preserved the issue, by opposing the motion for a new trial on damages and by rejecting the proposed additur, and would have held the additur to be excessive.
The Appellate Division granted defendants leave to appeal (87 AD3d 1414 [4th Dept 2011]). We now modify the Appellate Division’s order to direct a new trial on the issue of damages for Mr. Oakes’s pain and suffering, and otherwise affirm.
II
We agree with the Appellate Division majority that the claimed excessiveness of the trial court’s additur was not properly before it, though we reach that conclusion by a route different from the one the Appellate Division took.
Defendants were not required, in order to preserve their claim that the additur was excessive, to identify a specific amount that they considered reasonable. As Justice Peradotto pointed out, defendants made clear in opposing plaintiffs’ motion for a new trial their view that the first jury’s award was reasonable, and “that an additur in any amount would be inappropriate” (Oakes v Patel, 87 AD3d at 823 [Peradotto, J., dissenting]). The Appellate Division regularly reviews, and sometimes accepts, arguments that an additur or remittitur granted by a trial court is either excessive or inadequate (see e.g. Perlin v King, 36 AD3d 495 [1st Dept 2007]; Rivera v Lincoln Ctr. for Performing Arts, Inc., 16 AD3d 274 [1st Dept 2005]; Carlos v W.H.P 19, 301 AD2d 423 [1st Dept 2003]). In no such case, as far as we are aware, has the appellant’s claim been held unpreserved for failure to specify a more reasonable *643increase or decrease in the damages, and imposing such a requirement would serve little purpose.
But a party that wants to challenge the amount of an additur or remittitur on appeal must do so before a new trial takes place. The chief benefit of the devices known as additur and remittitur is that, when they are accepted, they spare the parties and the court the burden and expense of a second trial. Deferring appellate review until after the second trial destroys that benefit. Such a deferral also gives the party opposing the additur or remittitur an unjustified tactical advantage: if successful on appeal, that party can choose whether to accept the new amount of the additur or remittitur, already knowing what the second jury has awarded.
No case has been cited to us in which an appellate court has modified an additur or remittitur after the increased or decreased judgment has been rejected and a new trial has taken place. The Supreme Court of Florida, in the only case we are aware of that addresses a similar question, was also unable to find any such decision anywhere (Poole v Veterans Auto Sales & Leasing Co., Inc., 668 So 2d 189, 191 [Fla 1996]). And under New York appellate practice, review of an additur or remittitur after final judgment would seem inconsistent with the rule that an order granting a new trial—the only kind of order that can include an additur or remittitur—is not one that “necessarily affects” a final judgment and so is not brought up for review when the final judgment is appealed (see CPLR 5501 [a] [1]; Karger, Powers of the New York Court of Appeals § 9:5 at 312 [3d ed rev]).
We see no unfairness in requiring a party dissatisfied with the size of an additur or remittitur to obtain appellate review before any retrial. If there is not time for such review, and neither the trial court nor the appellate court will grant a stay, the party’s remedy is to reject the proffered stipulation and retry the case. Defendants here pursued that remedy. They are not entitled to another remedy because they are displeased with the result.
ra
Kaleida argues that its motion to amend its answer to assert a defense of release, made between the first and second trials, should have been granted. The other defendants join the argument, because a release of Kaleida would reduce their exposure to damages under General Obligations Law § 15-108. Before considering the argument, however, we must decide whether we *644have the power to do so. The question arises because, as we mentioned above, under CPLR 5501 (a) (1) an appeal from a final judgment brings up for review a nonfinal judgment or order only when the nonfinal decision “necessarily affects the final judgment.” The reviewability of the order denying Kaleida’s motion to amend depends on whether it meets that description.
Our opinions have rarely discussed the meaning of the expression “necessarily affects” in CPLR 5501 (a) (1). (Matter of Aho, 39 NY2d 241, 248 [1976] and Siegmund Strauss, Inc. v East 149th Realty Corp., 20 NY3d 37 [2012] are exceptions.) We have never attempted, and we do not now attempt, a generally applicable definition. Various tests have been proposed, but how to apply them to particular cases is not self-evident, and our decisions in this area may not all be consistent (see generally Karger, Powers of the New York Court of Appeals § 9:5 at 304-314 [3d ed rev 2005]).
The application of the “necessarily affects” rule to orders granting or denying motions to amend pleadings has been particularly vexing. We have at times reviewed such orders, thus implicitly assuming that they necessarily affected the final judgment (see Whalen v Kawasaki Motors Corp., U.S.A., 92 NY2d 288, 293 [1998]; Edenwald Contr. Co. v City of New York, 60 NY2d 957, 959 [1983]). On the other hand, in Best v Yutaka (90 NY2d 833, 834 n [1997]) and in Arnav Indus., Inc. Retirement Trust v Brown, Raysman, Millstein, Felder & Steiner (96 NY2d 300, 303 n 1 [2001]) we dismissed appeals insofar as they were taken from orders ruling on motions to amend pleadings. In each of these cases we said in a footnote, without further elaboration, that the order did not “necessarily affect the final determination. ’ ’
We now conclude that we cannot adhere to the rule that the grant or denial of a motion to amend is always unreviewable on appeal from a final judgment. There will be times, of which this is one, when such a ruling “necessarily affects” the final judgment under any common sense understanding of those words. Here, Kaleida’s motion to amend, if granted, would have added a new defense to the case—one that defendants argue, with at least colorable justification, would have significantly changed the case’s result. We hold that in such cases—when an order granting or denying a motion to amend relates to a proposed new pleading that contains a new cause of *645action or defense—the order necessarily affects the final judgment. Best and Arnav, to the extent they hold otherwise, are overruled.
Having found the order denying the motion to amend reviewable, we must begin our review by explaining the rather complicated background of Kaleida’s asserted release defense. Kaleida was insured, subject to a $2 million self-insured retention, by PHICO Insurance Company. PHICO entered liquidation in Pennsylvania in 2002, while this action was pending. In 2003, plaintiffs, seeking to protect their right to whatever insurance proceeds might become available, filed claims with the liquidator.
A Pennsylvania statute says that, when a third party having a claim against an insured files a claim with the liquidator of an insurer, the filing “shall operate as a release of the insured’s liability to the third party ... in the amount of the applicable policy limit” (40 Pa Cons Stat § 221.40 [a]). The Pennsylvania Supreme Court has held that, under this statute, a party filing such a claim makes a binding election to grant a release to the insured, and cannot later change his mind (Koken v Reliance Ins. Co., 586 Pa 269, 290-295, 893 A2d 70, 82-85 [2006]). The third party must “make a choice between pursuing his claim against the insured and presenting his claim in the liquidation” (586 Pa at 293, 893 A2d at 84). Thus, when plaintiffs here made their claim in liquidation, they effectively released their claims against Kaleida to the extent those claims were covered by insurance—i.e., all their claims in excess of the $2 million retention. Indeed, the claim forms plaintiffs signed said, immediately above their signatures: “The undersigned hereby releases any and all claims which have been or could be made against such PHICO insured ... up to the amount of the applicable policy limits and subject to coverage being accepted by the Liquidator, regardless of whether any compensation is actually paid to the undersigned.”
The Pennsylvania statute also provides, however, that a claimant’s release “shall be null and void if the insurance coverage is avoided by the liquidator” (40 Pa Cons Stat § 221.40 [a])—and here the liquidator has avoided coverage of Kaleida for part of the claims now in suit. Relying on Kaleida’s decision to abandon a third-party claim against Dent, the liquidator has declined to accept coverage to the extent that Kaleida is held vicariously liable for Dent’s conduct. The liquidator has acknowledged liability for the 75% of the fault attributed *646directly to Kaleida by the first jury, but not for the 19% attributed to Dent, and thus to Kaleida vicariously. The impact of this on plaintiffs’ releases of Kaleida is unclear under Pennsylvania law; presumably, it voids the releases at least to the extent of the Dent share of the obligation, but whether it voids them entirely—or, perhaps, gives plaintiffs the option of voiding them entirely—is at least disputable.
The Appellate Division resolved this question of Pennsylvania law against Kaleida, holding that the releases were entirely “null and void,” and that therefore the trial court did not err when it barred Kaleida from relying on them. But unlike the Appellate Division, we do not decide the merits of the release defense. Rather we hold that, even assuming that Kaleida’s release defense had merit, the trial court did not abuse its discretion in denying the motion to amend as untimely.
Kaleida claims that the releases did not come to its counsel’s attention until 2008, after the first jury’s verdict. It is clear, however, that Kaleida could have discovered the existence of the releases much sooner. Indeed, in 2007, before the first trial, Kaleida itself, as PHICO’s insured, submitted a claim to the liquidator—using a form identical to the ones, including release language, that plaintiffs had signed in 2003.
The trial court found, and had a basis for finding, that Kaleida’s unexcused delay in relying on the releases prejudiced plaintiffs. Under the theory that Kaleida advanced for the first time between the two trials, it was released for its own actions, but not for the actions of Dent. Had plaintiffs been aware of that defense earlier, they would surely have made an effort to persuade the jury at the first trial that more fault should be attributed to Dent than to Kaleida itself—and they might have succeeded in getting a more favorable apportionment of fault between Dent and Kaleida than the 19% and 75% the first jury adopted. Instead, during the first trial plaintiffs, uncertain of establishing Kaleida’s vicarious liability, directed significant effort at showing that Kaleida’s predecessor Millard Fillmore Hospitals, rather than Dent, was at fault; they even joined Kaleida in an unsuccessful effort to persuade the court not to submit the issue of Dent’s fault to the jury.
It is also reasonable to infer that plaintiffs’ approach to settlement would have been altered if a defense of release had been in the case before the first trial. A claim that is subject to a release defense obviously has a lower settlement value than one that is not.
*647The decision whether to grant the belated amendment of its answer that Kaleida sought was within the discretion of the trial court (e.g. Whalen, 92 NY2d at 293). That discretion was not abused here.
IV
The trial court erred in prohibiting defendants from litigating issues of causation at the second, damages-only trial.
It is often true, as it is in this case, that causation issues are relevant both to liability and to damages. Thus, in a medical malpractice case, liability cannot be established unless it is shown that the defendant’s malpractice was a substantial factor in causing the plaintiffs injury (Mortensen v Memorial Hosp., 105 AD2d 151, 158 [1st Dept 1984]). But even where liability is established, the plaintiff may recover only those damages proximately caused by the malpractice. More specifically, where a condition existing before the malpractice occurred may have contributed to the plaintiffs injury, the plaintiff is not entitled to recover those damages that the preexisting condition would have caused in the absence of malpractice (see McCahill v New York Transp. Co., 201 NY 221, 224 [1911] [in a wrongful death case, “the probability of later death from existing causes for which a defendant was not responsible would probably be an important element in fixing damages”]; Schneider v Memorial Hosp. for Cancer & Allied Diseases, 100 AD2d 583, 584 [2d Dept 1984] [medical malpractice award excessive because “it failed to take into account the fact that the underlying disease . . . could in the future be expected to be the cause” of much of the damage complained of]; Steinhauser v Hertz Corp., 421 F2d 1169, 1173 [2d Cir 1970] [where an accident precipitated plaintiffs schizophrenia, her “latent psychotic tendencies . . . may have a significant bearing on the amount of damages”]).
Here, Mr. Oakes had a preexisting condition, an aneurysm in a blood vessel near his brain. Defendants’ malpractice did not cause the aneurysm. Defendants should have been allowed to show that, even with appropriate medical care, some of the injuries that Mr. Oakes suffered were inevitable. The trial court was mistaken in thinking that the first jury’s verdict resolved this question. The first jury did decide that the malpractice was a substantial factor in causing Mr. Oakes’s stroke, but defendants were entitled to show that some of the pain and suffering that Mr. Oakes endured was not preventable.
Thus, the trial court should not have granted plaintiffs’ motion in limine to preclude causation proof at the second trial. A *648closer question is whether this error had any practical impact on the case. Defendants did not respond to the trial court’s ruling by making a specific offer of proof. Dr. Mongia had served an “expert disclosure” pursuant to CPLR 3101 (d) (1) (i), saying that an expert witness would testify “with respect to the degree of likelihood, if any, that the Plaintiff DANIEL OAKES’s alleged pain and suffering could have been avoided, the degree to which his alleged future medical costs could have been prevented, if at all, and consequently, the degree to which the Plaintiff LISA OAKES’s alleged losses could have been prevented, if at all.” But the disclosure is phrased in general terms and does not tell us specifically what the expert would have said. While it may be adequate for disclosure purposes, it does not serve the function of an offer of proof, which is to enable an appellate court to understand clearly what evidence a trial court has excluded. If we had no more than the expert disclosure to go on, we would not feel justified in reversing any part of the judgment in this case, despite our disagreement with the in limine ruling.
During the course of the second trial, however, Dr. Mongia did present causation evidence, which the court, relying on its in limine ruling, ordered the jury to disregard. Plaintiffs had proved in support of their pain and suffering claim, through the testimony of Mrs. Oakes and an expert, that Mr. Oakes’s treatment for his stroke had resulted in a wound in his groin that became infected and caused serious difficulty. Mrs. Oakes had testified:
“Q. Did you see that infection?
“A. Yes.
“Q. Can you describe what that looked like and where it was?
“A. Well, when I saw when it was opened up and left open ... it was about the size of a pita pocket, and it was just like raw meat, just raw meat in there, pussy and smelled terrible.
“Q. And how long did that wound remain open?
“A. That remained open for a long time. It started healing when he was at the last hospital that he was at. But the slightest little thing would rip it all open again, and he had a lot of problems with that area.”
*649Responding to this evidence, counsel for Dr. Mongia elicited testimony, while cross-examining plaintiffs’ expert witness, that the groin wound was the result of an angiogram (a means of obtaining images of the inside of blood vessels), and that an angiogram would have been necessary to deal with Mr. Oakes’s aneurysm even if his stroke had not occurred. The trial court ruled this and certain other testimony inadmissible, saying to the jury: “the issues of the responsibility and the tie to the injuries as far as causation have been determined, and we’re not revisiting that again.” The court thus told the jury, in substance, that it could not consider the extent to which plaintiffs’ injuries resulted from the malpractice.
We conclude that this error requires reversal of the second jury’s awards (totaling $9.6 million) for pain and suffering. Plaintiffs argue that the groin injury was “minor” in the context of Mr. Oakes’s other very serious injuries, and perhaps it was; but plaintiffs chose to present detailed and vivid testimony about it. We cannot say that this testimony had no significant effect on the jury’s verdict.
We see no need, however, to disturb the awards of the second jury (totaling $7.12 million) in the other categories of damages that were submitted to it: custodial care and supportive services for Mr. Oakes, Mrs. Oakes’s past and future loss of Mr. Oakes’s services and society, and Mrs. Oakes’s future loss of household services. Defendants have identified no evidence and made no argument which suggests to us that proof on causation could have affected these awards. Plaintiffs’ damages in these categories, as far as we can infer from the record, are attributable simply to the fact that Mr. Oakes had a disabling stroke. Any contention that the stroke would have occurred even in the absence of malpractice is essentially a defense to liability, not damages, and is foreclosed by the verdict of the first jury.
V
There is one remaining argument before us: Kaleida claims that there was insufficient evidence before the first jury to support the finding of fault by Dent. The argument is without merit.
The evidence showed that there was an arrangement between Dent and Millard Fillmore Suburban Hospital, by which any scans of the head taken at Fillmore Suburban were read and interpreted by Dent. There was expert testimony that the CT scan of Mr. Oakes’s head, taken 15 days before his stroke, would, *650if properly interpreted, have disclosed a hemorrhage, and that that finding should have led to the taking of measures that could have stopped the stroke from occurring. It was undisputed that no written report was made of Dent’s reading of the CT scan, and that the scan was orally reported to Mr. Oakes and to doctors treating him as showing no abnormality. From these facts, the inference that the scan was either misread or not read at all, and that Dent was at fault for this error, is compelling.
Accordingly, the order of the Appellate Division should be modified, without costs, in accordance with this opinion, and as so modified, affirmed. The certified question is unnecessary and should not be answered.
Chief Judge Lippman and Judges Graffeo, Read, Pigott and Rivera concur.
Order modified, without costs, in accordance with the opinion herein, and as so modified, affirmed, and certified question not answered upon the ground that it is unnecessary.
Mr. Oakes died while this appeal was pending. We use “plaintiffs” to include Mr. Oakes or Mrs. Oakes as his executrix, as the context requires.