We granted certification, 130 N.J. 399, 614 A.2d 620 (1992), to consider whether the Appellate Division had correctly ordered yet another retrial of this fifteen-year-old action based on errors in the trial court’s jury charge. With but one modification we affirm the judgment below substantially on the basis of the Appellate Division opinion. See Feldman v. Lederle Lab., 257 N.J.Super. 163, 608 A.2d 356 (1992). Our disagreement is limited to so much of that court’s opinion as addressed the admissibility of a notation on a letter — the “Swanzey notation.”
I
Plaintiff’s complaint, originally filed in 1978 against Lederle Laboratories, charged that Lederle’s tetracycline-based drug Declomycin, which had been administered to plaintiff from her *343birth in 1960 until some time in 1965, had permanently discolored her teeth. She contended under a strict-liability theory that the product had been defective as marketed at that time because it had failed to warn that pediatric administrations could cause tooth discoloration. Defendant prevailed at the first trial, and the Appellate Division affirmed in an unpublished opinion. We granted certification, Feldman v. Lederle Lab., 91 N.J. 266, 450 A.2d 579 (1982), and remanded the cause to the Appellate Division for reconsideration in light of Beshada v. Johns-Manville Products Corp., 90 N.J. 191, 447 A.2d 539 (1982). The Appellate Division again affirmed, Feldman v. Lederle Lab., 189 N.J.Super. 424, 460 A.2d 203 (1983).
We again granted certification, 94 N.J. 594, 468 A.2d 230 (1983), reversed the judgment, and remanded for a new trial, holding that when a strict-liability defect consists of an improper warning, reasonableness of the defendant’s conduct is a factor in determining liability. Feldman v. Lederle Lab., 97 N.J. 429, 451, 479 A.2d 374 (1984) (Feldman I). The question in such cases is whether, assuming that the manufacturer knew of the product’s harmful propensities, the manufacturer acted as a reasonably-prudent manufacturer in marketing the product without an adequate warning. Ibid. Thus, once the law imputes knowledge to a manufacturer, strict-liability analysis, like negligence, focuses on the reasonableness of a defendant’s conduct. Ibid.
At the close of the evidence in the second trial in 1985, the trial court denied Lederle’s motion to dismiss the complaint on the basis that the claim had been preempted by federal law. The jury returned a verdict for plaintiff on her strict-liability failure-to-warn claim. The verdict form reflected the following findings of fact:
1. Plaintiff ingested Declomycin between 1960 and 1963.
2. There was not a defect in Declomycin based on a failure to warn of tooth discoloration in 1960 and in 1961.
3. There was a defect in Declomycin based on its failure to warn of tooth discoloration in 1962 and in 1963.
*3444. The failure to warn was a proximate cause of plaintiff’s losses and injuries.
5. The conduct of Lederle was not a deliberate act or omission with knowledge of a high degree of probability of harm to the plaintiff and reckless indifference to the consequences.
The finding of “no defect” before 1962 apparently reflects partial acceptance of the testimony of defendant’s three experts that the “discoloration” side effect associated with tetracycline was unknowable before that date.
Lederle appealed, renewing its contention that plaintiff’s claim based on failure to warn had been preempted by federal laws and regulations governing antibiotic and prescription-drug labeling. Alternatively, Lederle argued that reversal was required because the trial court had erroneously instructed the jury on the applicable law and had made improper evidentiary rulings. \
The Appellate Division reversed and remanded for entry of judgment in favor of Lederle on the preemption issue. Feldman v. Lederle Lab., 234 N.J.Super. 559, 564, 561 A.2d 288 (1989) . In doing so, the court noted:
Our failure to address the alternative issues should not be understood to mean that they have no merit. Indeed, we perceive several errors in the jury instructions and evidentiary rulings which, but for our ruling on preemption, would require a reversal and new trial. [Id. at 564 n. 3, 561 A.2d 288.]
We again granted certification, 122 N.J. 348, 585 A.2d 360 (1990) , and held that under the Supremacy Clause, U.S. Const. art. VI, cl. 2, federal law had not preempted plaintiff’s strict-liability cause of action based on failure to warn: in short, Lederle could have warned about the possibility that Declomycin caused tooth discoloration without violating federal law during the time period in which plaintiff alleged injury. Feldman v. Lederle Lab., 125 N.J. 117, 122, 592 A.2d 1176 (1991), cert, denied, — U.S. —, 112 S.Ct. 3027, 120 L.Ed.2d 898 (1992) (Feldman II). We remanded to the Appellate Division to consider those issues previously raised on defendant’s appeal but not decided.
*345Holding true to its prediction, see 234 N.J.Super. at 564 n. 3, 561 A.2d 288, n. 3, the Appellate Division reversed and remanded for a new trial. 257 N.J.Super. at 178, 608 A.2d 356.
II
The Appellate Division found three bases for reversal in the trial court’s jury charge. We agree with its reasoning in each instance.
At the outset of our analysis, we pause to acknowledge the concerns of our dissenting colleagues who remind us of the dangers of reading isolated portions of a jury charge out of context. See post at 350-352, 625 A.2d at 1071-1072. We are mindful that “[rjeversible error will not be found where the charge, considered as a whole, adequately conveys the law and would not confuse or mislead the jury, even though part, standing alone, might be incorrect.” Latta v. Caulfield, 79 N.J. 128, 135, 398 A.2d 91 (1979) (citing Jurman v. Samuel Braen, Inc., 47 N.J. 586, 592, 222 A.2d 78 (1966); Stackenwalt v. Washburn, 42 N.J. 15, 26-27, 198 A.2d 454 (1964)). We note further that the charge, as a whole, must be analyzed “under the evidence and circumstances of the trial.” Fitzmaurice v. Van Vlaanderen Mach. Co., 110 N.J.Super. 159, 164, 264 A.2d 740 (App.Div.1970), affd, 57 N.J. 447, 273 A.2d 561 (1971).
However, unlike our dissenting colleagues, we further recognize other equally-important principles bearing on the. issue of prejudicial error in a jury charge. First, statements delimiting the duty owed by a party in a strict-liability failure-to-warn case involving complicated issues of causation, proof, and preemption should, in the interests of justice, be unmistakably clear. See Eden v. Conrail, 175 N.J.Super. 263, 278, 418 A.2d 278 (App.Div.1980), modified, 87 N.J. 467, 435 A.2d 556 (1981); Phillips v. Scrimente, 66 N.J.Super. 157, 162, 168 A.2d 809 (App.Div.1961). Second, firmly entrenched in our law is the principle that conflicting or inconsistent statements in a jury charge do not magically counteract each other to produce a *346correct jury instruction; otherwise-reversible error dissipates only if the erroneous portions of the charge are expressly withdrawn by the trial court. See, e.g., Jelinek v. Sotak, 9 N.J. 19, 22-24, 86 A.2d 684 (1952) (finding reversible error when court first instructed that train engineer had duty to stop at crossing and then instructed that he was obliged only to use reasonable care at crossing); Girardin v. New York & Long Branch R.R., 135 N.J.L. 135, 138, 50 A.2d 872 (E. & A.1947); Horton v. Smith, 128 N.J.L. 488, 489, 27 A.2d 193 (E. & A.1942); Price v. Phillips, 90 N.J.Super. 480, 485, 218 A.2d 167 (App.Div.1966); Davidson v. Fornicola, 38 N.J.Super. 365, 371, 118 A.2d 838 (App.Div.1955), certif. denied, 20 N.J. 467, 120 A.2d 275 (1956); Terminal Constr. Corp. v. Bergen County Hackensack River Sanitary Sewer District Auth., 34 N.J.Super. 478, 518-19, 112 A.2d 762 (App.Div.1954), modified, 18 N.J. 294, 113 A.2d 787 (1955); cf. Mohr v. B.F. Goodrich Rubber Co., 147 N.J.Super. 279, 285, 371 A.2d 288 (App.Div.1977) (finding that clearly erroneous portion of strict-liability charge in which “and” was substituted for “or” did not compel reversal because methods of proving strict liability were explained clearly throughout course of trial so as to render charging error minor non-prejudicial inaccuracy), certif. denied, 74 N.J. 281, 377 A.2d 685 (1977).
We turn now to the three points of error in the charge. First, the Appellate Division held that the trial court had improperly charged the jury regarding the relevancy of Lederle’s decision to seek Food and Drug Administration (FDA) approval before issuing new warnings about the possibility of tooth discoloration. The Appellate Division’s holding, that the instruction in question effectively directed a verdict for plaintiff, keys in on the trial court’s statement that “the defendant still owes a duty to warn its users in the exercise of reasonable care.” That part of the charge was erroneous and contributed to the verdict; indeed, it bears directly on the critical liability issue in the case.
*347The trial court’s error lay in telling the jury outright that Lederle had a “duty to warn.” Because the jury heard evidence that Lederle had not incorporated a tooth-discoloration warning until late 1963, it could have reached no conclusion other than that Lederle had breached a legal “duty to warn.” The charge effectively told the jury that as a matter of law, defendant had not exercised reasonable care by attempting only to comply with FDA regulations without issuing its own warning. The language in the charge immediately following that excerpted above, which informs the jury that it may consider Lederle’s correspondence with the FDA during the relevant period on the issue of the reasonableness of the company’s decision not to warn until late 1963, becomes meaningless. The damage had already been done.
The jury must decide whether Lederle breached a duty of reasonable care by failing to warn of a known danger in its product — in other words, whether the product was defective for failure to warn at a given time. We reiterate what we stated in Feldman I: the manufacturer’s duty on learning of Declomycin’s dangerous side effects was to warn “as soon as reasonably feasible,” and once knowledge of the danger is imputed to the manufacturer, the focus is on the reasonableness of defendant’s conduct, just as in an ordinary negligence case. Feldman I, supra, 97 N.J. at 451, 456, 479 A.2d 374. Lederle’s duty to warn may not have arisen until late 1963, if that is when a reasonable manufacturer would have acted. The charge left no room for the permissible finding Lederle did not have a duty to warn above and beyond FDA compliance until 1963, and that its decision not to warn was reasonable.'
Even though we held in Feldman II, supra, 125 N.J. at 147, 592 A.2d 1176, that as a matter of law nothing in the federal statutes and regulations prevented Lederle from placing a warning on Declomycin while awaiting FDA approval in 1962 and 1963, Lederle’s attempt to comply with existing FDA regulations still bears on the reasonableness of its conduct. The fact issue for the jury was whether a reasonable drug *348manufacturer with the same knowledge Lederle had in 1960 to 1963 would have placed a tooth-discoloration warning on Declomycin packaging in addition to corresponding with the FDA. In that light, the FDA regulations were simply “minimum standards” — compliance with them was mandatory, but they did not preclude Lederle from taking additional action. See Feldman II, supra, 125 N.J. at 117, 592 A.2d 1176 (holding that although federal law still applied to Lederle, it did not prevent placing of additional warnings on drug packaging); Feldman I, supra, 97 N.J. at 461, 479 A.2d 374 (citing cases holding that FDA regulations are minimum standards); Duffy v. Bill, 32 N.J. 278, 291-93, 160 A.2d 822 (1960) (holding that once trial court finds sufficient evidence to warrant submission of question, jury must decide fact issue of whether reasonable care requires warnings in addition to what is called for by statute); McQuaid v. Burlington County Memorial Hospital, 212 N.J.Super. 472, 475-76, 515 A.2d 796 (App.Div.1986) (stating that although manufacturer’s compliance with FDA request to remove part of warning label may not have provided absolute defense, that decision was clearly relevant to reasonableness of its conduct); Hatfield v. Sandoz-Wander, Inc., 124 Ill.App.3d 780, 80 Ill.Dec. 122, 126, 464 N.E.2d 1105, 1109 (1984) (holding that evidence of FDA compliance in strict-liability failure-to-warn cases is relevant and not overly prejudicial); Hines v. St. Joseph’s Hospital, 86 N.M. 763, 765, 527 P.2d 1075, 1078, cert, denied sub nom. Hines v. Blood Services, Inc., 87 N.M. 111, 529 P.2d 1232 (1974) (holding that compliance with federal regulations could be used to show that defendant had acted with due care under circumstances); O’Gilvie v. International Playtex, Inc., 821 F.2d 1438, 1442 (10th Cir.1987) (charging that “if you find that defendant * * * met all government regulations and requirements, which are minimum standards, such compliance is not a defense if a reasonable and prudent manufacturer would have taken added precautions”), cert, denied, 486 U.S. 1032, 108 S.Ct. 2014, 100 L.Ed.2d 601 (1988); Restatement (Second) of Torts § 288C (1965).
*349 The second error in the charge was that the trial court misstated defendant’s burden of proof when it instructed the jury that it could find for Lederle if the company had “met its burden of proving that it [had] acted as a reasonably prudent manufacturer in distributing Declomycin between 1960 and 1963 * * *.” The trial court seems to have melded concepts of knowledge and reasonableness. The trial court’s assignment to Lederle of the burden to prove reasonableness, and Feldman’s statement, in her supplemental brief in the Appellate Division, that “unless defendant demonstrated that * * * [it did not have] knowledge of the harmful propensity of Declomycin * * *, it fell short of establishing the reasonableness of its conduct to a legal sufficiency,” evidences their confusion. We have speculated on the possibility that their positions stem from an inaccurate interpretation of our statement in Feldman I, that “generally conduct should be measured by knowledge at the time the manufacturer distributed the product.” 97 N.J. at 452, 479 A.2d 374.
Whatever the genesis of the error, knowledge and reasonableness are clearly separate questions, and a shift of the burden to rebut the imputation of knowledge does not shift the burden to prove reasonableness of conduct. In Feldman I, we shifted the burden to defendant to prove only that it had lacked the actual or constructive knowledge of its product’s defect otherwise imputed to it by law. 97 N.J. at 455-56, 458, 479 A.2d 374. Feldman I directed that once knowledge is imputed to the manufacturer, the analysis then focuses on the reasonableness of defendant’s conduct. Id. at 451, 479 A.2d 374. Feldman II, in reviewing defective-product law, similarly instructed that the issue was whether the manufacturer had known of the danger, and “if so, whether ” it had acted reasonably. 125 N.J. at 144, 592 A.2d 1176 (emphasis added). Clearly the process requires two steps, the second step being functionally equivalent to a negligence analysis in which the burden does not shift to defendant, and the reasonableness of the manufacturer’s conduct given its knowledge is at issue. In an *350ordinary negligence case, the plaintiff bears the burden of showing the unreasonableness of the defendant’s conduct (in other words, the defendant’s breach of a duty owed). Dziedzic v. St. John’s Cleaners & Shirt Launderers, Inc., 53 N.J. 157, 161, 249 A.2d 382 (1969); Long v. Landy, 35 N.J. 44, 54, 171 A.2d 1 (1961).
The third error in the charge lies in the trial court’s failure to instruct the jury properly on the issue of damage apportionment. Not only should the court have tailored its general causation instruction more closely to the facts of the case and the proofs adduced, but it should have specifically instructed the jury to apportion damages caused by the administrations of the product only in those years in which it found the product to have been defective. See Scafidi v. Seiler, 119 N.J. 93, 102, 574 A.2d 398 (1990) (stating that-in such contexts standard by which jury evaluates causation must be expressed in terms consistent with operative facts); Campos v. Firestone Tire & Rubber Co., 98 N.J. 198, 210, 485 A.2d 305 (1984) (stating that strict-liability charge should be tailored to factual situation to assist jury in performing its fact-finding responsibility). Owing to the complexity of this case, the standard proximate-cause charge given could have easily confused or misled the jury into awarding damages for all of plaintiff’s injuries if it found Declomycin defective in any one year.
Our decision in Ostrowski v. Azzara, 111 N.J. 429, 545 A.2d 148 (1988), presents an analogy. In Ostrowski, a medical-malpractice case, the jury had to consider the plaintiff's pretreatment health condition and post-treatment conduct in assessing whether the defendant’s negligent medical treatment had proximately caused the plaintiff’s injuries. Id. at 448, 545 A.2d 148. The trial court’s interrogatories asked only whether the plaintiff’s failure “ ‘to exercise that [reasonable] degree of care for her own safety and well-being’ was ‘a proximate cause of her injuries and damages.’ ” Ibid, (quoting interrogatories). We found that interrogatory improper because, among other *351things, it “did not adequately separate or define the concepts that were relevant to the disposition of the plaintiffs case”; “did not specifically cordon off the jury’s consideration * * * to either the pre-treatment period or the post-treatment period”; and “did not distinguish between the patient’s pre-operative and post-operative conduct.” Id. at 448-49, 545 A.2d 148.
Similarly, in this case the trial court’s general instruction on proximate cause and jury interrogatory number three (which asked simply whether the failure to warn had been a proximate cause of plaintiff’s injuries and losses) did not allow the jury an opportunity to find that all or part of plaintiff’s damages had been caused by pre-1962 and post-1963 administrations of Declomyein. Notably, the trial court in Ostrowski made a greater effort to tailor its instructions to the facts of the case than did the trial court in the ease before us, despite which we found error. See ibid.
The dissent argues that-the jury award reflects a rational conclusion, supported by the evidence, that Lederle’s failure to warn in 1962 and 1963 proximately caused $300,000 worth of injuries to plaintiff’s teeth. See post at 373-376, 625 A.2d at 1083-1085. However, that most substantial award is more reasonably seen as reflecting the jury’s confusion on how properly to apportion damages for only those injuries caused by defendant’s failure to warn. Our interpretation is informed by those portions of the record suggesting that most of plaintiff’s damages were caused early in the period in which liability is alleged and for which the most accurate dosage and administration data are available.
The trial court should have, specifically instructed the jury to apportion damages for only those harms caused by administrations of Declomyein during which it found the product defective for a failure to warn. Although retrial will be limited to-liability for 1962 and 1963, the jury must be instructed to apportion damages for pre-1962 and post-1963 administrations of Declomyein if it finds that those caused plaintiff’s damages. *352Because the point of administration of the drug within each year assumes significance, the trial court might consider breaking each year down into smaller units of time, especially within 1962 and 1963.
In sum, we conclude that each point of error, standing alone, could potentially warrant retrial. When considered in combination, those errors had the cumulative effect of rendering the charge capable of confusing, misleading, and otherwise prejudicing the jury. See Borowicz v. Hood, 87 N.J.Super. 418, 424, 209 A.2d 655 (App.Div.) (finding that “cumulative effect” of multiple errors in charge mandated reversal and retrial even though some of those errors did not constitute plain error), certif. denied, 45 N.J. 298, 212 A.2d 170 (1965).
Ill
The Appellate Division further commented on other deficiencies in the charge, which, although not warranting reversal, should nevertheless be corrected at retrial. 257 N.J.Super. at 174, 608 A.2d 364. The cohrt below ruled that on retrial Lederle would be permitted to call its expert witness, Dr. Goddard, who had not been allowed to testify as an expert or as a fact witness because his report had not been furnished to plaintiff on time and because he had lacked personal knowledge of the facts relevant to the time periods in question. Id. at 176-77, 608 A.2d 364. The court also limited retrial to Lederle’s liability in 1962 and 1963 only, and for compensatory damages during those years. Id. at 177, 608 A.2d 364. We approve of the court’s instructions and findings on those issues.
However, we part company with the court below in its decision to classify the notation made by Dr. Swanzey, then director of Regulatory Agencies Relations at Lederle, on a letter from Dr. Barzilai, Medical Officer in the Division of Antibiotic Drugs at the FDA’s Bureau of Medicine, as admissible hearsay under Evidence Rule 63(13), the “business entries” exception. See ibid.
*353Dr. Swanzey testified that he had received a letter dated December 3, 1962, from Dr. Barzilai. At about the time the letter was sent, the two men had a phone conversation. Immediately after he received Barzilai’s letter, Swanzey noted the substance of that phone conversation on the bottom of the letter in his own handwriting, followed .by his initials. That notation reads:
Telephone conversation with Barzilai, December 4. He advised against putting any statement in circular. He is studying this matter and will call a meeting after the holidays with interested parties. E.H.S.
Plaintiff did not object to the typed portions of the letter, but twice objected to the handwritten portions as inadmissible hearsay. The trial court ruled that
the notations on correspondence that result from a telephone conversation do not in any way constitute business records * * *. In fact, * * * that's certainly an out of court statement that’s being admitted for the truth and I am satisfied that it does not fit within the category of exceptions to the hearsay rule * * *.
The Appellate Division reversed, stating that
[t]he evidence was offered to show why Lederle did not issue a warning at that time. Dr. Swanzey was a witness in the case and subject to cross-examination. In our view, he established the requisites of admissibility under Evidence Rule 63(13). While we are not satisfied that this error, standing alone, was capable of affecting the result, we hold that on retrial the notation should be admitted with the letter on which it was written. [257 N.J.Super. at 177, 608 A.2d 364 (citation omitted).]
We agree that the exclusion of the notation did not affect the result in this case. The letter on which Swanzey penned his note communicated a similar message and was admitted into evidence. In that letter Dr. Barzilai acknowledged receipt of Dr. Swanzey’s correspondence of November 16,1962, proposing a new warning statement for all oral and parenteral tetracycline products concerning reports that those products might cause tooth discoloration. He then informed Dr. Swanzey that the FDA was in the process of investigating the matter, and that Dr. Swanzey would be notified when a conclusion had been reached.
However, Dr. Swanzey’s notation is not admissible as a business entry under Evidence Rule 63(13). Under Evidence *354Rule 63, hearsay is “[ejvidence of a statement offered to prove the truth of the matter stated which is made other than by a witness while testifying at the hearing * * All hearsay is inadmissible unless it falls within a recognized exception under Evidence Rules 63(1) to 63(33).
The requisites for admissibility under that rule are:
First, the writing must be made in the regular course of business. Second, it must be prepared within a short time of the act, condition or event being described. Finally, the source of the information and the method and circumstances of the preparation of the writing must justify allowing it into evidence. [State v. Matulewicz, 101 N.J. 27, 29, 499 A.2d 1363 (1985).]
The Appellate Division incorrectly concluded that Swanzey’s testimony had met those requirements. After having Swanzey read his notation to the trial court, Lederle’s attorney examined him as follows:
Q. And did you make that notation at the time of telephone conversation or immediately thereafter?
A. I made it immediately thereafter upon receipt of this letter which crossed in the mail. When I received the letter I put the notation on it immediately. MR. MELHUISH: When you made that notation, sir, was it a notation you entered as to facts that you knew and understood to be true when you wrote them and in the ordinary course of business in your responsibility as you have described for the Lederle Laboratories?
THE WITNESS: Yes, sir.
Moreover, Lederle argued that as a business record, the notation was just as admissible under Evidence Rule 63(13) as the laboratory-research reports that had been received in evidence.
However, unlike lab reports, the Swanzey notation lacks sufficient indicia of trustworthiness and reliability normally found in business records. Even records that are properly kept and have the required circumstantial probability of trustworthiness are not admissible if “ ‘the trial court, after examining them * * * and hearing the manner of their preparation explained, entertains serious doubt as to whether they are dependable or worthy of confidence.’ ” Matulewicz, supra, 101 N.J. at 30, 499 A.2d 1363 (quoting Mahoney v. Minsky, 39 N.J. 208, 218, 188 A.2d 161 (1963)). Such doubts clearly exist where, as *355here, the alleged “business entry” is an informal, handwritten note on the bottom of a piece of correspondence.
Other than Swanzey’s statement that he had made the notation in the ordinary course of business, the record contains no evidence that he had regularly kept, or had been under a duty to keep, written memorializations of his phone conversations with persons on letters sent by those persons, or that he had ever engaged in a regular practice o.f memorializing his phone conversations at the time he made the notation on Barzilai’s letter. Swanzey made no similar notations on any of the other correspondence either sent or received by him included in the appendices to the Appellate Division brief. Comment 2 to Evidence Rule 63(13) instructs that “[pjroof that entries comparable to the one at issue were usually made at or about the time of the event recorded should be sufficient.” No such proof was adduced, nor was any necessity to keep such records demonstrated. Phoenix Assocs. v. Edgewater Park Sewer Auth., 178 N.J.Super. 109, 116, 428 A.2d 508 (App.Div.1981), affd, 89 N.J. 2, 444 A.2d 51 (1982). Thus, because no independent check exists on the veracity of the statements made in the notation other than Swanzey’s word, the notation is not admissible as a business entry under Evidence Rule 63(13).
IV
The judgment of the Appellate Division is modified and, as modified, affirmed.