D.W. NELSON, Circuit Judge:

John Doe and John Smith filed separate suits against the manufacturers of a blood clotting factor and the United States. The district court granted summary judgment in favor of all defendants in both cases. Because the two cases raise similar issues, we have consolidated them for the purposes of this appeal.

OVERVIEW

John Doe and John Smith are hemophiliacs who have tested HIV positive. They received a clotting agent known as Factor VIII from Tripler Army Medical Center (TAMC). Factor VIII enables the blood of hemophiliacs to clot. In the United States, Factor VIII is manufactured and sold by the four appellees in these cases, Alpha Therapeutic Corporation (Alpha), Cutter Biological (Cutter), Armour Pharmaceutical Company (Armour) and Baxter Hyland Healthcare Corporation (Baxter).

Appellants claim that they were infected with the AIDS virus from Factor VIII sometime during 1983. Accordingly, they sued the four manufacturers of Factor VIII for negligence and strict liability. Because the appellants were uncertain as to which manufacturer provided TAMC with the infected clotting agent, appellants brought suit against all four manufacturers of the agent. They also sued the United States for negligence and for breach of its duty to warn appellants while they were treated at TAMC. Appellants originally filed suit in Hawaii state court; their cases were subsequently removed to federal district court.

On January 12, 1989, the district court granted summary judgment in favor of the defendants in the Doe action. On May 18, 1989, the court granted summary judgment in favor of the defendants in the Smith action. The district court held that because appellants could not identify exactly which manufacturers' product had caused their infection, under Hawaii law they could not bring a negligence suit. In addition, the court found that appellees were entitled to summary judgment for several other reasons. These reasons included the court's findings that the appellants could not prove the date of their infections and that because of the limited knowledge about AIDS at the time the infections occurred, appellees were not negligent as a matter of law. The court also found that the United States was not liable because "until 1984 there was no medical consensus that AIDS was transmitted through blood." Therefore, the court found that the appellants' treatment at TAMC had not fallen below the proper standard of care. Finally, the court held that appellants could not bring either a strict liability or negligence suit under Haw.Rev.Stat. § 327-51, Hawaii's Blood Shield Law.

Smith and Doe filed timely appeals. In addition, both appellants filed motions requesting that the Ninth Circuit certify two questions of law to the Hawaii Supreme Court pursuant to Haw.R.App. P. 13. Because the cases presented important state law issues of first impression, we granted Smith's motion. On August 24, 1990, we filed an order certifying the following questions to the Hawaii Supreme Court:

(1) Does Hawaii's Blood Shield Law, Haw.Rev.Stat. § 327-51 preclude Smith from bringing a strict liability claim?

(2) Does Hawaii's Blood Shield Law, Haw.Rev.Stat. § 327-51, preclude Smith from bringing a negligence claim?

(3) Would Hawaii allow recovery in this case when the identity of the actual tortfeasor cannot be proven? If Hawaii would allow recovery, what theory (i.e., burden-shifting, enterprise liability, market share or other) would the Hawaii Supreme Court adopt?

See Smith v. Cutter Biological, Inc., 911 F.2d 374 (9th Cir. 1990). The Hawaii Supreme Court accepted our request, and on November 29, 1991, filed an opinion answering these questions. See Smith v. Cutter Biological, Inc., 72 Haw. 416, 823 P.2d 717 (1991) ( "Cutter Biological"). With the benefit of that Court's responses, we now conclude that the district court erred in granting summary judgment. Accordingly, we reverse and remand this case for proceedings consistent with this opinion and that of the Hawaii Supreme Court.

DISCUSSION [9] I. ISSUES RELATING TO GRANT OF SUMMARY JUDGMENT

A. Legal Theories

The district court concluded that the appellants' inability to identify precisely which manufacturer's product caused them harm prevented them from recovering against any manufacturer. In its response to our certified questions, the Hawaii Supreme Court has made clear that this was an erroneous legal conclusion, and has delineated the legal theories available to appellants.

1. Recovery When Identity of Actual Tortfeasor Not Known

States have developed essentially three types of approaches that permit plaintiffs to bring lawsuits when they do not know the identity of the actual tortfeasor. First, under the theory of alternative liability, if several defendants act negligently and it is not possible to determine which defendant caused plaintiff's injury, the burden shifts to the defendants to prove that they did not cause the injury. This approach was developed by the California Supreme Court in the classic case of Summers v. Tice, 33 Cal.2d 80, 199 P.2d 1 (1948). Under the second approach, the enterprise liability theory, if the plaintiff can prove that an entire industry was negligent, the burden shifts to the members of the industry to prove that they did not supply the specific product that caused the injury. Hall v. E.I. Du Pont, 345 F. Supp. 353 (E.D.N.Y. 1972). Finally, in Sindell v. Abbot Lab., 26 Cal.3d 588, 163 Cal.Rptr. 132, 607 P.2d 924 (1980), cert. denied sub nom E.R. Squibb Sons, Inc. v. Sindell, 449 U.S. 912, 101 S.Ct. 285, 66 L.Ed.2d 140 (1980), the California Supreme Court developed the theory of market share liability. Under this third approach, when it is impossible for a plaintiff alleging injury to prove which of the numerous manufacturers produced the offending product, each manufacturer is responsible for a percentage of the plaintiff's recovery corresponding to its share of the market for the drug.

The Hawaii Supreme Court has now endorsed, with some modifications, market share liability for this case. Cutter Biological, 823 P.2d at 728. The Court further indicated that for the purposes of this case, the relevant market should be the national market. Therefore, all four Factor VIII manufacturers, suppliers of the clotting agent to the national market, are properly defendants to the suit. Limitations for liability are as follows: a given defendant will never be liable for more than its share of the national market, and exculpatory allowances are ordinarily available for defendants that can show that they had no product on the national market at the time of a plaintiff's injury. Id. at 728-29. In the words of that court:

As a result of our determination that a national market is appropriate, as long as defendant is actually one of the producers of Factor VIII, there is little to justify exculpation of defendant. However, the exception would occur where defendant would prove that it had no product on the market at the time of the injury. As far as the defendants in this suit are concerned, it appears that none of them would be able to escape liability on that basis.

Id. at 729 (footnote omitted). On remand, the district court shall permit appellants to proceed with their tort suit against the manufacturer defendants under the theory of market share liability adopted by the Hawaii Supreme Court and against the United States for negligence.

2. Hawaii Blood Shield Law

The Hawaii Blood Shield Law protects the donors and preparers of "blood or component[s] thereof" from liability except for their "own negligence or willful misconduct." Haw.Rev.Stat. § 327-51. The Hawaii Supreme Court has interpreted this statute and held that it bars suits based on strict liability, but permits suits based on negligence. Cutter Biological, 823 P.2d at 722-23. Therefore, on remand, appellants may sue under this statute so long as their suit alleges negligence on the part of the appellees.

3. Concert of Action

Although we did not certify to the Hawaii Supreme Court the question whether a concert of action theory might be available to appellants, that Court was free to consider that question. See Toner v. Lederle Lab., 779 F.2d 1429, 1433 (9th Cir. 1986). In adopting a theory of market share liability, the Hawaii Supreme Court indicated that it did not wish to impose joint liability on the defendants. Consistent with that goal, the Court explicitly held that it would not permit a concert of action theory to be advanced in this case. Cutter Biological, 823 P.2d at 726. Therefore, we uphold the grant of summary judgment as to the concert of action claim.

B. Other Grounds for Summary Judgment

In addition to finding that appellants could not proceed as a matter of law, the district court made additional findings that it held supported a grant of summary judgment for appellees. These findings relate in main to whether appellants could tie appellees to their injuries, to the extent of knowledge about AIDS transmission at the time appellants were allegedly infected, and to whether appellees were negligent as a matter of law. On appeal, appellees argue that the district court's findings and grant of summary judgment should be affirmed. For their part, appellants challenge the court's findings, arguing that in each instance there is a genuine issue of material fact that should preclude a grant of summary judgment. We agree with the appellants. We discuss each issue in turn.

1. Date of Seroconversion

Appellants contend that Doe and Smith cannot prove with certainty the dates of their seroconversions and that because they may have seroconverted before AIDS was identified as being transmissible through blood, summary judgment was properly granted. Appellees and appellants agree that the exact dates appellants acquired the HIV virus cannot be determined.

Doe claims to have seroconverted between mid-1983 and 1984. He supports this claim with the affidavit of Dr. Barbara Weiser, a virologist specializing in HIV. Weiser testified that "based upon her examination of Plaintiff's medical records, her knowledge of the history of the AIDS epidemic in the world, and T cell abnormalities in seropositive persons; that the plaintiff most likely became infected with HIV as the result of exposure to the virus in Factor 8 during the period from mid-1983 to 1984." Weiser then described the change in Doe's T-cell levels and how those changes reflect progression of the disease.

Appellees claim that Doe was infected in December 1980. They support this claim with a letter by one of Doe's experts, Dr. William O'Connor, positing that a rash Doe was treated for at that time was probably a symptom of HIV infection. In an affidavit signed by O'Connor on a later date, he changed his mind and stated that his earlier conclusion that Doe had been infected in 1980 had been tentative, and that he had been unaware that Doe had acquired the rash after working in his yard. Weiser reviewed the same medical records discussing the 1980 rash and concluded that it was unlikely that the rash reflected HIV infection. Appellees' expert, Dr. Peter Levine, a hematologist, shared O'Connor's initial belief that the rash reflected HIV infection.

Smith's claim is complicated by the fact that his medical records have been lost, through no fault of his own. Therefore, his medical history is more speculative. Dr. Harold Burger, a virologist, reviewed the changes in Smith's T-cell levels and concluded that "it is likely that he was infected after mid-1983." Appellees in Smith's case do not point to another date on which they believe Smith was infected. They argue simply that because it is impossible to determine when he seroconverted, summary judgment is appropriate.

What is clear from all of this is that there is a factual dispute regarding the dates of Doe and Smith's seroconversions. Whether at trial appellees can demonstrate that they seroconverted during the period between the point at which the manufacturers could be considered negligent in not testing Factor VIII for AIDS and the time the manufacturers began routinely screening their product remains to be seen. However, we see no basis for appellees' arguments that the conflicting testimony regarding the dates in question somehow supports a grant of summary judgment in appellees' favor. To the contrary, the existence of that conflict only underscores the inappropriateness of a grant of summary judgment as to this issue.

2. Factual Determination Regarding the Medical Community's Knowledge About AIDS

In granting summary judgment for appellees, the district court in Doe made the factual finding that:

The medical community did not reach a consensus that AIDS was blood-borne until 1984. See Judge Flannery's detailed and well-reasoned discussion of the medical chronology of AIDS in Kozup v. Georgetown University, 663 F. Supp. 1048, 1051-1053 (D.D.C. 1987), aff'd in relevant part, 851 F.2d 437 (D.C.Cir. 1988), and the medical literature cited therein. See also, McKee v. Miles Laboratory, 675 F. Supp. 1060 (E.D.Ky. 1987).

The court reached the same conclusion in Smith, citing these and some additional cases. Apparently, the court relied on this finding to support its conclusions that the various appellees were not negligent in their handling of Factor VIII, as discussed below. Our review of the record indicates that in reaching this critical conclusion, the district court did not rely on any of the evidence submitted in the Doe and Smith cases. Rather, the court appears to have based this factual determination solely on the findings of these other courts.

Appellants argue that the district court erred in relying on the findings of fact in Kozup. We agree. The idea that courts can rely on the opinions of other courts for legal principles but not for findings of fact is axiomatic. Doe and Smith had no opportunity to challenge the factual findings of the Kozup court — for example, they could not cross examine the witnesses or counter the evidence produced in documents in that case. We think it clear, therefore, that Doe and Smith cannot be bound by the factfinding in that case.

In Kozup, the parents of an infant who contracted AIDS through a blood transfusion sued the hospital at which he had received the transfusion as well as the American Red Cross. The court found that the hospital and the Red Cross were not liable because they were not aware of the risk and because there were no steps either party could reasonably have been expected to take to prevent transmission of the virus. Kozup, 663 F. Supp. at 1061. Subsequently, in McKee v. Miles Laboratory, 675 F. Supp. 1060 (E.D.Ky. 1987), aff'd, 866 F.2d 219 (6th Cir. 1989), a Kentucky district court considered whether the manufacturers of Factor VIII could be liable under circumstances similar to those presented in the instant cases. In concluding that they could not, the McKee court relied in main on its finding that there was no consensus in the medical community that AIDS was blood borne until 1984. For that finding, the McKee court cited the findings of fact in Kozup. It appears that the court did not assess independently the evidence advanced by the litigants before it.

The differences between Kozup on the one hand and Doe and Smith on the other illustrate the danger in relying on the factual decisions of other courts. Kozup involved transmission through a blood transfusion, not through the use of Factor VIII. This difference is significant in several material ways. First, Factor VIII involves a higher concentration of blood donations and therefore involves a higher risk of transmission than the risk associated with a single blood transfusion obtained from a single donor. Second, Factor VIII is manufactured by only four producers, as opposed to the hundreds of blood banks and medical entities involved in the taking and storing of blood donations. Information about the threat of AIDS and various strategies to meet the threat could have easily been discussed and coordinated by the four appellees. With blood banks, by contrast, such pooling of information would have been extremely difficult. Finally, Factor VIII is a "product" produced from blood, whereas blood transfusions involve transferring whole blood directly to recipients. Therefore, Factor VIII manufacturers may have had a greater ability to intervene and take precautions in the production of Factor VIII in order to prevent the transmission of disease than did the entities responsible for administering blood banks.

Moreover, some specific findings of the Kozup court are obviously in conflict with the evidence in the instant cases. For example, the Kozup court concluded that the defendants could not have been liable for failure to utilize hepatitis surrogate testing because "plaintiffs can point to no organization, governmental or medical, which advocated the use of this test as a means for screening AIDS." 663 F. Supp. at 1057. Doe and Smith, by contrast, presented the district court with direct evidence that surrogate testing was discussed as early as December, 1982.

In short, it was inappropriate for the district court to rely on the findings of fact in Kozup. The fact that the McKee court committed the same error does not alter our analysis. Based on the evidence contained in the record, there is a genuine issue of material fact regarding the extent of the medical community's knowledge about AIDS transmission through the Factor VIII clotting agent and the manufacturers' ability to take steps to prevent the transmission of the disease. Because the state of the medical community's knowledge about AIDS is a controverted issue of fact, on remand, it is an issue to be submitted to the trier of fact.

3. Negligence

Finally, appellees contend, and the district court agreed, that they are entitled to summary judgment because they were not negligent as a matter of law. Once again, because we find that the question of negligence depends on the resolution of genuine issues of material fact, we conclude that summary judgment is inappropriate.

(a) Factor VIII Manufacturers

Since 1985, Factor VIII has been both screened for HIV and heat treated to kill the AIDS virus. Appellants claim that the manufacturers of Factor VIII were negligent in not utilizing heat treatment or surrogate testing to screen the clotting factor for HIV when they first suspected that AIDS was a blood borne virus, perhaps as early as late 1982. Appellants also argue that the manufacturers were negligent in failing to warn the users of Factor VIII of the risk of contracting the AIDS virus. Appellees respond that the prevailing standard of the industry was not to test or treat Factor VIII, that there was no consensus that AIDS was a blood borne virus until 1984, and that in the case of surrogate testing, there was no guarantee that even with such testing that appellants would have avoided infection. Therefore, appellees contend, they were not negligent as a matter of law.

We find appellees' arguments unpersuasive. The fact that a defendant in a negligence action was following the standards of its industry does not necessarily immunize that defendant from liability. Certainly, evidence of custom, usage, or industry practice is relevant in determining whether a particular defendant has met the appropriate standard of care. However, it is well-settled that "[p]roof of adherence to an industry practice or custom is not dispositive on the issue of negligence," Martinez v. Korea Shipping Corp., 903 F.2d 606, 610 (9th Cir. 1990), because "what ought to be done is fixed by a standard of reasonable prudence, whether it usually is complied with or not." Texas Pacific Ry. Co. v. Behymer, 189 U.S. 468, 470, 23 S.Ct. 622, 622, 47 L.Ed. 905 (1903). The possibility that industry standards may fall short of reasonable care is particularly acute, we believe, in a situation such as this where the entire industry is comprised of only four manufacturers. Here, the individual manufacturers may have a far greater influence and control over "industry" standards than do members of industries with greater numbers of participants.

The fact that surrogate testing may not have guaranteed pure blood is not necessarily a defense to liability. There is strong indication that, at a minimum, surrogate testing would have seriously curtailed the amount of HIV infected Factor VIII. A safety measure need not be perfect before its use is considered necessary to satisfy the standard of care. Again, the critical inquiry is whether the defendant acted reasonably. Martinez, 903 F.2d at 609.

As for the question of the medical community's consensus, it simply cannot be resolved on summary judgment. According to the evidence submitted to the district court, cases of hemophiliacs with AIDS were first reported in July 1982. At that time, appellees first became aware of the potential danger to their product. From that date on, representatives of the Factor VIII industry were engaged in discussions about ensuring the safety of the clotting factor. Thus, although it is true that the virus was not isolated until 1984, there is evidence supporting appellants' claim that the industry strongly suspected the new disease was caused by a blood borne virus before 1984 and that means of testing the blood supply were both available and considered.

Surrogate testing was discussed as early as December 1982. In January 1983, the National Hemophilia Foundation called for a screening of the blood supply. Two months later, a doctor from the Center for Disease Control asked Factor VIII manufacturers to screen their products for hepatitis antibodies. In December 1983, appellees planned a delaying tactic to ward off government-imposed surrogate testing. See supra note 6. In March 1984, industry representatives once again discussed and rejected the idea of surrogate testing. Although such testing may have prevented many cases of HIV infection, such testing was never used by the industry.

The technology for heat treating blood was developed in the late 1970s. The record reflects some dispute as to whether the technology was available in the early 1980s to heat treat Factor VIII specifically. However, it appears that the technology definitely was available by March 1983. It is possible that had the industry concentrated on applying knowledge about heat treatment to the treatment of Factor VIII, that technology could have become available at an earlier date. Heat treatment was never used by the industry during the period at issue in these cases. Since heat treatment is 100% effective in destroying HIV, the process could have been an important tool in ensuring the safety of blood products.

In short, appellants have introduced significant evidence indicating that the manufacturers of Factor VIII could have taken steps to prevent appellants' infection with HIV contaminated Factor VIII. As this appeal is from a grant of summary judgment, we view the facts in the light most favorable to appellants. Viewing the evidence in that light, we find that there are genuine issues of material fact in dispute as to whether appellees knew that AIDS was a blood borne virus prior to 1984 and whether measures were available to counteract the transmission of the disease. We therefore conclude that the district court was incorrect in holding that appellees were not negligent as a matter of law, and find that the grant of summary judgment was inappropriate.

(b) United States

Appellants claim that TAMC's, and therefore the United States', negligence is manifested in its breach of its duty of care to the appellants because it did not warn them of the possible danger of infection through Factor VIII. Appellants also claim that the United States was negligent for failing to switch from the administration of Factor VIII to the administration of cryoprecipitate, a crude form of clotting factor made from frozen plasma, and for failing to dispense only heat-treated Factor VIII. Appellants claim that had they known about the risk they would have altered their lifestyles in order to curtail the frequency of necessity for the clotting factor and that they would have switched from Factor VIII to cryoprecipitate. The United States responds that it had no duty to warn appellants or abandon the use of Factor VIII since the risk of HIV infection from blood transfusions was not a recognized risk of harm at the time of appellants' seroconversions. With regard to the utilization of heat treatment, the United States makes similar arguments about the medical community's knowledge of AIDS and AIDS transmission. Appellants, however, have presented considerable evidence that the medical community was aware of the fact that AIDS was caused by a blood borne virus before the virus itself was isolated, as discussed above. We are presented, once more, with a factual question not amenable to summary judgment. Whether TAMC physicians were aware of the danger of Factor VIII and therefore had a duty to warn appellants or cease utilizing Factor VIII is a question best resolved at trial.

II. CHALLENGES TO OTHER TRIAL COURT DECISIONS

A. Exclusion of Expert Testimony

Doe challenges the district court's refusal to admit the testimony of his expert witnesses. We review a decision of the district court to exclude expert testimony for an abuse of discretion. McGlinchy v. Shell Chemical Co., 845 F.2d 802, 806 (9th Cir. 1988).

1. Exclusion for Lack of Personal Knowledge

The district court granted appellees' motion to strike the affidavits of J. Garrot Allen, William T. O'Connor, and Thomas Drees, which were submitted by Doe in support of his Response to Defendants' Motion for Summary Judgment. The court struck the affidavits because they were "not based on any facts within the affiant's personal knowledge." Therefore, the court concluded, the affidavits violated Fed.R.Civ.P. 56(e) and Fed.R.Ev. 702 and 703. In so holding, the court did not specify what facts were beyond the affiants' expertise.

William O'Connor is a doctor who has testified as an expert on AIDS in two other lawsuits and before the California State Assembly and the U.S. House of Representatives. His affidavit concerned the history of blood screening and his review of Doe's medical records. Appellees argue that the district court was correct in striking his affidavit because he is not a hematologist and his expertise was based upon knowledge of medical literature on AIDS and not actual research or clinical work.

Thomas Drees is the former CEO of Alpha and is an expert on the manufacturing of blood products. His affidavit described what he learned about AIDS at meetings of Factor VIII manufacturers and what the prevailing safety practices of the industry were in the period at issue in this case. Appellees argue that he is not a doctor and that he admitted at deposition that his views about a "conspiracy" among Factor VIII manufacturers to delay testing were based on hindsight.

J. Garrot Allen is a surgeon and an expert on blood products and hematology. He has testified as an expert on AIDS in another lawsuit. He has been very active over the years in the legislative enactment of laws and regulations governing the "paid," as opposed to volunteer, donation of blood. His affidavit concerned regulation of the blood supply and blood products. Appellees claim that his affidavit should have been struck because he is not a licensed hematologist, he retired in 1977 and he has not personally worked with AIDS.

Although we are mindful of the deference generally accorded trial courts on evidentiary matters, we find that in this case the court abused its discretion in striking these affidavits. The fact that the experts were not licensed hematologists does not mean that they were testifying beyond their area of expertise. Ordinarily, courts impose no requirement that an expert be a specialist in a given field, although there may be a requirement that he or she be of a certain profession, such as a doctor. McCormick, On Evidence, § 13, at 34 (3rd ed. 1984). See also United States v. Viglia, 549 F.2d 335, reh'g denied, 553 F.2d 101 (5th Cir.), cert. denied, 434 U.S. 834, 98 S.Ct. 121, 54 L.Ed.2d 95 (1977) (pediatrician with no background treating obesity allowed to testify about the effect of drug on obese people). Therefore, we fail to see the basis for appellees' assertion that the experts in this case must be hematologists. Moreover, we note that AIDS is not a blood disorder, it is a virus. As a result, virologists and infectious disease specialists may have more knowledge about the transmissibility of the disease than do hematologists, who are knowledgeable about hemophilia and other blood disorders.

We therefore conclude that there was no basis for excluding the affidavits of Allen and O'Connor. The fact that their expertise may have been acquired through the study of medical literature is not a bar to admissibility. McCormick, § 13 at 34. Finally, we think it improper to have struck Drees' affidavit merely because he was not a doctor. Drees' testimony did not concern medical knowledge, it concerned the practices of a manufacturing industry of which he was an integral member. We therefore find that Drees was highly competent to testify as to the standards and practices of the industry.

In short, all of these experts were knowledgeable about the issues their testimony concerned. All of them had important information to contribute to the factfinding process. Since we do not see any valid bases for the district court's decisions to strike these affidavits, we conclude that the court's granting of the motion to strike was an abuse of discretion. On remand, the district court is instructed to admit this testimony.

2. Exclusion for Creating A Genuine Issue of Material Fact

The district court struck the affidavit of William Robinson because the court found that his affidavit contradicted his own sworn testimony and therefore was being used to create issues of material fact. Robinson, an internist and infectious disease specialist, testified as to the generic nature of Factor VIII, the percentage of Factor VIII that has been tainted with HIV and as to the likely time period when Doe seroconverted. In his affidavit Robinson stated that Doe had seroconverted in late 1983 or 1984. In his deposition, he admitted that he could not pinpoint the date with certainty.

It is true that a party cannot create its own issue of fact by having a witness contradict himself between affidavit and deposition testimony. Foster v. Arcata Associates, Inc., 772 F.2d 1453, 1462 (9th Cir. 1985), cert. denied, 475 U.S. 1048, 106 S.Ct. 1267, 89 L.Ed.2d 576 (1986). We do not believe, however, that Robinson's affidavit and deposition testimony were in conflict. Robinson never claimed to know exactly the date of Doe's seroconversion. In his affidavit, he ventured an approximate date, and at his deposition he merely stated that he could not provide the date with any greater accuracy. In any event, we have already held that there is a genuine issue of material fact with regard to the dates of seroconversion that exists independent of Robinson's testimony. On remand, the district court is therefore instructed to admit Robinson's affidavit.

B. Motion to Amend Complaint

Doe filed a motion before the district court requesting leave to amend his complaint to add a products liability allegation. The district court denied Doe's motion on the theory that Hawaii's Blood Shield Law precluded such liability. The Hawaii Supreme Court has affirmed that legal conclusion. Cutter Biological, 823 P.2d at 722. Accordingly, we affirm the denial of the motion to amend.

C. Inadequate Discovery

Smith challenges the district court's decision to end discovery and enter a judgment. We review such decisions for an abuse of discretion. Landmark Development Corp. v. Chambers Corp., 752 F.2d 369, 373 (9th Cir. 1985) (per curiam).

Smith claims that the district court prematurely cut off discovery when it entered summary judgment. He argues that there are issues which he could have further explored and supported had he been able to conduct further investigation. For example, he claims he could have acquired more information about the supply of Factor VIII to TAMC, and that such information would have supported further his negligence claim.

Because we have found that summary judgment was prematurely granted, we agree that on remand the district court should permit additional discovery as part of the factfinding process.

CONCLUSION

The grants of summary judgment in favor of defendants-appellees Cutter, Armour, Baxter, Alpha and the United States are REVERSED. Both Doe and Smith are REMANDED to the district court for further proceedings consistent with this opinion and the opinion of the Hawaii Supreme Court.