OPINION

Plaintiffs Halia S. Marsh, executrix of the estate of Robert Sidney Marsh, Thomas R. Speer, and Theodore McLeod; executor of the estate of Beulah M. McLeod, appeal from the district court’s decision granting summary judgment to defendants W.R. Grace & Co. (Grace), Frit Industries, Inc. (Frit), and Continental Casualty Company. Marsh v. W.R. Grace & Co., No. 3:90CV00613 (M.D.N.C. May 26, 1998). Plaintiffs brought a product liability action against the defendants stemming from their use of fertilizer contaminated with the chemical picloram. They alleged that the picloram caused Marsh, Speer, and Mrs. McLeod to develop cancer. The plaintiffs had one expert, Dr. Scott Levin, who testified that the contaminated fertilizer caused the plaintiffs’ cancers. The district court found that Dr. Levin’s testimony did not satisfy the admissibility requirements for expert testimony required by the Supreme Court in Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993), and consequently, granted the defendants’ motions for summary judgment because the plaintiffs provided no additional evidence that the picloram caused their injuries. Plaintiffs appeal from that decision and from the district court’s denial of the motions of Mrs. Moore and Speer to amend their complaints to include chemicals other than picloram as the cause of their injuries. Marsh v. W.R. Grace & Co., No. 3:90CV00613 (M.D.N.C. Oct. 13, 1992); Speer v. W.R. Grace & Co., No. 3:91CV00048 (M.D.N.C. Dec. 9, 1994). For the reasons set forth below, we affirm the decisions of the district court.

I.

The underlying facts of this case are not complicated and largely not in dispute. Because we are reviewing disposition of this case on summary judgment, we examine the facts as found by the district court after it construed the case in the light most favorable to the plaintiffs. Beall v. Abbott Labs., 130 F.3d 614, 618-19 (4th Cir. 1997). In 1984, Grace manufactured and sold Gold Dollar tobacco fertilizer using a micronutrient mix provided by Frit which included picloram. All of the plaintiffs are North Carolina farmers—Marsh and Speer worked on Marsh’s son’s, Waymon Marsh’s, farm, and the McLeods worked on their own farm. The plaintiffs purchased and used, either or both, Grace’s Gold Dollar fertilizer on tobacco crops and on vegetable gardens. Each of the plaintiffs subsequently discovered that the Gold Dollar fertilizer contained several ingredients it should not have, including picloram, a broadleaf herbicide. For several years, the plaintiffs were exposed to the picloram while handling the fertilizer itself, from contact with soil and water contaminated by picloram from the fertilizer, by eating vegetables from gardens contaminated by the fertilizer, and from eating fish from a pond contaminated by fertilizer run-off.

Between 1984 and 1988, all of the plaintiffs suffered from some type of cancer. Marsh was diagnosed with lung cancer in October 1988, and he died two weeks later at the age of 78. Speer discovered he had colorectal cancer in March 1988 at the age of 65, but he received successful treatments for the disease. Mrs. McLeod was diagnosed with cervical cancer in August 1986. She died from the disease two years later at the age of 74.

*885 In 1990, each of the plaintiffs filed their suits against Frit, Grace, and Continental Casualty Company in North Carolina state court. All three complaints alleged that the defendants failed to warn the plaintiffs of the contaminated fertilizer, made false representations concerning the contaminants in the fertilizer, negligently manufactured the fertilizer, breached their express and implied warranties, and participated in unfair settlement practices. The plaintiffs all claimed damages incurred as a result of their respective cancers, which they alleged were proximately caused by the picloram in the fertilizer.

The defendants removed all three cases to federal court in December 1990. The parties then engaged in extensive discovery, including an exhaustive deposition of the plaintiffs’ expert witness, Dr. Levin. On October 28, 1991, the defendants moved for summary judgment in the Marsh case. In March 1992, Mrs. Marsh made a motion pursuant to Federal Rule of Civil Procedure 15(a) to amend her complaint to include chemicals other than picloram, specifically 2,4-D. The court declined her motion on the basis that she had unduly delayed making the motion and that the defendants would be unfairly prejudiced if it was granted. In October 1994, soon after the close of discovery, Speer made a motion to amend his complaint under Rule 15(a) that was essentially identical to the one Mrs. Marsh had previously made. In November 1994, the defendants filed motions for summary judgment in the Speer and McLeod cases. The court denied Speer’s motion to amend his complaint citing the same reasons it denied Mrs. Marsh’s motion in addition to the fact that the parties had again reached summary judgment stage in the proceedings. On May 26, 1998, after reviewing the defendants’ motions for summary judgment in all three cases, the district court found that Dr. Levin’s testimony that picloram caused the plaintiffs’ cancers did not meet the requirements for admissibility of expert testimony set forth in Daubert. Marsh v. W.R. Grace & Co., No. 3:90CV00613, slip op. at 12 (M.D.N.C. May 26, 1998). Without Dr. Levin’s testimony, the district court found that there was no genuine issue of material fact with respect to proximate causation and granted the defendants’ motions for summary judgment. Marsh, No. 3:90CV00613, slip op. at 3-5. The plaintiffs now appeal the district court’s exclusion of Dr. Levin’s testimony and denial of their requests to amend their complaints.

II.

A.

We review a district court’s decision granting a motion for summary judgment de novo. Metropolitan Life v. Pettit, 164 F.3d 857, 860 (4th Cir. 1998). The district court’s decision on the motion for summary judgment in this case turned on the exclusion of expert testimony under Daubert. We review a district court’s decision to exclude expert testimony for abuse of discretion. General Elec. Co. v. Joiner, 522 U.S. 136, 139, 118 S.Ct. 512, 139 L.Ed.2d 508 (1997). We first address the plaintiffs contention that the district court erroneously excluded Dr. Levin’s testimony and, second, the court’s subsequent decision to grant summary judgment for the defendants.

In Daubert, the Supreme Court set forth the two-part inquiry a district court must engage in when determining whether to admit expert testimony. The district court must exercise a gatekeeping function in which it only admits expert testimony that is both reliable and relevant. Daubert, 509 U.S. at 589, 113 S.Ct. 2786. We are only concerned with the reliability prong of the test in this case because the evidence *886 was relevant to the causation element of the plaintiffs’ case.

The Daubert analysis is a flexible one, Daubert, 509 U.S. at 592-93, 594-95, 113 S.Ct. 2786, and, to aid courts engaging in the reliability analysis, the Court provided four non-exclusive factors that the district court could consider when evaluating proposed expert testimony: (1) whether the evidence can be tested, (2) whether the theory has been subjected to peer review or publication, (3) the known or potential rate of error of the technique, and (4) whether the relevant scientific community has generally accepted the theory. Daubert, 509 U.S. at 593-94, 113 S.Ct. 2786. The focus of the district court’s inquiry should be the principles and methodology of the expert, not the conclusions reached. Daubert, 509 U.S. at 595, 113 S.Ct. 2786. The overall goal of the Daubert analysis is to ensure that “an expert, whether basing his testimony upon professional studies or personal experience, employs in the courtroom the same level of intellectual rigor that characterizes the practice of an expert in the relevant field.” Kuhmo Tire Co. v. Carmichael, 526 U.S. 137, 152, 119 S.Ct. 1167, 143 L.Ed.2d 238 (1999).

The Court instructs us to provide district courts with “broad latitude,” p. 142, or “considerable leeway,” p. 156, both as to the method used to decide whether to admit expert testimony and to the decision whether to admit or exclude the testimony. Kuhmo Tire, 526 U.S. pp. 142, 156, 119 S.Ct. 1167. Under the abuse of discretion standard, we examine the record and the rationale of the district court and reverse its decision only if it made a clear error in judgment. Westberry v. Gislaved Gummi AB, 178 F.3d 257, 261 (4th Cir. 1999). We are of opinion that the record supports the district court’s analysis and decision to exclude Dr. Levin’s testimony.

Our review of the district court’s opinion discloses that the court found Dr. Levin’s methods and logical processes used to reach his opinion on causation were unreliable. Dr. Levin is a medical doctor specializing in pathology, immunology, and cancer who has researched and written about cancer. In determining the cause of his patients’ cancers, Dr. Levin relies on his evaluation of individual responses to a given intrusive agent. His theory is that all disease is a function of the host, not of the etiologic agent. So, if the body is set up for cancer, then the etiologic agent can cause cancer. Dr. Levin does not focus on epidemiological studies, which indicate percentages of cancer occurrences from particular causes in designated populations to aid scientists in determining whether a specific agent can cause cancer. We have held that that epidemiological evidence is not necessarily required for a valid expert opinion on causation so long as the expert’s methods are otherwise sound. Benedi v. McNeil-P.P.C., Inc., 66 F.3d 1378, 1384 (4th Cir. 1995). However, the district court, in the case at hand, engaged in a lengthy analysis of Dr. Levin’s methods using each of Daubert’s four suggested factors and found that Dr. Levin’s methods were not reliable.

Daubert first suggests that we should evaluate whether an expert’s opinion is subject to testing. Daubert, 509 U.S. at 593, 113 S.Ct. 2786. The district court found that Dr. Levin’s testimony was not subject to testing because he started from the conclusion that picloram caused the plaintiffs’ cancers and then generated, without testing, the hypotheses to support that conclusion. Marsh v. W.R. Grace & Co., No. 3:90V00613, slip op. at 7 (M.D.N.C. May 26, 1998). The record supports the district court’s decision that Dr. Levin’s opinion is unreliable because it cannot be tested or verified consistently, and therefore, the court did not abuse its discretion.

*887 Dr. Levin used the results of only one laboratory test performed on two of the plaintiffs to reach his opinion on causation. This test, however, did not identify the agent that caused the plaintiffs’ cancers. Dr. Levin provided no other medical tests to verify his hypothesis that picloram caused the plaintiffs’ cancers. We conclude that the district court did not abuse its discretion when it found that Dr. Levin needed more verification for his opinion because the defendants proposed several alternative causes for the plaintiffs cancers. The defendants presented evidence that showed the plaintiffs had been exposed to a multitude of chemicals that could cause cancer during their lifetimes, that one plaintiff had a family history of the type of cancer he contracted, and that another contracted lung cancer after a 50-year history of smoking. Dr. Levin, however, failed to verify or provide a means to verify his conclusion by ruling out any alternative causes for the plaintiffs’ cancers. While an expert need not rule out all alternative causes to provide a valid opinion, he must at least explain why his opinion remains sound in light of alternative causes proposed by the other parties. Westberry v. Gislaved Gummi AB, 178 F.3d 257, 265-66 (4th Cir. 1999). This record supports the district court’s finding that he did not do so adequately.

The district court also found that Dr. Levin’s testimony failed the second Daubert factor because his theories underlying his conclusion had not been subject to peer review or publication. Marsh, No. 3:90V00613, slip op. at 10. While we acknowledge that the Daubert analysis must be flexible and that not all medical experts’ theories and methods will need to be published or peer reviewed, this is a valid consideration in considering the facts of a particular case. Dr. Levin’s determination that picloram is a carcinogen is inapposite to much of the scientific literature on the subject. Neither the National Toxicology Program nor the Environmental Protection Agency have listed picloram as a carcinogen. To support his conclusion that picloram is a carcinogen, Dr. Levin stated that he read abstracts of several articles and looked at other studies. But in the face of literature and findings contradictory to Dr. Levin’s determination, the district court did not abuse its discretion in finding that Dr. Levin needed further support for his independent conclusion that picloram causes cancer.

The district court noted that Daubert’s third factor, the potential rate of error of a technique, is not directly applicable to Dr. Levin’s testimony because his technique is not susceptible to a measurable rate of error. However, we find that the district court did not abuse its discretion in deciding that Dr. Levin’s failure to identify a no-effect level for picloram, and his untested assumptions with respect to the amounts of pecloram to which plaintiffs were exposed, introduce an unacceptable potential for error into his opinion that the amount of picloram to which plaintiffs were exposed caused their cancers. While Dr. Levin agreed that picloram had a no-effect level, he did not state what it was or whether the amounts plaintiffs were exposed to exceeded that level. Apparently he assumed that the plaintiffs’ exposure to the chemical was about three years, and he assumed the plaintiffs had received adequate exposure to picloram to cause cancer. We find that the record supports the district court’s decision that this means of analysis by assumption could lead to an unacceptable potential for error.

After reviewing Dr. Levin’s affidavits and deposition testimony, we find sufficient support for the district court’s determination that Dr. Levin’s methods are not generally accepted by the medical community, the final Daubert factor. While plaintiffs argue that Dr. Levin relies on a meth *888 odology used by other immunologists and cite a portion of an article in the Journal of the American Medical Association in support of that contention, the article only supports Dr. Levin’s proposition that population studies are not determinative. While potentially true, the district court, supported by the record, found that Dr. Levin did not demonstrate that his methods used in this case are also used by experts in his field. Additionally, the record supports the district court’s conclusion that Dr. Levin’s opinion that picloram is a carcinogen is not generally accepted in the scientific community for essentially the same reasons we noted that this conclusion needed support either in scientific testing or through validation by peer review or publication.

B.

We grant a party’s motion for summary judgment under Rule 56(c) only if the case presents no genuine issues of material fact. Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 247-48, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986). The only evidence the plaintiffs provided indicating that the picloram in Grace’s fertilizer caused their cancers was the expert testimony of Dr. Levin. Without that evidence, the plaintiffs’ case is devoid of the proximate causation element. Thus, there is no genuine issue as to causation, and we are of opinion the district court properly granted the defendants’ motion for summary judgment on that basis.

III.

We review a district court’s decision denying leave to amend a complaint under Rule 15(a) for abuse of discretion. Foman v. Davis, 371 U.S. 178, 182, 83 S.Ct. 227, 9 L.Ed.2d 222 (1962). A district court should grant leave to amend unless the motion was made in bad faith or after undue delay, or with a dilatory motive, or the amendment would cause undue prejudice to the non-moving party, or the amendment would be futile. Island Creek Coal Co. v. Lake Shore, Inc., 832 F.2d 274, 279 (4th Cir. 1987) (quoting Foman, 371 U.S. at 182). A district court should not deny leave to amend because of mere delay. Island Creek, 832 F.2d at 279. Delay must be accompanied by prejudice, bad faith, or futility to comprise sufficient basis for denial of leave to amend. Island Creek, 832 F.2d at 279. Under the facts of this case, we are of opinion that the district court did not abuse its discretion in denying the plaintiffs leave to amend their complaints.

The district court denied Mrs. Marsh’s 1992 motion to amend her complaint to include 2,4-D as a cause of her cancer after finding not only that the plaintiff had unduly delayed in making the motion, but also that granting the motion would cause prejudice to the defendants. The court observed that the parties had engaged in extensive, continuing, and expensive discovery directed to picloram. Adding another chemical at that point in the proceedings, after defendants had filed a motion for summary judgment, would basically force the parties to start the discovery proceedings anew. Additionally, the court noted that the plaintiff had known of the presence of 2,4-D since early in the litigation and, essentially, had no justifiable reason for waiting to file her motion to amend when the court had granted one such motion in early 1991. Unlike the case in Island Creek, discovery here did not progress at a desultory pace, the plaintiff did not discover the new contaminant recently, and both parties had concentrated their discovery and efforts on picloram. See Island Creek, 832 F.2d at 279-80 (finding district court abused its discretion in denying leave to amend complaint to include a new, recently discovered basis for relief).

*889 The district court was also acting within its discretion when it denied Speer’s 1994 motion to amend his complaint. The four year delay before Speer’s motion to amend was unnecessary. The same attorney represents all three plaintiffs, and he made an almost identical motion to amend in Mrs. Marsh’s case. The two year delay between that motion and the one in Speer’s case is unjustifiable. All of the facts necessary for such a motion had been available to the attorney for over four years. Additionally, as the district court noted, granting the motion at that time would prejudice the defendants. The case had reached summary judgment stage and the parties had completed years of discovery that the defendants would be forced to repeat if the court allowed a new causal chemical into the case. It appears that the plaintiffs’ amendment was no more than a last chance attempt to ward off summary judgment, which is not well taken. Cf. Sandcrest Outpatient Servs., P.A. v. Cumberland County Hosp. Sys., Inc., 853 F.2d 1139, 1149 (4th Cir. 1988).

While we acknowledge that plaintiffs should have the right to amend their complaints freely as the ends of justice require, we are of opinion that the facts in this case support the district court’s discretionary decisions to deny amendment to the complaints.

For the reasons expressed in the opinion of the district court, its judgment is accordingly

AFFIRMED. *