FOR PUBLICATION
  UNITED STATES COURT OF APPEALS
       FOR THE NINTH CIRCUIT

RANCHERS CATTLEMEN ACTION                
LEGAL FUND UNITED
                                               No. 06-35512
STOCKGROWERS OF AMERICA,
              Plaintiff-Appellant,
                                                D.C. No.
                                             CV-05-00006-RFC
              v.
                                                  OPINION
U.S. DEPARTMENT OF AGRICULTURE,
             Defendant-Appellee.
                                         
        Appeal from the United States District Court
                for the District of Montana
        Richard F. Cebull, District Judge, Presiding

                    Argued and Submitted
               July 13, 2007—Portland, Oregon

                     Filed August 28, 2007

   Before: Cynthia Holcomb Hall and Milan D. Smith, Jr.
  Circuit Judges, and Kevin Thomas Duffy,* Senior Judge.

            Opinion by Senior Circuit Judge Hall




  *The Honorable Kevin Thomas Duffy, Senior United States District
Judge for the Southern District of New York, sitting by designation.

                              10697
10700         RANCHERS CATTLEMEN v. USDA


                     COUNSEL

Russell S. Frye, FryeLaw PLLC, Washington, DC, for the
appellant.
                 RANCHERS CATTLEMEN v. USDA                10701
Joshua Waldman and Mark B. Stern, Department of Justice,
Civil Division, Washington, DC, for the appellees.

Maureen E. Mahoney, Latham & Watkins LLP, Washington,
DC, for amicus Government of Canada.

Edward J. Farrell, Blank Rome LLP, Washington, DC, for
amicus Canadian Cattlemen’s Association.

Simeon M. Kriesberg, Mayer, Brown, Rowe & Maw LLP,
Washington, DC, for amicus Alberta Beef Producers.

Philip C. Olsson, Olsson, Frank and Weeda, P.C., Washington
DC, and John W. Ross, Brown Law Firm, P.C., Billings,
Montana, for amici National Meat Association et al.


                          OPINION

HALL, Senior Circuit Judge:

   This case involves a challenge to the government’s regula-
tion of Canadian cattle imports in the wake of the “mad cow
disease” scare of the late 1990s. Ranchers Cattlemen Action
Legal Fund United Stockgrowers of America (“R-CALF-
USA” or “R-CALF”) argues that the United States Depart-
ment of Agriculture (“USDA”) issued an arbitrary and capri-
cious rule relaxing a ban on Canadian beef and cattle imports.
See Bovine Spongiform Encephalopathy; Minimal-Risk
Regions and Importation of Commodities, 70 Fed. Reg. 460
(January 4, 2005) (hereinafter “the Final Rule”).

   R-CALF argues that recent incidents of mad cow disease in
the Canadian herd, and in American cows imported from the
Canadian herd, cast doubt on the agency’s rulemaking proce-
dure. With additional references to scientific studies and inter-
national regulations, R-CALF challenges the agency’s
10702             RANCHERS CATTLEMEN v. USDA
assessment that the “multiple, interlocking safeguards” imple-
mented by both the United States and Canada will be effective
at preventing human infection domestically.

   The district court granted summary judgment to the USDA,
and we have jurisdiction to review this order under 28 U.S.C.
§ 1291. The facts have been provided in prior related deci-
sions and will not be recited exhaustively here. See Ranchers
Cattlemen Action Legal Fund United Stockgrowers of Am. v.
United States Dep’t of Agric., 

(D.
Mont.) (hereinafter “R-CALF I”), rev’d 

(9th
Cir. 2005) (hereinafter “R-CALF II”).

  We affirm.

                            Background

   Commonly referred to as “mad cow disease,” Bovine
Spongiform Encephalopathy (“BSE”) is a degenerative neuro-
logical disease that was first discovered in 1986 and has since
infected more than 187,000 cattle worldwide, with 95 percent
of the cases occurring in England. In the mid-1990s, public
health officials discovered that cattle feeding practices were
the likely cause of an outbreak of BSE in England. At the
time, cattle feed typically contained recycled or “rendered”
cattle parts that gave it a higher protein content.1

   In 1996, the British government discovered that consump-
tion of BSE-contaminated meat could could cause variant
Creutzfeld-Jakob Disease (vCJD) in humans. There have been
approximately 150 human cases of vCJD, including one case
in the United States in a woman who had probably contracted
the disease while living in England. Scientists are still learn-
ing how these diseases develop, incubate and spread. See R-
CALF 

.
  1
    See generally PL 107-9 Federal Inter-agency Working Group, “Animal
Disease Risk Assessment, Prevention, and Control Act of 2001 (PL 107-
9), Final Report,” January 2003.
                    RANCHERS CATTLEMEN v. USDA                        10703
   The Food and Drug Administration and the USDA
responded to the BSE outbreak with new regulations.2 These
rules prohibited the use of mammalian proteins in cattle feed,
see 21 C.F.R. § 589.2000, and prohibited the use of “specified
risk materials” — such as cattle brains, spinal cords, and
nerve tissue — in human food, see 9 C.F.R. § 310.22. The
USDA, working closely with the World Organization for Ani-
mal Health, developed guidelines and proposed protective
measures to prevent the spread of BSE to the United States.
See 70 Fed. Reg. at 463.3 Chief among these measures was a
ban on imports of all cattle products from countries where
BSE was known to exist. See 9 C.F.R §§ 93.401, 94.18
(2003). The USDA added Canada to this list of countries in
May 2003, after a cow in Alberta was diagnosed with BSE.
Change in Disease Status of Canada Because of BSE, 68 Fed.
Reg. 31939 (May 29, 2003). Though Canada had instituted its
own feed ban in 1997, it was likely that the cow had been
exposed before the ban and that the disease had incubated for
a period of years.4 The USDA estimates that the disease has
an incubation period of two to eight years. 70 Fed. Reg. at
470.

   In August 2003, the agency partially changed course and
announced that certain “low-risk” cattle products could be
imported from Canada, including meat from cows under 30
months of age. See 70 Fed. Reg. at 536. In November 2003,
it also announced a proposed rule creating a new category of
“minimal risk” regions — a category that would include Can-
  2
     The FDA is under the Department of Health and Human Services. The
USDA acted through two subsidiary agencies, the Food Safety and Inspec-
tion Service and the Animal and Plant Health Inspection Service. We will
refer to actions by these agencies as actions by the USDA.
   3
     This organization is also called the Office International des Epizooties,
or “OIE.” It is responsible for the development of standards and recom-
mendations regarding animal health and “zoonoses” (diseases that are
transmissible from animals to humans). See 70 Fed. Reg. at 463.
   4
     Scientists thus far believe that BSE is caused by protein-based infec-
tious agents called “prions.” 70 Fed. Reg. at 461.
10704              RANCHERS CATTLEMEN v. USDA
ada and possibly other countries. See Bovine Spongiform
Encephalopathy; Minimal Risk Regions and Importation
of Commodities, 68 Fed. Reg. 62386 (Nov. 4, 2003).

   Shortly before the comment period on this rule was to end,
a Canadian-born cow in Washington state was diagnosed with
BSE, likely caused by feed ingested before the Canadian feed
ban went into effect. The USDA reopened the comment
period in March with an expiration date of April 7, 2004. See
Bovine Spongiform Encephalopathy; Minimal Risk Regions
and Importation of Commodities, 69 Fed. Reg. 10,633 (Mar.
8, 2004). By the close of the comment period, the agency had
received 3,379 comments. See 70 Fed. Reg. at 465.

   On January 4, 2005, after a struggle with R-CALF over an
interim regulation,5 the USDA published the Final Rule,
which modified existing regulations to allow imports of Cana-
dian cattle under 30 months of age for purchase by feedlots
or meat packing companies. See 

9 CFR §§ 93.420,
93.436, 94.0, 94.18, 94.19, 95.4; R-CALF 

n.10. The rule at this stage also allowed in Canadian beef
products from cattle of all ages. 70 Fed. Reg. at 494.

   Shortly after the rule was published, two older cows in
Alberta were diagnosed with BSE, and the USDA attributed
the disease to contaminated feed manufactured before the
Canadian feed ban. The USDA then announced its intention
to suspend the part of the rule that would relax the ban on
  5
   The USDA had moved, without public notice, to expand the types of
ruminant products eligible to be imported from Canada. R-CALF sued to
prevent this move, and the district court granted a temporary restraining
order on April 26, 2004. See Ranchers Cattlemen Action Legal Fund
United Stockgrowers of Am. v. United States Dep’t of Agriculture, 

(D. Mont. April 26, 2004). The USDA included a formal
version of this amendment in the Final Rule but ultimately suspended its
implementation. See Bovine Spongiform Encephalopathy; Minimal-Risk
Regions and Importation of Commodities; Partial Delay of Applicability,
70 Fed. Reg. 12,112 (March 11, 2005).
                RANCHERS CATTLEMEN v. USDA               10705
meat from cattle over 30 months old. See Bovine Spongiform
Encephalopathy; Minimal-Risk Regions and Importation of
Commodities; Partial Delay of Applicability, 70 Fed. Reg.
12,112 (March 11, 2005).

  The Final Rule was set to go into effect on March 7, 2005,
but was blocked by a preliminary injunction from the district
court in R-CALF I, a ruling stemming from related proceed-
ings in this case.

                      Prior Proceedings

   On January 10, 2005, six days after the Final Rule was pub-
lished, R-CALF filed a complaint alleging that the USDA’s
rulemaking violated the Administrative Procedure Act (APA),
the Regulatory Flexibility Act, and the National Environmen-
tal Policy Act. It applied for a preliminary injunction, which
the district court granted on March 2, 2005. See R-CALF 

. The district court found that R-
CALF had demonstrated a likelihood of success on its claim
that the rule was arbitrary and capricious, in violation of the
APA. See 5 U.S.C. § 706(2). While an initial appeal to this
court was pending, the parties filed cross-motions for sum-
mary judgment with the district court.

   We reversed the preliminary injunction ruling on July 14,
2005, and issued a final amended opinion on August 17,
2005. See R-CALF II, 

. (We found that the dis-
trict court had not accorded adequate deference to the
USDA’s determinations and concluded that the agency “had
a firm basis for determining that the resumption of ruminant
imports from Canada would not significantly increase the risk
of BSE to the American population.” 

  In light of this order and opinion, the district court post-
poned, and ultimately never scheduled, a hearing on the cross-
motions for summary judgment. It denied R-CALF’s motion
10706            RANCHERS CATTLEMEN v. USDA
for summary judgment, and granted summary judgment to the
USDA on April 5, 2006.

   The district court’s order set out the deferential standard of
review and provided only one paragraph of analysis. It quoted
this court’s holding in R-CALF II, and then stated: “Based
upon this, the District Court’s hands are tied. The Ninth Cir-
cuit has instructed this court to ‘abide by this deferential stan-
dard,’ and ‘respect the agency’s judgment and expertise.’ ” It
offered no analysis of the record, which had been supple-
mented several times while the preliminary injunction appeal
was pending.

   R-CALF filed its timely appeal of the district court’s deci-
sion on June 2, 2006. Only the APA claim is before us.

              A Second Remand is Unnecessary

   R-CALF argues that the district court improperly deter-
mined it was bound by our decision reversing the preliminary
injunction, and therefore R-CALF requests that we remand to
the district court for analysis of the record that was developed
in support of the motion for summary judgment.

   R-CALF correctly points out that the ruling on the motion
for a preliminary injunction “leaves open the final determina-
tion of the merits of the case.” Ross-Whitney Corp. v. Smith
Kline & French Labs., 

, 194 (9th Cir. 1953).
This rule acknowledges that “decisions on preliminary injunc-
tions are just that — preliminary — and must often be made
hastily and on less than a full record.” S. Ore. Barter Fair v.
Jackson County, 

, 1136 (9th Cir. 2004) (citation
omitted).

   Accordingly, the district court should abide by “the general
rule” that our decisions at the preliminary injunction phase do
not constitute the law of the case. See id.; see also City of
Anaheim v. Duncan, 

, 1328 n.2 (1981). Any of
                 RANCHERS CATTLEMEN v. USDA                  10707
our conclusions on pure issues of law, however, are binding.
See This That And The Other Gift And Tobacco, Inc. v. Cobb
County, 

, 1284-85 (11th Cir. 2006); 18 Charles
Alan Wright & Arthur R. Miller, Federal Practice and Proce-
dure § 4478.5 (2002) (“A fully considered appellate ruling on
an issue of law made on a preliminary injunction appeal . . .
become[s] the law of the case for further proceedings in the
trial court on remand and in any subsequent appeal.”). The
district court must apply this law to the facts anew with con-
sideration of the evidence presented in the merits phase. See

; accord Washington Capitols
Basketball Club, Inc. v. Barry, 

, 476 (9th Cir.
1969).

   [1] This administrative law case presents a thornier issue
because this court has reviewed the administrative record for
the purpose of the injunction, and, with some narrow excep-
tions, neither we nor the district court may consider any other
evidence. See Asarco, Inc. v. E.P.A., 

, 1160 (9th
Cir. 1980). For summary judgment, R-CALF has presented
only new, extra-record evidence, of arguable relevance to this
court’s review. Still, technically, the district court was not
bound by our earlier conclusions.

   Though the district court erroneously determined other-
wise, remand is not the only option available at this stage of
the litigation. Our review of a summary judgment order pro-
ceeds de novo, see The Lands Council v. Powell, 

, 1026 (9th Cir. 2005), and in administrative appeals,
where the court reviews only the record before the agency,
“[t]he factfinding capacity of the district court is . . . typically
unnecessary to judicial review of agency decisionmaking.”
Fla. Power & Light Co. v. Lorion, 

, 744 (1985).
Simply put, this court’s task on appeal is the same as the dis-
trict court’s task in the initial review: “Both courts are to
decide, on the basis of the record the agency provides,
whether the action passes muster under the appropriate APA
standard of review.” 

CATTLEMEN v. USDA
   [2] Because all of R-CALF’s new evidence is outside the
administrative record and of very limited use, and because we
agree with amici that extension of this litigation will aid R-
CALF in its attempt to “create, on a rolling basis, a one-sided
evidentiary record that supersedes USDA’s administrative
record,” Brief of the Government of Canada at 15, we decide
to reach the merits of this case.

                        Legal Standard

    Under the APA, an agency action may be set aside only if
it is “arbitrary, capricious, an abuse of discretion, or otherwise
not in accordance with law.” 5 U.S.C. § 706(2)(A). We must
determine whether the agency “considered the relevant factors
and articulated a rational connection between the facts found
and the choices made.” City of Sausalito v. O’Neill, 

, 1206 (9th Cir. 2004) (citation omitted). This standard
of review is “highly deferential, presuming the agency action
to be valid and affirming the agency action if a reasonable
basis exists for its decision.” Indep. Acceptance Co. v. Cali-
fornia, 

, 1251 (9th Cir. 2000) (citations omit-
ted).

   In its paradigmatic statement of this standard, the Supreme
Court explained that an agency violates the APA if it has “re-
lied on factors which Congress has not intended it to consider,
entirely failed to consider an important aspect of the problem,
offered an explanation for its decision that runs counter to the
evidence before the agency, or is so implausible that it could
not be ascribed to a difference in view or the product of
agency expertise.” Motor Vehicle Mfrs. Ass’n v. State Farm
Mutual Auto. Ins. Co., 

, 43 (1983).

               The USDA’s Rulemaking Proces

   The Animal Health Protection Act authorizes the Secretary
of Agriculture to ban imports of animals if “necessary to pre-
vent the introduction . . . of any pest or disease of livestock.”
                RANCHERS CATTLEMEN v. USDA                10709
7 U.S.C. § 8303. We previously remarked that the statutory
language “indicate[s] a congressional intent to give the Secre-
tary wide discretion in dealing with the importation of plant
and animal products.” R-CALF 

. The stat-
ute does not contain a strict requirement that the USDA elimi-
nate all risk that BSE will enter the country. 

Rule at issue in this case establishes a lower tier
of import restrictions for regions that pose a “minimal-risk”
of exporting BSE to the United States. Though the World
Organization for Animal Health would have based any lower-
risk designation on the statistical incidence rate of BSE in a
given country, the USDA disagreed with “holding a country
to a rigid criterion without consideration of compensatory risk
reduction measures,” such as surveillance programs. 70 Fed.
Reg. at 464. The agency instead defined a minimal-risk region
as one that had (1) implemented risk mitigation measures
(including import restrictions, surveillance and a feed ban)
before BSE was detected in the country, (2) conducted an epi-
demiological investigation after BSE was detected, and (3)
took additional risk mitigation measures after the BSE out-
break. See 9 C.F.R. § 94.0. The USDA did not establish
numerical criteria.

  [3] The agency accordingly designated Canada as a
“minimal-risk region” based on the following documents,
actions and considerations:

    •Regulations in the U.S., specifically the bans on the
    use of “specified risk materials” in human food from
    cattle over 30 months old, and the ban on meat from
    non-ambulatory cattle. 70 Fed. Reg. at 466.

    •The 2003 Revised Harvard-Tuskegee Study of BSE
    risk in the U.S., which concluded that there was a
    “very low risk” of BSE becoming established
    domestically if it were introduced. The study found
    that bans on U.K.-imported cattle and feed bans were
10710           RANCHERS CATTLEMEN v. USDA
    the most effective measures to prevent BSE intro-
    duction, and that the biggest risks for human expo-
    sure to BSE were non-compliance with the feed
    bans, use of other infected farm animals in feed, and
    use of high-risk tissues in products for human con-
    sumption. 

    •A memorandum from researchers at the Harvard
    Center for Risk Analysis, which updates the model
    from the Harvard-Tuskegee Study. 

by Canada prior to the discovery of
    BSE in 2003, including import restrictions on U.K.
    cattle and Canada’s 1997 feed ban. Because most
    infected cattle show clinical signs of BSE within
    seven years of infection, any cattle born before the
    feed ban would show clinical signs before the time
    of the Final Rule and therefore would be detected by
    surveillance. 

    •A 2002 assessment of BSE risk in Canada, finding
    that the 665 cattle imported from Europe between
    1979 and 1997 resulted only in a “low potential” for
    introduction of BSE infection. 

    •A 2003 epidemiological investigation and report
    after BSE detection, which found little exposure to
    BSE and determined that Canada’s protective mea-
    sures were effective and proposed additional mea-
    sures. 

taken in Canada, including a
    ban on specified-risk materials from cattle at slaugh-
    ter, a new epidemiological investigation, and
    increased surveillance. 

    •The agency’s update to its own risk analysis of Can-
    ada that provides a more detailed analysis of its rules
    and their application to Canada. 

                 RANCHERS CATTLEMEN v. USDA                10711
In its later affirmation of the Final Rule, the USDA empha-
sized that “the cumulative effect of all the measures in place
in Canada and the United States . . . is an extremely effective
set of interlocking, overlapping and sequential barriers to the
introduction and establishment of BSE in the United States.”
Bovine Spongiform Encephalopathy; Minimal-Risk Regions
and Importation of Commodities; Finding of No Significant
Impact and Affirmation of Final Rule, 70 Fed. Reg. 18,252,
18,255 (April 8, 2005).

   [4] We endorsed this holistic approach at the preliminary
injunction phase, when we chose to “evaluate the cumulative
effects of the multiple, interlocking safeguards” instead of fol-
lowing a “divide and conquer” strategy. R-CALF 

. Because this approach represents a legal conclusion
about the construction of the regulations, it is the law of the
case, and therefore we adopt it for our decision now.

                     R-CALF’s Allegations

   R-CALF initially brought five claims for declaratory and
injunctive relief but appeals only on the basis of its first APA
claim, in which it argues that the Final Rule is arbitrary and
capricious because it is based on faulty assumptions about the
efficacy of the American and Canadian feed and import
restrictions.

   [5] As we evaluate each argument under this claim, we will
consider whether R-CALF’s new evidence is relevant to our
review. It is an established rule that “the focal point for judi-
cial review should be the administrative record already in
existence, not some new record made initially in the review-
ing court.” Camp v. Pitts, 

, 142 (1973). Under
limited circumstances, however, extra-record evidence can be
admitted and considered.

   At the district court level, extra-record evidence is admissi-
ble if it fits into one of four “narrow” exceptions: (1) if admis-
10712           RANCHERS CATTLEMEN v. USDA
sion is necessary to determine whether the agency has
considered all relevant factors and has explained its decision,
(2) if the agency has relied on documents not in the record,
(3) when supplementing the record is necessary to explain
technical terms or complex subject matter, or (4) when plain-
tiffs make a showing of agency bad faith. Southwest Ctr. for
Biological Diversity v. U.S. Forest Serv., 

, 1450
(9th Cir. 1996) (internal punctuation omitted). R-CALF also
relies heavily on one statement in the case law that extra-
record information might be admitted if it tends to show that
the agency relied on assumptions that were “entirely fictional
or utterly without scientific support.” Ass’n of Pac. Fisheries
v. E.P.A., 

, 812 (9th Cir. 1980).

   In Asarco, Inc. v. E.P.A., 

(9th Cir. 1980), the
benchmark case on this issue in this circuit, the district court
had held a four-day hearing, which included testimony from
two experts who had not helped the agency make the chal-
lenged decision. This court disapproved because this testi-
mony was “plainly elicited for the purpose of determining the
scientific merit of the EPA’s decision.” 

Consider-
ing evidence outside this record is inappropriate, we
explained, because it “inevitably leads the reviewing court to
substitute its judgment for that of the agency.” 

Under the APA, courts must refrain from de novo review of
the action itself and focus instead on the agency’s decision-
making process. 

   [6] Under these principles, R-CALF’s arguments can only
carry the day if they show flaws in the USDA’s approach,
rather than in its predictions. We address each of R-CALF’s
arguments in turn below.

   1) The BSE Incidence Rate in Canada: R-CALF argues that
the agency relied on a Canadian report that used an insuffi-
cient sample size based on data collected in 2001, before the
Canadian and American BSE-infected cows were discovered.
The district court agreed, R-CALF 

-
                  RANCHERS CATTLEMEN v. USDA                   10713
66, but we held that the district court improperly substituted
its judgment for the agency’s, see R-CALF 

. We found that the USDA had based its calculations on
international standards, and that the World Organization for
Animal Health had ranked Canada in its minimal risk range
in 2003. 

   [7] R-CALF argues now that “data not available during the
preliminary injunction proceedings and appeal indicate, if
anything, an increasing prevalence of BSE, with five of the
nine cases in Canadian-born cattle having been diagnosed in
just the past year.” R-CALF argues that this post-decision
empirical data shows that the USDA was relying on faulty
assumptions that lacked scientific support.

   [8] While these new incidents are certainly cause for con-
cern, they do not suggest that the agency made an incomplete
or unreasoned review of the evidence before it in 2004. The
agency was entitled to rely on the reasonable opinion of its
experts at that time, see Marsh v. Ore. Natural Res. Council,

, 378 (1989), and the agency continues to moni-
tor BSE in Canada, see Bovine Spongiform Encephalopathy;
Minimal-Risk Regions; Importation of Live Bovines and
Products Derived From Bovines, 72 Fed. Reg. 1102 (Jan. 9,
2007). Because the Final Rule does not anticipate an inci-
dence rate of zero in Canada or the U.S., these subsequent
BSE cases do little to impugn the agency’s decision-making
process. If recent cases have cast doubt on the agency’s scien-
tific predictions, the proper remedy is to petition to reopen
rulemaking under 5 U.S.C. § 553(e), not to challenge the
existing rule as arbitrary and capricious.6

  2) The Effectiveness of the Canadian Feed Ban: R-CALF
argues that BSE may be transmitted through blood and saliva,
not just contaminated feed, that one of the BSE-infected cows
  6
   R-CALF’s related argument about the risk of BSE entering the United
States also fails for these reasons.
10714           RANCHERS CATTLEMEN v. USDA
was rendered into feed, and that the recent diagnoses of BSE
show that the feed ban is not working because of alleged non-
compliance. The district court credited this argument, see R-
CALF 

, but we held that the
agency had properly considered and rejected these alternative
theories of transmission. R-CALF 

. We
also held that it was appropriate for the agency to assume the
longer incubation rates for BSE in Canada to explain the more
recent cases of infected cattle. 

now argues that the recent incidents of BSE in
cows born after the feed ban prove that the feed ban is inef-
fective. It also argues that a government study on the U.S.
feed ban shows some noncompliance. See United States Gov-
ernment Accountability Office, Mad Cow Disease — FDA’s
Management of the Feed Ban Has Improved, but Oversight
Weaknesses Continue to Limit Program Effectiveness, Feb.
2005. It also refers to a statement by Secretary Johanns sug-
gesting that there are “questions that must be answered” about
the number of BSE incidents in Canada and remarking that
South Korea has continued to close its borders to American
beef. These post-decisional statements are far outside the
record and of little persuasive weight.

   [10] Though these recent incidents in younger cattle cer-
tainly cast doubt on the effectiveness of the feed ban, the
agency — at the time it made its decision — properly relied
on studies from both the World Organization for Animal
Health and the Harvard Center on Risk Analysis finding that
feed bans were the most effective way to prevent the spread
of BSE, see 70 Fed. Reg. at 463, 467, and, again, considered
them as a part of a system of safeguards, not as a sole preven-
tative measure. It bears repeating that the agency did not
assume 100 percent effectiveness of its measures. See 70 Fed.
Reg. at 511.

   In a related argument, R-CALF claims that the agency
incorrectly assumed that the Canadian feed ban, which
                   RANCHERS CATTLEMEN v. USDA                      10715
exempted products made from animal blood or fat, would be
as effective as the European feed ban, which does not have
these exemptions. The agency expressly considered this argu-
ment and rejected it because Canada’s feed ban was equiva-
lent to the feed ban in the United States, which also allowed
these products. 70 Fed. Reg. at 491. The agency’s research
showed that about 96% of the “infectivity” of any given cow
was contained in certain tissues, and that the only examples
of blood transmission of BSE occurred in blood transfusions.

we noted in our preliminary injunction ruling, the
agency properly relied on studies rejecting the idea of trans-
mission through tallow, and we held that the district court
erred when it criticized the “gaps” in the Canadian feed ban.
See R-CALF 

. R-CALF does not offer any
evidence or arguments to support a different result at this
phase. In light of the science available at the time, the agen-
cy’s partial reliance on the feed ban was justified.7

   Finally, R-CALF argues that the agency showed its own
lack of confidence in the feed ban when it suspended the part
of the Final Rule allowing meat products from cattle over 30
months old. This argument fails as well. While the Final Rule
prohibited importing cattle over 30 months of age, see 9
C.F.R. §§ 93.436 (a)(1), (b)(1), and banned the use of speci-
fied risk materials from cows over 30 months of age, see 9
C.F.R. 310.22 (a)(1), it allowed meat products from cattle of
any age, see 70 Fed. Reg. at 494. The Final Rule, however,
stated that international guidelines recommended allowing
meat from cattle of any age as long as there were measures
  7
   For these reasons, we also reject R-CALF’s related argument that the
agency assumed that cattle under 30 months old would not be infected
with BSE because this claim is an implicit attack on assumptions about the
feed ban.
10716              RANCHERS CATTLEMEN v. USDA
in place to segregate highly infective tissues from the nervous
system. Id.8

  Two months after issuing the Final Rule, the agency
decided to suspend this part of the rule and continue to ban
beef derived from older cattle. See 70 Fed. Reg. 12,112. It
essentially left in place the pre-2004 practice of allowing in
meat from cattle under 30 months of age. See 70 Fed. Reg. at
536.

   R-CALF argues that this change of heart shows the agen-
cy’s “lack of confidence” in its initial assumptions about the
effectiveness of the feed ban, the ban on specified risk materi-
als, and the BSE incubation period. To be considered by the
courts, however, this evidence would have to show a “lack of
reasons” for the parts of the rule that are currently being chal-
lenged, rather than a subsequent “lack of confidence” in them.
R-CALF has failed to make a connection between the uncer-
tainty about this provision and the lack of justification for any
other provisions of the Final Rule.

   [12] 3) Blood Transmission: R-CALF claims that the
agency incorrectly assumed that the ban on “specified risk
materials” in products for human consumption would elimi-
nate the risk of BSE in spite of information that BSE can also
be transmitted by blood that affects other tissues. R-CALF
points out that the agency continues to ban fetal bovine
serum, see 70 Fed. Reg. at 502-03, but, it argues, inconsis-
tently permits the use of tallow in cattle feed, see 

01.
   8
     It did not provide a citation to any provision from the World Organiza-
tion for Animal Health guidelines, and the current guidelines do not
appear to “recommend” allowing beef imports from older cattle. See OIE
Terrestrial Animal Health Code, 2.3.13.10 (stating that meat products may
be imported from cattle that were born after the imposition of an
appropriate feed ban). See http://www.oie.int/downld/SC/2007/
en_chapitre_2.3.13. pdf.
                RANCHERS CATTLEMEN v. USDA                10717
   [13] The agency’s commentary in the Final Rule explains
that fetal bovine serum “might pose a risk for livestock if used
in certain applications such as bovine vaccine production or
bovine embryo transfer, or for other products brought into
direct exposure with ruminants.” 70 Fed. Reg. at 502. In R-
CALF II, we noted the special risk posed by the serum
because it is injected directly into the bloodstream. R-CALF

. The agency’s ban on fetal bovine serum
represents caution in the face of unknown risk. It does not
imply a more general finding of risk from feed products that
may have come into contact with cattle blood. As the agency
explains in the Final Rule, cattle blood only appears to pose
a risk when it is directly transfused into other cattle. See 70
Fed. Reg. at 491. Regarding the agency’s decision to allow
imports of tallow, the agency was entitled to follow interna-
tional standards and previous practices requiring that the tal-
low be protein-free and accompanied by certification. See 

In light of our previous endorsement of the feed ban,
we find that the agency has justified its different treatment of
tallow, fetal bovine serum, and cattle feed.

   R-CALF, in a similar vein, argues that the agency’s subse-
quent rule prohibiting imports of pregnant cattle shows that it
has since come to recognize the possibility of other types of
BSE transmission. See Bovine Spongiform Encephalopathy;
Minimal-Risk Regions and Importation of Commodities;
Technical Amendments, 71 Fed. Reg. 12,994 (Mar. 14, 2006).
However, we previously adopted the agency’s interpretation
that the Final Rule, even before the amendments, banned
breeding cattle. See R-CALF 

. Moreover,
the amendments specifically state that the agency is merely
clarifying the Final Rule. 71 Fed. Reg. at 12,994. R-CALF is
therefore incorrect that the Final Rule did not ban breeding
cattle.

  [14] The agency fully considered the possibility of other
types of BSE transmission and gave reasons for banning some
products and not others. Its analysis satisfies our review.
10718              RANCHERS CATTLEMEN v. USDA
   [15] 4) Ban on Specified Risk Materials. R-CALF argues
that more recent science shows that the ban on these cattle
parts will be less effective than the agency assumed. We pre-
viously endorsed the agency’s reliance on this ban because its
decision was based on the Harvard-Tuskegee Study. See R-
CALF 

.9 As R-CALF’s summary judg-
ment motion points out, the study’s authors have since revised
their certainty about the ban from 95% to 80%. The agency
has also acknowledged the scientists’ downward adjustment.
See Substances Prohibited From Use in Animal Food or
Feed, 70 Fed. Reg. 58,570, 58,587 (Oct. 6. 2005). This post-
decisional revision, however, does not show that the agency,
at the time it made its decision on the Final Rule, failed to
consider relevant factors or rested its decision on completely
baseless assumptions.10

   5) Other Arguments: R-CALF argues that the agency
assumed that non-ambulatory cattle (who are more likely to
have BSE) will not be slaughtered for human consumption,
but the agency stated in the Final Rule that Canada does not
allow non-ambulatory cattle to be slaughtered for export, 70
Fed. Reg. at 491, and R-CALF offers no reason to distrust that
statement.

   R-CALF also argues that the agency relied on the Harvard-
Tuskegee study’s findings of low risk without considering the
risk of errors and mislabeling. We find that the agency con-
sidered these risks and found them covered by existing regu-
lation and monitoring by the USDA. See 70 Fed. Reg. at 499.
    9
      R-CALF has also submitted a declaration from Dr. Stanley Prusiner,
who discovered the “prions” that cause BSE. Dr. Prusiner makes several
conclusions that run counter to the findings of the Harvard-Tuskegee study
but, as in Asarco, this declaration serves only to attack the merit of the
agency’s decision and does little to suggest flaws in the process leading
up to that decision.
    10
       We previously held that the agency’s reliance on this study gave it a
“firm basis” for its assumptions that R-CALF’s speculative arguments did
little to undermine. See R-CALF 

.
                RANCHERS CATTLEMEN v. USDA               10719
   Finally, on summary judgment, R-CALF contends that,
overall, the agency’s actions were contrary to the purposes of
the Animal Disease Risk Assessment, Prevention and Control
Act of 2001, Pub. L. No. 107-9, 115 Stat. 11, which requires
the Secretary of Agriculture to submit a report to Congress on
the USDA’s plans to research and monitor BSE and gauge the
effectiveness of its prevention measures. This argument was
not pled in the complaint, and in any event is unavailing. The
Act merely requires a report on these factors, and the USDA
continues to provide these reports.

   [16] Therefore, under the APA standard of review, none of
the claims as stated in R-CALF’s complaint warrant remand
to the agency.

                         Conclusion

   [17] Having reviewed the merits of this case, we conclude
that the agency considered the relevant factors and articulated
a rational connection between the facts found and its decision
to designate Canada a minimal-risk country. R-CALF’s extra-
record evidence has failed to convince us that the agency’s
review was unauthorized, incomplete, or otherwise improper.
The district court’s order granting summary judgment to the
USDA is therefore

  AFFIRMED.