Opinion for the Court filed by Circuit Judge STEPHEN F. WILLIAMS.
STEPHEN F. WILLIAMS, Circuit Judge:ICORE, Inc., is a consulting firm representing a consortium of local telephone companies. It and several such companies bring this petition for review to challenge a decision of the Federal Communications Commission as to how such phone companies are compensated for the interconnections they provide between their customers and providers of interstate long-distance telephone service.
The charges are based in one way or another on estimates of the local phone companies’ costs of providing the service. Some companies — known as “cost companies” — make a detailed study of the actual costs. The rest — known as “average schedule companies” — avoid the expense of such a study by using an “average schedule”, compiled from data designed to show the cost structure of a hypothetical local carrier. Most smaller companies follow the second path.
Until the early 1980s, both types of companies used parallel methods for computing fixed costs. Cost companies used a concept called “subscriber plant factor” (“SPF”); average schedule companies used a ratio known as the “ARPM method”.' The details of the schemes do not matter for our purposes; both were flawed in that they used volume-sensitive measures to estimate costs that did not vary with volume.
In 1982 the FCC became concerned that the SPF method allowed cost companies to recover more than the real costs of their interstate connection service and directed them to reduce the proportion of fixed costs allocated to these calls. This partial reform of course created a disparity between the treatment of cost companies and average schedule companies. The Commission in 1983 started to correct the disparity by adopting a new rule, 47 CFR § 69.606(a) (1991), which in effect required that average schedule companies be compensated under principles paralleling those for cost companies. This rule has not been and is not here challenged.
In 1985 the National Exchange Carrier Association (“NECA”), an association of local phone companies established by the Commission to administer the average schedule process, proposed changes designed to bring the method of calculating average schedule companies’ costs into line with the method for cost companies. The proposal called for replacing the prior formula with a fixed payment of $8.06 per access line. As the new method would cause substantial revenue changes for many companies, NECA proposed a gradual transition for those that stood to lose.
For most such companies, the reduction under NECA’s plan would start slowly. The then-current compensation would be reduced $1.25 per phone line per year for four years, after which revenue would drop to $8.06 per phone line. For others, however, NECA proposed a so-called “flash cut” — an immediate cut in revenues, with the reduced figure then serving as the starting point for the annual $1.25 per line step transition. As the cost companies were subject to a rule under which they could recover no more than 85% of their non-traffic-sensitive (“NTS”) costs out of interstate settlements, NECA aimed the “flash cut” at those companies which, it estimated, were recovering more than 85% of their NTS costs from interstate service under the prior, traffic-sensitive measure. For example, one ICORE client, petitioner Northwest Iowa, was collecting more than $800 a month per phone line under the old method; obviously a cut of $1.25 per line for four years would not bring it much of the way towards the $8.06 per line representing parity with cost companies.
The 85% criterion could not be applied with complete accuracy; the whole problem with average schedule companies is that their actual costs are unknown, because the costs of computation are high in relation to the stakes. To identify companies likely to be receiving more than an 85% recovery, NECA carried out a statistical study. On the basis of the study’s conclusion that carriers handling more than 15.9 *19interstate messages per phone line per month were likely to be over-recovering, NECA proposed that they be subject to the flash cut.
The Commission approved the NECA proposal in 1986. MTS and WATS Market Structure: Average Schedule Companies, No. 78-72, Phase 1,103 F.C.C.2d 1017, 1023 (1986). The ICORE companies challenged the new rule in this court. We found that the Commission had failed “to demonstrate a rational basis” for its adoption of NECA’s proposal. City of Brookings Municipal Telephone Co. v. FCC, 822 F.2d 1153, 1171 (D.C.Cir.1987). This finding covered both the basic NECA proposal and the flash cut. As to the latter we complained that the Commission had offered no persuasive data to show that its 15.9 messages per phone line cut-off point closely approximated the line of 85% NTS recovery. Id. Although we remanded the ease to the Commission, we declined to vacate the revised formula, expressing diffidence about our grasp of the “technical intricacies” involved and saying that such an order would disrupt the settlement process and cause hardship to many companies. Id. at 1171-72.
During the remand proceedings the FCC Common Carrier Bureau “directed 35 questions to NECA concerning the data, statistical tests, methodologies, and safeguards that NECA had employed in preparing its proposed revisions”, and invited public comment on NECA’s answers. MTS and WATS Market Structure: Average Schedule Companies, No. 78-72, Phase 1 (“1991 Order”), 6 FCC Red. 6608, 6609 (1991). After reviewing the augmented record, the FCC reaffirmed its adoption of the NECA proposal. Id. at 6614. It found the 1986 schedule and transition plan reasonable, soundly derived, superior to the old method, and more in accord with FCC rules. Id. at 6609-11. The Commission also ruled that it had properly applied the NECA schedule during the period of remand from this court and that it should not change that application retroactively. Id. at 6613-14.
ICORE and the other petitioners no longer challenge the Commission’s basic revisions in the treatment of average schedule companies, but challenge only the flash cut. We reject all of their claims.
Rationality of the flash cut
When this court originally rejected the Commission’s flash cut proposal, it held merely that the FCC had failed to provide data supporting its assertion as to the likely accuracy of the 15.9 messages per phone line cut-off point, or to show that it had “resolved this dispute in a reasoned fashion.” City of Brookings, 822 F.2d at 1171. The court particularly noted the lack of support for NECA’s claim that the cut-off correctly classified 97% of the average cost companies. Id. Petitioners now claim that the Commission has stuck to the flash cut rule “without citing any new data or providing any additional explanation that supported a correlation between 15.9 messages and an 85% recovery ratio.” Petitioners’ Brief at 20 (emphasis in original). The claim is sharply contradicted by the record.
Two of the questions addressed to NECA (questions 1 and 2) asked it to provide and discuss all data, calculations, mathematical functions and statistical tests used to arrive at the 15.9 message per line cut-off. Joint Appendix (“J.A.”) 206. NECA’s response discussed the data sample, the formulas for calculating company costs, the translation of messages per line to 85% of NTS costs, and set out a detailed step-by-step analysis of the statistical methodology. J.A. 212-28. Petitioners’ brief makes no effort to identify flaws in that methodology.
Further, in response to question 5, asking for all studies “that confirm or deny the validity of the 15.9 message ‘flash-cut’ criterion”, J.A. 206, NECA explained the origins of the 97% accuracy claim. This consisted of a statistical study appearing to establish that the 15.9 messages per line criterion was correct for 458 out of 471 companies, i.e., 97%, or an error rate of 3%. J.A. 244-45.
Petitioners claim (vaguely in their opening brief, more seriously in their reply *20brief)1 that the Commission’s 97% accuracy argument misses the point. For them, the point is that for companies with more than 15.9 messages per line — the ones that would be subject to the flash cut — the error rate was far higher: three out of six or 50%.2 There are a number of answers to this. First, without offering a serious critique of NECA’s carefully explained methodology, petitioners are in a weak position to raise the claim at all. Second, petitioners do not explain why the error rate in the group above 15.9 messages should be the sole focus of inquiry. Reducing the error rate for false positives (companies wrongly subject to the flash cut) would necessarily increase the false negatives (companies wrongly spared the flash cut). Petitioners offer no principled argument for a shift in that direction. The flash cut’s purpose is to accelerate the most excessive over-recoverers toward the now undisputed $8.06 per line settlement rate. Petitioners never explain why the Commission should tilt against application of the flash cut, i.e., in favor of delaying a basically corrective process.
Third, it is far from clear that in fact half the companies with more than 15.9 messages per line recovered materially less than 85% of NTS costs. The 1985 “scatter-gram” shows six companies with more than 15.9 messages per line (in fact there were seven including Northwest Iowa). J.A. at 249. Of the three companies with estimated recoveries of less than 85% of NTS costs, one with 21.47 messages per line was recovering 84%, and another with 15.95 messages per line was recovering 78%. Id. The only truly problematic classification was of a company with 21.91 messages per line with an estimated recovery of only 65% of NTS costs. Id. But NECA went on to replicate the analysis with revised data' from 1987. J.A. at 246-47, 252-53. This yielded a lower cut-off point (15.3 messages per lines). Seven companies exceeded the 15.3 level; five of them had estimated recoveries exceeding 85%, and the remaining two very nearly so. See id. at 247, 252. These statistical results hardly indicate an irrational cut-off point.
Moreover, in response to questions 3 and 4, NECA identified 13 companies originally subject to the flash cut (minus Northwest Iowa, which had been excluded from the statistical analysis because it was an “outlier”, see J.A. at 219) and discussed whether the 15.9 messages per line cut-off accurately classified those companies as recovering in excess of 85% of NTS costs.3 J.A. 229-43. Of the five companies for which NECA had financial data, it found three to have been accurately classified, and the other two “close to 85% recovery.” J.A. at 230. Petitioners do not attack these findings, or claim that the two companies merely “close” were far enough off to suggest a problem with NECA’s calculation.
NECA also provided “company specific reviews” of the 13 companies. J.A. 230-38. The investigation revealed unusually high messages per phone line (typically because the firm operated right next to a state line), but no “extraordinary cost factors.” J.A. at 560. NECA estimated that each of the 13 companies was receiving reimbursement per line in excess of 85% cost recovery. *21J.A. 232-38. As the initial problem with average schedule companies is lack of data, obviously the figures are estimates, but petitioners make no real effort to show inaccuracy. Only one company protested its classification, Respondent's Brief at 32, and seven of the 13 companies were exempted after “true-up data” showed that their volume fell below 15.9 messages per phone line.
Of course, as the FCC recognized explicitly, the absence of cost data imposed “unavoidable limitations on NECA’s ability to implement an absolutely accurate surrogate” for 85% recovery. 1991 Order, 6 FCC Red. at 6613. See also J.A. at 553 (similar recognition by NECA). But — especially in light of petitioners’ failure to pinpoint a single methodological flaw— NECA's detailed responses provide a substantial basis for the Commission decision.
We have compared the record here with the one that produced the remand in City of Brookings. There the joint appendix contained only a few pages of assertion that the transition plan was reasonable, City of Brookings Joint Appendix at 38-40, 281-82, plus three unexplained statistical charts leading up to a scattergram relating cost recovery to messages per phone line. Id. at 945-50. The record here, in contrast, contains substantial and detailed material that was considered by the Commission in reaching its decision to affirm the flash-cut. See, e.g., J.A. 213-53, 546-61.
In defending the flash cut, the Commission has relied in part on the fact that it gave any affected company the option of seeking to justify exemption by doing a cost study. Petitioners argue that such a safety valve cannot save an irrational rule, citing Alltel Corp. v. FCC, 838 F.2d 551, 561 (D.C.Cir.1988). There we rejected a rule that would have converted all average schedule companies affiliated with cost companies into cost companies, saying that affiliation had no relevance to the company’s proper status. Id. at 558. Not surprisingly, we rejected a Commission argument that the rule was saved by an escape hatch provision exempting companies with revenues below a specified level; we likened the Commission’s effort to a rule that denied all brown-haired people broadcasting licenses but “softened” the impact by limiting the rule to brown-haired people born in odd-numbered years. Id. at 558-59. At the same time, we recognized that a rule with a rational basis, yet otherwise impermissibly broad, “ ‘can be saved by the “safety valve” of waiver or exemption procedures.’ ” Id. at 561 (quoting KCST-TV, Inc. v. FCC, 699 F.2d 1185, 1200 (D.C.Cir. 1983) (Scalia, J., dissenting)). On this record it is far from clear that the flash cut would otherwise be impermissibly broad, but, to the extent that it may be, it is amply shored up by the allowance of exemption through a cost study.
Retroactive application of the flash cut
Petitioners claim that the 1991 Order, by failing to order compensation reversing the effects of the flash cut for the period between the original order and that of 1991, is in effect a retroactive rulemaking in violation of the Supreme Court’s holding in Bowen v. Georgetown University Hospital, 488 U.S. 204, 109 S.Ct. 468, 102 L.Ed.2d 493 (1988). The Commission does not question the understanding that Georgetown University Hospital bars retroactive rulemaking in the absence of congressional assent, see id. at 208, 109 S.Ct. at 471, nor does it claim that there has been such assent. The issue is whether forbidden retroactive rulemaking has occurred when (1) a rule is remanded to an agency for want of adequate reasoning and explanation, but not vacated, (2) the agency on remand provides the necessary support, and (3) the agency applies the rule during the remand period.
It seems apparent that the Court does not view “curative” rules as permissible under a statutory scheme that generally withholds authority for retroactive rule-making. In Georgetown University Hospital, a lower court had “struck down” a 1981 rule on calculation of certain Medicare reimbursements for want of any notice- and-comment. Id. at 206, 109 S.Ct. at 470. The agency set about proper promulgation of the rule (meanwhile applying the prior rule adopted in 1979), and, on re-adopting *22the 1981 rule, made it applicable retroactively to its original date. After reviewing the statutory scheme and finding no authority in the agency to promulgate retroactive rules on the subject, the Court struck down the rule’s applicability to periods prior to its proper issuance. The Court then noted the agency’s argument that “curative” retroactive rules could be justifiable, in view of the agency’s special need and the weakened reliance interests of affected parties, even in a context where retroactive rules generally were not. Id. at 215,109 S.Ct. at 475. The Court responded that, whatever the validity of the contention “in other contexts”, the statutory scheme denying the agency “authority to promulgate retroactive cost-limit rules”, made it unnecessary to address the claim. Id.4 The Court thus declined to acknowledge an exception for “curative” rules.
Our own decision below in Georgetown University Hospital is more explicit in refusing to carve out a general exception for curative rulemakings, saying that to permit retroactive rulemakings “to remedy a procedural defect in a rule would ... make a mockery of the provisions of the [Administrative Procedure Act].” Georgetown University Hospital v. Bowen, 821 F.2d 750, 758 (D.C.Cir.1987).
Here, of course, in contrast to Georgetown University Hospital, the court considering the rule initially found it inappropriate to set the rule aside. The court’s decision on that point represented a careful consideration of the risk of disruption and of the likelihood that the rule was altogether sound at the core (see, e.g., the reference to the court’s diffidence about “understanding] the technical intricacies that loom so large in this case”, Brookings, 822 F.2d at 1171). Petitioners offer no reason why a rule so treated, and in fact applied during the entire interim period, should be treated the same as the rule initially “struck down” in Georgetown University Hospital. Petitioners cite no case employing Georgetown University Hospital to cancel the effect of rules deliberately left standing by a court pending a remand.
This reduces petitioners to two devices to undercut the Brookings panel’s choice of remedy. First they argue that the court did not, in fact, refuse to vacate the 1986 average schedule rules, contending that it left them in place only as to non-party companies. In support they cite the court’s expression of concern that altering settlements would cause economic hardship for non-party companies. Brookings, 822 F.2d at 1171. But economic hardship to non-parties was only one of the court's reasons for not vacating the rule. After alluding to the technical intricacies that the court did not “pretend to understand”, id., it went on to say:
We also forebear from ordering reassessment of settlement payments made under [the revised average schedule ... because such an] order at this late stage would disrupt the settlement process and would, among other things, cause economic hardship to many companies that are not parties to the petition for review.”
Id. at 1171-72 (emphasis added). The court then left to the Commission’s discretion how to accommodate the various interests during remand, clearly leaving the revised schedule in place as to all parties. Id. We suspect that if petitioners had thought for one moment that the Brookings judgment entitled them to reimbursement they would have promptly demanded payment.
Petitioners also claim that it would have been improper for this court not to vacate the rule once it found it arbitrary and capricious, evidently suggesting that we rewrite history to validate the conduct of the Brookings panel. Happily, vindication of that panel requires no such maneuver. This and other federal circuit courts have repeatedly found it appropriate to remand an agency action without vacating it. See, e.g., Commonwealth of Massachusetts v. NRC, 924 F.2d 311, 336 (D.C.Cir.1991) (“In appropriate cases, we will remand without *23vacating an agency’s order where the reason for the remand is a lack of reasoned decisionmaking.”); International Union, UMW v. FMSHA, 920 F.2d 960, 967 (D.C.Cir.1990) (stating considerations to be balanced in resolving the issue); Rodwa-y v. USDA, 514 F.2d 809, 817 (D.C.Cir.1975) (noting practical necessity for preserving invalidly issued regulation pending remand); Western Oil & Gas Ass’n v. EPA, 633 F.2d 803, 813 (9th Cir.1980) (similar). To the extent that petitioners claim these decisions are invalidated by Georgetown University Hospital, they confuse (1) the use of retroactive rulemaking to cure a gap created by judicial vacatur of a rule with (2) the judicial decision not to create such a gap. Nothing in Georgetown University Hospital speaks to the latter question. Nor do we share petitioners’ apparent assumption that Georgetown University Hospital’s strictures against filling a gap through retroactive rulemaking yield an inference in favor of courts exercising their remedial discretion in favor of creating gaps that as a result will be impossible to fill.
Accordingly, we find no invalid retroactive rulemaking in the 1991 Order.
The so-called second flash cut
Petitioners argue that the Commission made a second flash cut in the revenues of average schedule companies in 1990, when it allowed the reduction of average schedule revenues to $8.06 per phone line to take effect as planned at the end of the four years of phased reduction. This supposed “second flash cut” is alleged to have been wrought in violation of the notice-and-comment requirements of the APA and to represent an abdication of the Commission’s statutory duty.
This claim is utter fantasy. The 1990 drop in revenues occurred when the transition rule expired by its express terms. Such expiration is not enactment of a new rule, so APA rulemaking requirements were completely inapplicable.
There is a grain of sand out of which petitioners try to grow their legal pearl. NECA observed in its 1985 proposed Modification of Average Schedules that “further study” was needed to determine what should be done at the end of the transition period for companies still receiving amounts higher than the new rule called for, J.A. at 129, and the FCC, in adopting the proposed modifications, referred to the future NECA study, J.A. at 15, 21. From these remarks, petitioners infer that the FCC either had an obligation to continue the transition payments, or else to invoke full notice-and-comment rulemaking procedures to justify its not continuing them.
Petitioner’s argument is baseless. Not modifying a rule is not the same as “formulating, amending, or repealing a rule”, the APA definition of “rule making”. 5 U.S.C. § 551(5). An agency statement in one rulemaking, that a pending study may generate need for another, neither initiates a second rulemaking nor cancels the timetable adopted in the first rulemaking. Finally, when the FCC in 1990 reviewed the promised NECA study and adopted its conclusion that no new transition rule was needed, it did not thereby adopt a new rule.
% % ss jfc ¡k
The petition for review is
Denied.
. Although we have often refused to address claims largely confined to the reply brief because the strategy tends to sandbag the respondent, see, e.g., Corson & Gruman Co. v. NLRB, 899 F.2d 47, 50 n. 4 (D.C.Cir.1990); Carducci v. Regan, 714 F.2d 171, 177 (D.C.Cir.1983), we give the petitioners the benefit of the doubt and consider their claim.
. The three-out-of-six calculation disregards Northwest Iowa, which had more than 15.9 messages per line and vast over-recovery of NTS costs.
. It is not entirely clear why the scattergram shows six companies with more than 15.9 messages per line, while the response to question 4 speaks of 13. It appears, however, that between the time the 13 companies were originally identified (July 10, 1985) and the time the scatter-gram was made (September 1985), "true-up” data became available that exempted seven of the thirteen companies by revealing fewer than 15.9 messages per line. J.A. at 231-38. The record is clear that seven companies were exempted, but is silent as to when, except to say that the exemptions occurred before June 1, 1986. J.A. at 231. We can only speculate that they account for the difference. In any case, this same information was available to petitioners, and still they did not attack the methodology-
. The Court saw no problem with duly promulgated prospective regulations that themselves authorized "retroactive corrective adjustments" in the calculation of Medicare reimbursement. See id. at 211-12, 109 S.Ct. at 473.