March 27, 2003
VIA FEDEX and E-MAIL (w/o Attachments)
1600 Clifton Road, N.E.
Atlanta, GA 30333
Information Quality Appeal
This letter is for the purpose of appealing CDC's denial of a request by the Animal Health Institute ("AHI") to correct certain information circulated or otherwise disseminated by CDC. This appeal is filed under Section 515 of Public Law 105-5541 and pursuant to the CDC "Guidelines for Ensuring the Quality of Information Disseminated to the Public" (the "CDC Information Quality Guidelines"2 The denial, dated March 13, 2003 and signed by Dr. Rima Khabbaza, essentially declined to make corrections to certain information alleged to be in error and circulated by CDC ("CDC Reply"). Please find attached a copy of the CDC Reply together with a copy of the original request. The reasons for this appeal are set forth in this letter and in the original request, as amended.
This letter is also for the purpose of requesting a meeting with CDC. The CDC Reply did not respond to AHI's March 12 request for a meeting. AHI renews its request for a meeting to discuss the nature of its complaint with the CDC official responsible for the appeal.
On December 6, 2002, as amended January 8, 2003 AHI ("Requester") filed a Request for Correction of Information ("Request") with CDC pursuant to the CDC Information Quality Guidelines. The Request sought the correction or deletion of information disseminated by the CDC that is in error, including generally information related to the extent, duration and causes of campylobacteriosis and resulting days of illness in the U.S., including fluoroquinolone-resistant campylobacteriosis and illness-days. More specifically, statements requested to be corrected included the following:
- Extended Illness: A person with a Campylobacter infection that is resistant to fluoroquinolones will have an extended illness or two days of diarrhea more than a person who is infected with a Campylobacter that is susceptible to fluoroquinolones.
- Rapid Rise In Resistance Attributable To Poultry: The data show that the percentage of fluoroquinolone resistant Campylobacterisolated from people in the United States is dramatically rising and that such rise is attributable to the use of fluoroquinolones in poultry.
- Occurrence of Campylobacteriosis in U.S.: There are approximately 2.4 million cases of campylobacteriosis in humans annually in the United States.
Extensive scientific documentation was submitted to support the Request, including expert analysis of data obtained from CDC. The CDC Reply largely does not respond specifically to the data presented in the Request. It appears to continue to discount data and analyses that are contrary to CDC's previously stated views.
Requestor believes the following additional facts are relevant to the Request and to the CDC Reply and further demonstrate why Requestor's position is correct and CDC is in error. ("Comment" below refers to the heading in the CDC Reply.)
AHI originally objected to the CDC and Smith conclusions that fluoroquinolone-resistance causes extended days of illness among domestically-acquired campylobacteriosis cases on the grounds that they attribute excess days of diarrhea to domestically-acquired campylobacteriosis cases, even though domestically-acquired campylobacteriosis cases do not in fact have excess days of illness. The Smithet al. and CDC conclusions to which AHI objected attribute to domestically-acquired cases excess days of illness that, in fact, occur solely among foreign travel-associated cases. They do this even though the Smith et al. and CDC data clearly show that no such excess occurs among domestically-acquired cases. This point has been made in published sources and peer-reviewed journal articles such as Cox L.A., 2002. Reexamining the causes of campylobacteriosis. International Journal of Infectious Diseases. Dec. 6. Supplement 3:S26-S36.
Data submitted with the Request, specifically including expert analyses of the CDC and Smith et al. data, substantiate that persons infected with a domestically-acquired fluoroquinolone-resistant Campylobacter have not been shown by Smith et al. or by CDC analyses to have longer average illness durations (and, indeed, do not have longer average illness durations) than persons infected with a fluoroquinolone-susceptible Campylobacter for the following reasons, among others. In both data sets:
- Foreign travel is statistically significantly associated with excess days of illness among campylobacteriosis cases.
- Foreign travel is statistically significantly associated with fluoroquinolone-resistance in campylobacteriosis cases
- There is no association between fluoroquinolone-resistance and excess days of illness among domestically acquired cases. To the contrary, the association between them in the Smith et al. data set and also in the CDC data set is explained entirely by confounding due to foreign travel.
- Foreign travel thus fully meets the epidemiological definition of being an important confounder for the relation between fluoroquinolone-resistance and excess days of illness, in both data sets.
- Correct causal conclusions from case-control data sets require that the effects of important confounders be removed.
- Neither the Smith analysis nor the CDC analyses removed the effects of foreign travel before drawing the conclusion that fluoroquinolone-resistance is causally associated with excess days of illness among domestically acquired campylobacteriosis cases.
- Therefore the conclusions asserted in various forums by CDC, that fluoroquinolone resistance causes and/or is associated with excess days of illness among domestically-acquired campylobacteriosis cases, are not justified by the Smith et al. data or the CDC data. We request that CDC cease disseminating these unjustified conclusions.
The CDC Reply does not respond to the content of this objection. It does not take issue with any of (a)-(g) above, as we understand it. Instead, the letter re-states the original conclusions from the Smith and CDC analyses, in effect citing them as support for themselves. But it is these very conclusions that AHI objected to, on the grounds just outlined: that they do not follow from the data that they purport to be based on, due to flawed analysis (failure to correct for confounding). They cannot appropriately be used as support for themselves.
AHI therefore appeals for reconsideration of our Request. Specifically, AHI requests that CDC either agree with each specific reason contained in items (a)-(g) above, or identify specific disagreements with one or more of these items.
The CDC Reply also states that "Because there are no demonstrated differences in the virulence or clinical impact of primary Campylobacterinfection among countries, there is no scientific reason to exclude travelers from the dataset or analysis." Yet, CDC's assertion notwithstanding, demonstrated differences in features of Campylobacter enteritis among countries are well documented and widely known (see e.g., Coker AO, Isokpehi RD, Thomas BN, Amisu KO, Obi CL., Human campylobacteriosis in developing countries. Emerg Infect Dis. 2002 Mar; 8(3):237-44.) For the Smith and CDC data sets specifically, highly statistically significant differences in clinical impact (duration of diarrhea) have been conjectured about in dissertations supervised by CDC members and have subsequently been demonstrated by statistical analysis in the scientific literature (e.g., Cox, 2002 op cit.).
Thus, we believe that CDC's premise that "there are no demonstrated differences in the virulence or clinical impact of primary Campylobacterinfection among countries" cannot be supported and that the conclusion "there is no scientific reason to exclude travelers from the dataset or analysis" is incorrect. A compelling statistical reason to remove them is that they have very different (longer) duration of illness (as well as many other distinguishing characteristics, such as much higher probability of fluoroquinolone-resistance) than domestically-acquired cases. Travel-associated cases and domestically acquired cases are two very statistically different populations, as is clearly shown by CDC's own data and in other published studies (e.g., The Campylobacter Sentinel Surveillance Scheme Collaborators' "Ciprofloxacin resistance inCampylobacter jejuni: case-case analysis as a tool for elucidating risks at home and abroad." J Antimicrob Chemother. 2002 Oct; 50(4): 561-8.)
Further, the CDC Reply states that "Reviewing original data of peer-reviewed journal articles before citation of these articles is not standard practice." Notwithstanding that it may not be standard practice to review original data of peer-reviewed articles, it surely cannot be and is not standard practice at CDC to quote from, cite or otherwise rely on data that has been demonstrated not to support the article's conclusions.
AHI asks that CDC acknowledge that (a) confounding was not controlled for in the Smith and CDC analyses that claim to establish a link between fluoroquinolone-resistance and excess days of illness; and (b) confounding provides a non-causal explanation for the association between them. Peer-reviewed publications based on a review of the original data already exist showing that there is no causal relation or statistical association between fluoroquinolone-resistance and excess days of illness once confounding effects are removed (Cox, 2002, op cit.). Thus, if CDC does not wish to review its own original data or that of Smith, it can rely on peer-reviewed journal articles instead. AHI now requests CDC, by whatever route it prefers, to confirm and acknowledge that the association they have sometimes asserted to exist between fluoroquinolone-resistance and excess days of illness does not exist among domestically acquired cases. We also request CDC to acknowledge that the failure to correct for confounding by foreign travel (and perhaps income, consumption of contaminated water, insurance coverage, and other potential confounders) exemplifies an analytic approach that does not "appear to be appropriate and support the conclusions of the authors", as stated in the CDC Reply. Although citations to introductory epidemiology texts can easily be provided if desired, we presume that CDC is already very well familiar with the necessity of controlling for confounders before assigning causal interpretations to statistical associations. We ask them to accept and support this methodological necessity in the specific case of fluoroquinolone-resistance and illness-days for campylobacteriosis and to cease to promulgate interpretations and conclusions about campylobacteriosis that violate it.
As an initial matter, I wish to point out that CDC's statement on page 3, paragraph 4 in the CDC Reply is in error and should be corrected. I believe that if CDC examines its FOIA log it will readily confirm that CDC has provided no data to AHI in reply to requests made under the Freedom of Information Act, or otherwise. Quite simply AHI has made no requests to CDC for any such data. Similarly paragraph 2 on page 4 is also in error to the extent it is a representation that any data on the 1989-1990 survey was requested by, or provided by CDC to AHI.
The CDC Reply asserts that "There have been consistent sampling and testing procedures used for Campylobacter surveillance through NARMS-EB." This assertion does not dispute or respond to the many inconsistencies that AHI has referred to for sampling and testing procedures across FoodNet sites and over time, many of which can be found in Ginevan (2002) and in DeGroot's testimony (Request, as amended, Exhibit 6). Ginevan's evaluation of the NARMS-EB was provided to CDC in conjunction with, and discussed at, the NARMS 2000 meeting in November 2002.
The CDC Reply also does not address the facts that high levels of resistance were already found in chickens and humans in the 1970s and early eighties; that resistance rates cited in the peer-reviewed literature prior to introduction of Baytril are in many instances higher than rates cited after the introduction of Baytril; or that the NARMS sample is not representative of the general US population (see Ginevan 2002, op cit; and (Svedhem A, Kaijser B, Sjogren E. Antimicrobial susceptibility of Campylobacter jejuni isolated from humans with diarrhoea and from healthy chickens. J Antimicrob Chemother. 1981 Mar;7(3):301-5.). Most of all, it does not justify the causal attribution of any temporal trends in resistance risks to Baytril use in poultry, as opposed to other sources such as increasingly widespread contamination of water with ciprofloxacin (see e.g., http://www.mindfully.org/Water/Wastewater-Contaminants-US-StreamsMar02…) and the decreasing fraction of total campylobacteriosis cases caused by chicken, which is less likely than other sources to convey resistance.
In work by the FDA's CVM, publicly endorsed and cited by the CDC, based on data collected prior to the introduction of Baytril, the fluoroquinolone-resistance fraction attributed to poultry among all domestically acquired, non-treatment cases has been estimated as being approximately 57%. Based on data collected by CDC after the introduction of Baytril, the resistance fraction attributed to poultry among all domestically acquired, non-treatment cases is less than half of this earlier number (see e.g., Friedman et al. 2000 and testimony of Dr. Fred Angulo (request, as amended, Exhibit 17)). We believe that this large decline in resistance attributed to poultry is inconsistent with and is misrepresented by CDC's assertions of a "Rapid rise in resistance attributable to poultry."
The Kassenborg analysis cited in the CDC Reply appears to make many mistakes and undocumented assumptions (e.g., about selection of variables to consider), based on our own analysis of the same data, as detailed in the supporting documentation filed with the Request, and we have been unable to replicate its findings. Moreover, it does not perform any analysis of causes (or of causal attribution). We therefore request that CDC stop citing this analysis as support for causal conclusions.
The other observations in CDC's letter (e.g., findings of resistant Campylobacter on chickens at retail) also do not address the issue of causation or of whether the ratio of resistant Campylobacter in chickens to resistant Campylobacter from other sources (such as non-poultry meats, contaminated water, fish, vegetables, pets, restaurant non-poultry foods, etc.) has been increasing or decreasing since Baytril was introduced. Thus, logically, they cannot serve as support for assertions of a "Rapid rise in resistance attributable to poultry", since the fraction attributable to poultry necessarily requires considering the fractions attributable to competing sources of risk that CDC's letter does not consider. Additionally, there is a significant question about the relevance of retail studies to disease given that the Friedman et al. data (as well as independent data of Effler et al. and others) all show that, to the extent chicken is a potential source of campylobacteriosis, the risk arises not from chicken purchased at retail and cooked at home, but rather from chicken consumed outside the home.
The CDC Reply asserts that "in many countries, consumption of undercooked poultry is the most often identified specific risk factor forCampylobacter infections." We ask CDC to acknowledge that in this context, "risk factor" does not mean "cause"; that undercooked poultry is often a marker for poor cooking and food preparation hygiene (e.g., "undercooked potatoes" or "undercooked non-poultry meats" might be equally strong risk factors, had they been asked about); that, contrary to CDC's assertion in the letter, this finding does not support a link between resistant infections and poultry; and that, in truth, undercooked chicken accounts for less than 1% as many cases of campylobacteriosis (resistant or not) as contaminated water (in many other countries and specifically in the US based on outbreak data of Friedman et al., in Table 4, p. 127 of Nachamkin and Blaser, 2000.)
More generally, we request CDC to reconsider its response to Comment 2 and to acknowledge that it has no scientifically credible data supporting the claim that the rate of resistance caused by poultry has increased at all since the introduction of Baytril, let alone increased "rapidly". If CDC will not acknowledge this, we ask that CDC state reasons for disagreement that concretely address the change in the fraction of resistant campylobacteriosis cases caused by chicken, rather than citing for support discussion of other topics (e.g., consumption of undercooked chicken in foreign countries) not relevant to the issue of whether there has been "a rapid rise in resistance attributable to poultry consumption in the United States."
The 27% decline in U.S. campylobacteriosis rates from 1996 through 2000 was published by CDC in MMWR Weekly in April 2002. CDC announced in MMWR Weekly in March 2000 a 19% reduction in campylobacteriosis between 1998 -1999 in the original five FoodNet sites. The 2000 MMWR Weekly publication by CDC is difficult to reconcile with the CDC Reply that "a declining trend in Campylobacter ….was first detected by FoodNet in 2002," notwithstanding the parenthetical "noting a significant decline since 1996."
However, even assuming that FoodNet first detected the trend in 2002, dissemination after April 2002 by CDC of the 2.4 million figure, without qualification, represents at best a partial truth and possibly a distortion of the U.S. incidence of campylobacteriosis. If such a statement were made in advertising, the Federal Trade Commission might find it literally true but a misrepresentation and misleading nevertheless. The support submitted with the Request substantiates instances where CDC personnel have continued to use and otherwise disseminate the 2.4 million figure, unqualified, after March 2000 as well as after April 2002. We specifically request that CDC correct such statements and not use the 2.4 million figure in the future. If they do use it, we request that the 2.4 million figure only be used in conjunction with a clear statement that it is based on an estimate made in 1999, using older data, and that a decreasing trend in campylobacteriosis in the U.S. has been detected in FoodNet of 27% between 1996 and 2000.
As noted in the CDC Reply "four documents … were from the mass media." Those documents, however, contain statements attributed to CDC, based on written or oral statements made by CDC. Since the statements attributed to CDC contain the same data otherwise disseminated by CDC it is unlikely that CDC was misquoted. For example, the Wall Street Journal attributes to CDC's Dr. Robert Tauxe statements to the effect that the rate of Ciprofloxacin resistance in Campylobacter has gone from 15% to 19% and that "that's going from being not a detectable problem at all to being a real problem." (Exhibit 13, Request.) Similarly, Food Chemical News quotes CDC's Alicia Anderson as citing a rise in resistance from 13% to 18%. (Exhibit 20, Request.)
AHI believes that it makes a good case that the CDC Information Guidelines cover quotations by CDC personnel appearing in the public media. In the current instance it is the oral statements made by CDC, such as those referenced above, that need to be corrected. The CDC Information Quality Guidelines seem to anticipate this situation since the guidelines clearly cover oral statements, including interviews with and presentations by CDC personnel. (CDC Information Guidelines, II. A., IV. D.) Such an interpretation appears correct since the articles cited by AHI in the Request merely memorialize the incorrect statements made by CDC personnel. Clearly it would be of little benefit to those aggrieved by inaccurate information disseminated by an agency to obtain correction of printed information disseminated by the agency, while at the same time not have the ability to have corrected the same inaccurate information, orally presented and faithfully recorded and disseminated by the media.
AHI appeals the denial of its Request and renews its request for correction of information, and for a meeting with CDC as more fully set forth in the Request.
Requester Contact Information
Name: Animal Health Institute
The Animal Health Institute is a trade organization representing the interest of developers and manufacturers of animal drugs and biological products.
1335 G Street, N.W.
Washington, D.C. 20005
Telephone number: (202) 637-2440
E-mail address: www.ahi.org
By: Kent D. McClure, General Counsel
Animal Health Institute
1335 G Street, N.W.
Washington, D.C. 20005
Counsel for AHI:
Robert B. Nicholas, Esq.
McDermott, Will & Emery
600 13th Street, N.W.
Washington, DC 20009
1Section 515 of the Treasury and General Government Appropriations Act for Fiscal Year 2001; P.L. 106-554, § 515 (2001)
2The CDC Guidelines may be found at http://www.hhs.gov/infoquality/.
Last Revised: August, 2004