Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Should the system of death certification be changed?
- Arpan Dutta (17 February 2003)
-
Was the result specific to gastrointestinal infections?
- Adam Jacobs (18 February 2003)
-
Mortality associated with foodborne bacterial gastrointestinal infections reexamined
- Louis A. Cox (19 February 2003)
-
Reply to Lewis A. Cox and Adam Jacobs
- Kåre Mølbak, Morten Helms and Pernille Vastrup (21 February 2003)
-
Re: matched cohorts can be useful
- Darren C Greenwood (4 March 2003)
-
About case selection and clinical data
- Sarah J. O’Brien, Roger A. Feldman (7 March 2003)
-
Re: About case selection and clinical data
- Kåre Mølbak, Morten Helms, Peter Gerner-Smidt (12 March 2003)
-
Over estimation of mortality rates in cases with foodborne bacterial infections
- Phillippa M Cumberland, Jenny Roberts and Laura Rodrigues (18 March 2003)
-
Re: Over estimation of mortality rates in cases with foodborne bacterial infections
- Kåre Mølbak, Morten Helms, Peter Gerner-Smidt (28 March 2003)
|
|
|||
|
Arpan Dutta, 4th Year Medical Student Warrington Hospital, Lovely Lane, Warrington, Cheshire WA5 1QG
Send response to journal:
|
Editor - Helm et al’s paper on mortality associated with foodborne bacterial gastrointestinal infections raised an interesting point with regard to mortality data.1 Mortality data from death certificates is used commonly by researchers, but there is published evidence to suggest it is often inaccurate. Estimates range between 28 and 80%. 2-5 Such errors would undoubtedly affect the validity of any study using mortality data and even Helms et al. did have problems in finding valid information on causes of death in some cases. A recent study suggested a system of feedback to improve accuracy,4 since the problem with mortality data is not a new phenomenon and not happening solely in the United Kingdom either.5 Perhaps death certificates should not be accepted if they are incomplete? Maybe we should be looking toward a computerised system of death certification? Either way if evidence-based medicine is all its cracked up to be, shouldn’t this have changed many moons ago? Arpan Dutta, 4th Year Medical Student, Warrington Hospital, Lovely
Lane, Warrington, Cheshire, WA5 1QG
1. Helms M, Vastrup P, Gerner-Smidt P, Molbak K. Short and long term mortality associated with foodborne bacterial gastrointestinal infections: registry based study BMJ 2003;326:357-360 2. Morton L, Omar R, Carroll S, Beirne M, Halliday D, Taylor KM. Incomplete and inaccurate death certification: the impact on research. Journal of Public Health Medicine 2000;22:133-137 3. Peach HG, Brumley DJ. Death certification by doctors in non- metropolitan Victoria. Australian Family Physician 1998;27:178-82 4. Maudsley G, Williams EM. Inaccuracy in death certification--where are we now? Journal of public health medicine 1996;18:59-66 5. Medical Services Study Group of the Royal College of Physicians of London. Death certification and epidemiological research. BMJ: 1978;2:1063-1066 Competing interests: None declared |
|||
|
|
|||
|
Adam Jacobs, Director Dianthus Medical Limited, London SW19 3TZ
Send response to journal:
|
Helms et al present convincing data, using an impressively large dataset, to show that patients with gastrointestinal infections have a higher mortality than control subjects in the year following the infection [1]. However, I am not convinced that they have shown that the effect is a specific result of a gastrointestinal infection. It could equally be due to a non-specific association between poor health and death, and arise because gastrointestinal infections are simply a marker for poor health. The authors’ attempt to control for this by using a comorbidity index seems rather crude. Their comorbidity index appears to capture information on only serious illness, and does not take account of less serious illness that could still represent a declining state of health. The lack of any comorbidity in the vast majority of patients, as shown in Table 1 of the paper, shows that the comorbidity index is unlikely to capture more subtle signs of poor health, unless it really is true that the majority of the patients were perfectly healthy other than having an isolated gastrointestinal infection. It is noteworthy that relative risks all fell after adjustment for this crude measure of comorbidity, suggesting that adjustment for a more sensitive measure of poor health might have reduced relative risks still further. Overall healthcare use is known to increase in the last year of life [2], and it is therefore not surprising that a higher proportion of those selected on the basis of having any illness are in their last year of life compared with randomly selected controls. My guess is that this study could be repeated using any of a great many illnesses in place of gastrointestinal infections, and the results would be the same. References 1. Helms M, Vastrup P, Gerner-Smidt P, Mølbak K. Short and long term mortality associated with foodborne bacterial gastrointestinal infections: registry based study. BMJ 2003:326;357–60 2. Stuart B, Coulson NE. Dynamic aspects of prescription drug use in an elderly population. Health Serv Res 1993;28:237–264 Competing interests: None declared |
|||
|
|
|||
|
Louis A. Cox, President, Cox Associates; Clinical Professor, University of Colorado Health Sciences Center Cox Associates, 503 Franklin Street, Denver, CO, 80218, USA
Send response to journal:
|
Helms et al. have recently made the startling claim that food-borne pathogens including Campylobacter, usually associated with self-limiting cases of diarrhea, are also associated with far more ominous outcomes: a significant increase in both short-term and long-term mortality rates [1]. Since this conflicts with the conventional clinical wisdom that, in developed countries, diarrhea and other symptoms of these gastrointestinal infections are seldom fatal [2, 3], it is worth scrutinizing the statistical methodology that Helms et al, have employed to reach their striking results. The major conclusion from Helms et al, was that "Infections with all these bacteria were associated with an increased short term risk of death, even after pre-existing illnesses were taken into account. Salmonella, Campylobacter, and Yersinia enterocolitica infections were also associated with increased long term mortality." Their methodology relies on comparing ill cases with healthier (general population) controls. A comorbidity index is used to help control for the fact that ill patients with compromised immune systems might be expected to have higher mortality rates and higher bacterial infection rates than controls. This is because, as noted by Helms et al., "Underlying conditions were more common among patients than in the control group, particularly AIDS related illness, metastatic cancers, and lymphomas or leukaemia." However, they find that correcting for comorbidity makes surprisingly little difference: "After we adjusted for comorbidity, the relative mortality fell from 3.10 to 2.56 (95 % confidence interval 2.38 to 2.76)" A simultaneous explanation for the small impact of comorbidity adjustments on relative risk and also for the reported association between gastrointestinal infections and increased morbidity rates is found in the known limitations of the Charlson comorbidity index used by Helms et al. This index, while often useful for some conditions, is far from perfect, as has been documented in many studies [4-8]. For example, for lung cancer survival, it has been reported that “Comorbidity count and the Charlson index were significant predictors but explained only 2.5% and 2.0% of the survival variation, respectively. …Comorbidity count and the Charlson index failed to capture much information.” [5] The comorbidity index has not been validated for applications to gastrointestinal infections among patients with AIDS, leukemia, and other diseases that were prevalent in the case group. Thus, Helms et al, have relied on a comorbidity index that accounts for only a relatively small proportion of the total variance in survival and mortality rates among subjects. The remaining variance they attribute to bacterial pathogens. But severe underlying illness provides a far more plausible explanation. In terms of causal graphs [9], the situation may be diagrammed as follows: mortality rate <- AIDS, leukemia, etc.-> bacterial infection | comorbidity index That is, the presence of AIDS and other severe immune-compromising conditions among cases simultaneously predicts increased mortality rate, increased likelihood of bacterial infections, and a worse prognosis on the comorbidity index. But the comorbidity index is an imperfect indicator of the underlying condition and its effects on mortality rate. For example, it does not fully reflect the mortality rate consequences of a severe immune-compromising condition, as it has not been designed for this use. Thus, there can be substantial residual confounding by the underlying illness, even after conditioning on the comorbidity index. This creates a significant statistical association between bacterial infection and mortality rate, even though the former does not necessarily have any causal impact on the latter [9]. The effects of imperfectly controlled residual confounding and of imperfect indicators of underlying explanatory variables have been well documented in the epidemiological and biostatistical literatures [10-13]. They can create strong biases that can readily explain a host of otherwise surprising and counterintuitive statistical associations between exposures and health outcomes. Before accepting the proposition that infectious diarrhea in developed countries leads to a three-fold increase in mortality rates, as Helms et al. suggest, it would be appropriate to consider more carefully the non-causal association between infection rates and mortality rates brought about by incompletely controlled confounding from the serious conditions they have identified. REFERENCES [1] Helms M, Vastrup P, Gerner-Smidt P, Molbak K.. Short and long term mortality associated with foodborne bacterial gastrointestinal infections: registry based study. BMJ. 2003 Feb 15;326(7385):357. [2] Chowdhury MN. Campylobacter jejuni enteritis; a review. Trop Geogr Med. 1984 Sep;36(3):215-22. [3] Andrews GP. The enteric Campylobacter: they are everywhere. Clin Lab Sci. 1998 Sep-Oct;11(5):305-8. [4] Birim O, Maat AP, Kappetein AP, van Meerbeeck JP, Damhuis RA, Bogers AJ. Validation of the Charlson comorbidity index in patients with operated primary non-small cell lung cancer. Eur J Cardiothorac Surg. 2003 Jan;23(1):30-4. [5] Tammemagi CM, Neslund-Dudas C, Simoff M, Kvale P. Impact of comorbidity on lung cancer survival. Int J Cancer. 2003 Mar 1;103(6):792- 802. [6] Ghali WA, Hall RE, Rosen AK, Ash AS, Moskowitz MA .Searching for an improved clinical comorbidity index for use with ICD-9-CM administrative data. J Clin Epidemiol. 1996 Mar;49(3):273-8. [7] Cleves MA, Sanchez N, Draheim M. Evaluation of two competing methods for calculating Charlson's comorbidity index when analyzing short- term mortality using administrative data. J Clin Epidemiol. 1997 Aug;50(8):903-8. [8] Bravo G, Dubois MF, Hebert R, De Wals P, Messier L. A prospective evaluation of the Charlson Comorbidity Index for use in long- term care patients. J Am Geriatr Soc. 2002 Apr;50(4):740-5. [9] Shipley, B. Cause and Correlation in Biology. A user’s guide to path analysis, structural equations and causal inference. Cambridge University Press. 2000. http://callisto.si.usherb.ca:8080/bshipley/my%20book.htm. [10] Armstrong BG.. Effect of measurement error on epidemiological studies of environmental and occupational exposures. Occup Environ Med. 1998 Oct;55(10):651-6. [11] Feldman RA. Confounding factors in observational and intervention studies. Ital J Gastroenterol Hepatol. 1998 Oct;30 Suppl 3:S248-53. [12] Grimes DA, Schulz KF. Bias and causal associations in observational research. Lancet. 2002 Jan 19;359(9302):248-52. [13] Danesh J. Is there a link between chronic Helicobacter pylori infection and coronary heart disease? Eur J Surg Suppl. 1998;(582):27-31. Competing interests: The author has conducted research since 1999 on Campylobacter causes and risk analyses. This work has been partly supported by funds received from the US FDA and from the Animal Health Institute (AHI). This comment reflects the author's opinions only and has not been supported by the US FDA or the AHI. |
|||
|
|
|||
|
Kåre Mølbak, Consultant Department of Epidemiology, Statens Serum Institut, DK-2300 Copenhagen, Denmark, Morten Helms and Pernille Vastrup
Send response to journal:
|
We thank Adam Jacobs and Louis A. Cox for their interest in our work (1), and would like to clarify the points they raise in the two rapid responses entitled “Was the result specific to gastrointestinal infections?” and “Mortality associated with gastrointestinal infections re -examined”. Our analysis of short and long term mortality after foodborne bacterial infections was unique in several ways. The study group comprised unselected patients with infectious gastroenteritis, by and large persons with no underlying illness, who sought care at their family doctor. It was therefore expected that only a fraction of these patients had severe underlying illness (i.e., a comorbidity diagnosis). Sec-ondly, our study was the first study that determined mortality after foodborne in-fections while adjusting for background mortality. This was pivotal because gas- trointestinal infections frequently occur among elderly persons. The unexposed group was not randomly selected as suggested by Jacobs, but carefully matched by age, gender, and place of residence. Finally, we adjusted for comorbidity by collecting data from the national hospital discharge registry. We applied the prin-ciples described by Charlson, but did not use the weight proposed in her original work (2). Based on the actual survival rates of the large background population, we calculated empirical weights. This approach was used to ensure that the weights were valid and appropriate in the given context, and had the correct pre- dictive power. Although this approach may have been improved by obtaining more detailed clinical information and data on additional confounding factors, our work represents a great improvement compared with previous attempts to deter-mine mortality after foodborne infections. The biologic plausibility is supported by the fact that our estimates are in line with common knowledge of the different agents. For example, mortality after salmo-nella was higher than after campylobacter, and within the group of salmonella-infections, serotype Dublin, known to be invasive, was associated with a marked excess mortality. Although a long term mortality was observed for salmonella, campylobacter and yersinia, the proportion of deaths attributable to the infection was highest in the acute phase. The table below was prepared based on the figures in table 2 in our paper (1). The relative mortality rate has been converted to the at -tributable proportion of deaths among exposed, ((RR-1)/RR), i.e., a measure of the probability of a death being related to the gastrointestinal infection. To deter-mine the estimated attributable risk of death in a given time-interval, the mortality risk must be multiplied by the attributable proportion. In our opinion, the pattern presented in the table makes sense from a clinical point of view, and support the notion that our findings are more than artefacts. Nonetheless, we agree with Jacobs and Cox that the causal direction of our obser-vations needs to be carefully investigated in future studies. It is possible that gas-trointestinal infection may be a marker of increased vulnerability for some indi-viduals. It is also likely that the events in the causal chain that led to the diagnosis of the infection and further death were very complex and insufficiently described by our approach for a subset of the cases. However, even with these limitations, our methodology is an improvement compared with previous estimations, and the findings corroborate the notion that foodborne infections shall be taken seriously. Morten Helms, Pernille Vastrup, and Kåre Mølbak, Statens Serum Institut. 1. Helms M, Vastrup P, Gerner-Smidt P, Mølbak K. Short and long term mortality associated with foodborne bacterial gastrointestinal infections: registry based study. BMJ 2003; 326:357-60 2. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classify-ing prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987; 40:373-383. Table. Mortality data among 48,857 patients with Salmonella,
Campylobacter, Shigella and Yersinia enterocolitica infection. The
mortality is expressed as the cumulative mortality risk in a time-interval
after infection and the proportion of these deaths attributable to the
gastrointestinal infection.
Time since infection (days)
0-30 31-180 181-365
M AP M AP M AP
Salmonella 12.3 92% 10.8 55% 8.1 35%
Campylobacter 2.7 80% 5.0 46% 4.1 26%
Shigella 3.0 95% NS NS
Yersinia 1.7 72% 3.5 60% NS
M, mortality (per 1000)
AP, attributable proportion
NS, no significant excess mortality
Competing interests: None declared |
|||
|
|
|||
|
Darren C Greenwood, Lecturer in Medical Statistics Biostatistics Unit, University of Leeds, 24 Hyde Terrace, Leeds. LS2 9LN
Send response to journal:
|
Stephen Evans is to be congratulated for raising awareness of the matched cohort study. Lack of familiarity with this design has lead to many studies being wrongly labelled as case-control studies in the past. A quick review of all studies referred to as "case-control" in four major medical journals during 1996 (BMJ, Lancet, NEJM and JAMA) showed that a quarter of them had been misclassified to some extent and analysis was subsequently inappropriate in half of these. In these instances the analysis often ignored any matching, leading to potentially biased results. Some analyses were presented counter- intuitively, leading to difficulty in interpretation. Helms et al demonstrate the correct approach by conditioning on the matched sets to allow for the matched design. Part of the explanation for the problem may lie with studies where the exposure is a clinical condition in itself. The researcher is therefore tempted to label these exposed individuals as "cases". However in the case-control study, the term "case" is reserved for the outcome only. Whilst epidemiologists and statisticians pretend not to fall into these traps, the problem of mislabelling matched cohort studies is still common and is associated with misleading results and interpretation. Competing interests: None declared |
|||
|
|
|||
|
Sarah J. O’Brien, Head of Gastrointestinal Diseases Division PHLS Communicable Disease Surveillance Centre, Colindale, UK, NW9, Roger A. Feldman
Send response to journal:
|
Concerning the BMJ article of 15 February, by Helms et al[1], and the commentary by Professor Evans[2], it is important to recognize the possible effects of case selection and of clinical data on the estimates of short and long term mortality from clinical illness. Clinicians develop their understanding of the characteristics of an illness by identifying the symptoms and signs, in relation to age, gender, and residence, the three variables for which Helms et al selected controls. But clinicians also decide, on the basis of symptoms and signs, whether it is necessary to obtain a bacterial culture, and , if the result is deemed unlikely to influence patient management or outcome investigations, the culture might not be undertaken.[3] Variations in laboratory testing protocols and/or methods influence what is detected and reported.[4] Medical epidemiologists are aware that the infecting dose may affect severity. For many infectious gastrointestinal illnesses, the larger the infecting dose, the more severe the illness, the more likely the patient is to present to a clinician 5 and therefore, the more likely a stool culture might be obtained, so that severe illnesses are more likely to be represented in those studied by Helms et al. Medical microbiologists are aware, as acknowledged by Helms et al, that among the several thousand different Salmonellae, some, such as S. Typhi, S. Choleraesuis, S. Dublin,and S. Virchow may be more likely to be associated with severe illness and bacteremia. Similar variation in severity is present with Shigellae, with disease from Shigella dysenteriae more likely to be severe than Shigella sonnei, and with Campylobacter where certain serotypes seem pre-disposed to cause severe sequelae.[6] One way to validate the Helms et al estimate of short and long-term mortality is to obtain clinical information concerning the mortality. For example, they suggest that deaths that occur within one year may relate to the bacterial cause of illness, but we, as readers, are unaware of the data relating to those deaths. If a review of the death certificates might give an incomplete picture, then a review of the hospital records could reduce the difficulty. As it stands, the method, as suggested by Evans commentary[2], relates to estimates of an exposure that affects mortality, and makes full use of the extraordinary data available in Denmark and other Scandinavian countries. Further study, utilizing clinical information about cause of death, is needed to clarify whether the observations concern clinically severe cases, or represent other factors, not related to the gastrointestinal infection. References: 1. Helms M, Vastrup P, Gerner-Smidt P, Molbak K. Short and long term mortality associated with foodborne bacterial gastrointestinal infections: registry based study. BMJ 2003; 326: 357-360. 2. Evans S. Commentary: matched cohorts can be useful. BMJ 2003; 326: 361. 3. Aranda-Michel J, Gianella RA. Acute diarrhea: a practical review. Am J Med 1999; 106: 670-6. 4. Chapman PA, Siddons CA, Manning J, Cheetham C. An outbreak of infection due to verocytotoxin-producing Escherichia coli O157 in four families: the influence of laboratory methods on the outcome of the investigation. Epidemiol Infect 1997; 119 :113-9 5. Tam CC, Rodrigues LC, O’Brien SJ. The study of infectious intestinal disease in England: What risk factors for presentation to general practice tell us about potential for selection bias in case-control studies of reported cases of diarrhoea. Int J Epidemiol (in press). 6. Nachamkin I, Allos BM, Ho TW. Campylobacter and Guillain-Barré syndrome. Clin Microbiol Rev 1998; 11: 555 67. Competing interests: Dr. Feldman reviewed published and unpublished epidemiological data relating to Campylobacter on behalf of Bayer for an administrative hearing involving Bayer and the U. S. Food and Drug Administration's Center for Veterinary Medicine |
|||
|
|
|||
|
Kåre Mølbak, Department of Epidemiology Statens Serum Institut, DK-2300 Copenhagen S, Denmark, Morten Helms, Peter Gerner-Smidt
Send response to journal:
|
We agree that age and gender as well as symptoms and signs are important criteria when clinicians decide to obtain or not to obtain a bacterial culture. In addition, bacterial cultures are often requested for epidemiological reasons and for public health surveillance. Our material consisted mainly of previously healthy persons with acute infectious gastroenteritis. This, and the fact that we did adjust for comorbidity, makes the concerns that O’Brien and Feldman raise less of an issue. Nonetheless, few observational studies are perfect, and we agree that further clinical and epidemiological studies are warranted, including studies that address case selection, e.g., by the identification of risk factors for obtaining bacterial cultures. Although the observation of a one-year excess mortality is a new finding, it does make sense from a clinical point of view. Many acute infections are associated with both short- and long-term complications(1- 4). For the foodborne pathogens, a small subset of patients (even in a developed country) will develop severe dehydration, whereas other patients presenting with, for example, abdominal cramps are misdiagnosed and undergo surgery etc. All the pathogens we studied may more or less commonly spread to the blood stream, and Salmonella, in particular, is well known as a cause of focal- and vascular infections including endocarditis and arterial infections (3,5). Campylobacter is associated with the Guillain Barré syndrome (GBS), which may lead to hospitalisation and a period of assisted ventilation (2,6). Although each of these events is uncommon, they may, taken together, account for our findings. Studies are in progress to further address some of these issues. We agree that both the infecting dose and subtype of bacterial species are of importance. In the analyses, we looked at the effect of specific zoonotic Salmonella serotypes. Beyond S. Enteritidis, S. Typhimurium, and S. Dublin, we could not see any differences. We are sure there are differences, but the number of each of the exotic serotypes was too small to see this. It is probably also important, in future studies, to look at the effect of certain Campylobacter species and subtypes. There may, for example, be differences between the effect of C. coli and C. jejuni. As mentioned by O’Brien and Feldman, certain C. jejuni serotypes have in particular been associated with GBS. Finally, antimicrobial drug resistance may be associated with adverse public health effects (7-10). It was, however, beyond the scope of the paper to discuss these issues in detail. Future studies will have to address these questions, but this uncertainty should not stop anyone from preventing foodborne infections or reducing misuse of antimicrobials among humans and animals. Kåre Mølbak, Morten Helms and Peter Gerner-Smidt Reference List (1) Neal KR, Hebden J, Spiller R. Prevalence of gastrointestinal symptoms six months after bacterial gas-troenteritis and risk factors for development of the irritable bowel syndrome: postal survey of patients [see comments]. BMJ 1997;314:779-782. (2) Havelaar AH, de Wit MA, van Koningsveld R, van Kempen E. Health burden in the Netherlands due to infection with thermophilic Campylobacter spp. Epidemiol Infect 2000;125:505-522. (3) Cohen JI, Bartlett JA, Corey GR. Extra-intestinal manifestations of salmonella infections. Medicine (Baltimore) 1987;66:349-388. (4) Saebo A, Lassen J. A survey of acute and chronic disease associated with Yersinia enterocolitica infection. A Norwegian 10-year follow-up study on 458 hospitalized patients. Scand J Infect Dis 1991; 23:517-527. (5) Lester A, Eriksen NH, Nielsen H, Nielsen PB, Friis-Møller A, Bruun B et al. Non-typhoid Salmonella bacteraemia in Greater Copenhagen 1984 to 1988. Eur J Clin Microbiol Infect Dis 1991;10:486-490. (6) Allos BM. Campylobacter jejuni infection as a cause of the Guillain- Barre syndrome. Infect Dis Clin North Am 1998; 12:173-184. (7) Barza M, Travers K. Excess infections due to antimicrobial resistance: the "Attributable Fraction". Clin Infect Dis 2002;34 Suppl 3:S126-S130. (8) Travers K, Barza M. Morbidity of infections caused by antimicrobial- resistant bacteria. Clin Infect Dis 2002;34 Suppl 3:S131-S134. (9) Mølbak K, Baggesen DL, Aarestrup FM, Ebbesen JM, Engberg J, Frydendahl K et al. An outbreak of multidrug-resistant, quinolone-resistant Salmonella enterica serotype typhimurium DT104. N Engl J Med 1999;341:1420 -1425. (10) Helms M, Vastrup P, Gerner-Smidt P, Mølbak K. Excess mortality associated with antimicrobial drug-resistant Salmonella typhimurium. Emerg Infect Dis 2002;8:490-495. Competing interests: Dr. Mølbak reviewed data relating to Campylobacter on behalf of U. S. Food and Drug Administration's Center for Veterinary Medicine for an administrative hearing concerning a pro-posed withdrawal of the fluoroquinolone enrofloxacin (Baytril, Bayer) for use in poultry. |
|||
|
|
|||
|
Phillippa M Cumberland, Senior research fellow Instutute of Child Health, London WC1N 1EH, Jenny Roberts and Laura Rodrigues
Send response to journal:
|
Helms et al describe a significant increase in mortality rates in cases with foodborne bacterial infections compared to controls.1 We consider that there are methodological issues that may have led to mortality rates being over estimated in this study. In Helms et al response [BMJ reply 21/02/03] the group studied is described as ‘unselected patients…..who sought care at their family doctor’. Study cases may have been representative of those presenting to their family doctor, but not all community cases of organism specific infection present to their family doctor and those who do are unlikely to be representative of all cases in the community. In the infectious intestinal disease (IID) study in England it was found that the proportion of cases not recorded by national surveillance was large and varied widely by organism.2 Only 27% of community cases with IID presented to the GP and cases presenting to the GP had symptoms that were more severe, frequent and of longer duration than all cases found in the community.3 Of the cases presenting 22% had stool specimens sent for analysis and of those 45% had no organism isolated. In the UK outbreak cases are also reported and there may be cases with severe infections identified via this route without presenting to their GP. Campylobacter and Salmonella infections have severe symptoms so it is more likely that patients with these infections would see a GP, but it cannot be assumed that cases in this study represent all cases in the community. A further issue is that of residual confounding for underlying illness. The limitations of the Charlson comorbidity index have been discussed elsewhere. [Cox LA. BMJ response 19/03/03]. Finally, the authors suggest that all mortality was a direct cause of the GI infection, and ignore a body of research suggesting that infections can trigger heart disease.4 In a three month prospective follow-up study of cases and controls in the IID study it was found that cases were significantly more likely to present to GPs (for all causes) than controls, both at 2-3 weeks and 4-12 weeks after IID episode.5 Our data suggested that the burden of IID when the long term sequelae are taken into account is considerable. However, we feel that the increased burden due to mortality as described in the Helms’ study 1 is probably over-estimated. Phillippa Cumberland, Senior Research Fellow Centre for Paediatric Epidemiology and Biostatistics Institute of Child Health, London WC1N 1EH Jenny Roberts, Professor Centre for Economics of Infectious Diseases London School of Hygiene & Tropical Medicine, London WC1E 7HT Laura Rodrigues, Reader Infectious Disease Epidemiology Unit London School of Hygiene & Tropical Medicine, London WC1E 7HT 1. Helms M, Vastrup P, Gerner-Smidt P, Molbak K. Short and long term mortality associated with foodbourns bacterial infections:registry based study. BMJ 2003; 326:357-360 2. Wheeler JG, Sethi D, Cowden JM, Wall PG, Rodrigues LC, Tompkins DS, Hudson MJ, Roderick PJ. Study of Infectious Intestinal Disease in England: Rates in the community, presenting to GPs and reported to national surveillance. BMJ 1999;318:1046-1050. 3. Roberts JA, Cumberland P, Sockett PN, Wheeler J, Rodrigues LC, Sethi D and Roderick PJ. The study of infectious intestinal disease in England: socio-economic impact. Epi Infect. 2003, 130, 1-11. 4. Zebrack JS, Anderson JL. The role of inflammation and infection in the pathogenesis and evolution of coronary artery disease. Curr Cardiol Rep 2002 Jul;4(4):278-88 5. Cumberland P, Sethi D, Roderick PJ, Wheeler JG, Cowden JM, Roberts JA, Rodrigues LC, Hudson MJ, Tompkins D. The study of infectious intestinal disease in England: A prospective evaluation of symptoms and health care use after an acute episode. Epi. Infect. (in press). Competing interests: None declared |
|||
|
|
|||
|
Kåre Mølbak, Consultant Department of Epidemiology, Statens Serum Institut, 2300 Copenhagen S, Denmark., Morten Helms, Peter Gerner-Smidt
Send response to journal:
|
We thank for the interest in our work. It is well documented that only a fraction of infections with gastrointestinal bacteria are culture confirmed (1-3). Our study was based on confirmed cases, and it would be wrong to apply these findings to non-diagnosed community cases. The issue about residual confounding has been discussed in the previous responses. To summarize: The comorbidity index we applied was a relevant measure of underlying illness because most patients with a pre-existing disease severe enough to alter the outcome of a foodborne gastrointestinal infection are likely to have come into some sort of contact with a hospital or an outpatient clinic within a five-year period before gastrointestinal infection. In addition, the weights used in the comorbidity index were developed on the basis of the actual survival of the very large group of unexposed persons. The concerns about bias introduced by "case-selection" may, furthermore, be less relevant in Denmark than in many other areas. Laboratory costs are reimbursed by the Danish counties, and therefore many Danish physicians request stool culture simply because they want to know the cause of the disease (even if it is relatively mild), and, for example, whether the patient was part of an outbreak or the infection was linked to other similar episodes of illness. We agree that some infections may trigger heart disease. Indeed, a large part of the long-term excess mortality may be explained by medium to long-term consequences of acute infections, such as campylobacter triggering Guillain-Barré syndrome, or respiratory distress syndrome which often may follow septicemia, and so on. Hitherto, estimates of mortality due to gastrointestinal infection have been based on case-fatality rates from outbreaks or population surveys (4-6). These figures have not even been adjusted for age-specific background mortality rate. Our method represent a more valid approach. 1. Mead PS, Slutsker L, Dietz V, McCaig LF, Bresee JS, Shapiro C et al. Food-related illness and death in the United States. Emerg Infect Dis 1999;5:607-625. 2. de Wit MA, Hoogenboom-Verdegaal AM, Goosen ES, Sprenger MJ, Borgdorff MW. A population-based longitudinal study on the incidence and disease burden of gastroenteritis and Campylobacter and Salmonella infection in four regions of The Netherlands. Eur J Epidemiol 2000;16:713- 718. 3. Wheeler JG, Sethi D, Cowden JM, Wall PG, Rodrigues LC, Tompkins DS, Hudson MJ, Roderick PJ. Study of Infectious Intestinal Disease in England: Rates in the community, presenting to GPs and reported to national surveillance. BMJ 1999;318:1046-1050. 4. Cohen ML, Tauxe RV. Drug-resistant Salmonella in the United States: an epidemiologic perspective. Science 1986; 234:964-969. 5. Ryan CA, Nickels MK, Hargrett-Bean NT, Potter ME, Endo T, Mayer L et al. Massive outbreak of antimicrobial-resistant salmonellosis traced to pasteurized milk. JAMA 1987;258:3269-3274. 6. Mishu B, Koehler J, Lee LA, Rodrigue D, Brenner FH, Blake P et al. Outbreaks of Salmonella enteritidis infections in the United States, 1985- 1991. J Infect Dis 1994;169:547-552. Competing interests: None declared |
|||