Validation of treatment strategies for enterohaemorrhagic Escherichia coli O104:H4 induced haemolytic uraemic syndrome: case-control study

BMJ 2012; 345 doi: http://dx.doi.org/10.1136/bmj.e4565 (Published 19 July 2012)
Cite this as: BMJ 2012;345:e4565
  1. Jan Menne, consultant nephrologist1,
  2. Martin Nitschke, consultant nephrologist2,
  3. Robert Stingele, professor of neurology3,
  4. Mariam Abu-Tair, consultant nephrologist4,
  5. Jan Beneke, medical student1,
  6. Jörn Bramstedt, consultant nephrologist5,
  7. Jan P Bremer, gastroenterologist6,
  8. Reinhard Brunkhorst, professor of nephrology7,
  9. Veit Busch, consultant nephrologist8,
  10. Reinhard Dengler, professor of neurology and director1,
  11. Günther Deuschl, professor of neurology and director3,
  12. Klaus Fellermann, professor of gastroenterology2,
  13. Helmut Fickenscher, professor of microbiology and director3,
  14. Christoph Gerigk, consultant infectologist9,
  15. Alexander Goettsche, medical student3,
  16. Jobst Greeve, professor of gastroenterology and director10,
  17. Carsten Hafer, consultant nephrologist1,
  18. Friedrich Hagenmüller, professor of gastroenterology and director6,
  19. Hermann Haller, professor of nephrology and director1,
  20. Stefan Herget-Rosenthal, nephrologist and director11,
  21. Bernd Hertenstein, professor of haematology and director12,
  22. Christina Hofmann, consultant neurologist2,
  23. Melanie Lang, registrar13,
  24. Jan T Kielstein, professor of nephrology1,
  25. Ulrich C Klostermeier, doctoral student3,
  26. Johannes Knobloch, professor of medical microbiology and assistant director2,
  27. Markus Kuehbacher, research fellow14,
  28. Ulrich Kunzendorf, professor of nephrology and director3,
  29. Hendrik Lehnert, professor of internal medicine and director2,
  30. Michael P Manns, professor of gastroenterology and director1,
  31. Tobias F Menne, consultant haematologist15,
  32. Tobias N Meyer, consultant nephrologist13,
  33. Claus Michael, medical student1,
  34. Thomas Münte, professor of neurology and director2,
  35. Christine Neumann-Grutzeck, registrar6,
  36. Jens Nuernberger, professor of nephrology and director16,
  37. Hermann Pavenstaedt, professor of nephrology and director8,
  38. Leyla Ramazan, house officer1,
  39. Lutz Renders, consultant nephrologist3,
  40. Jonas Repenthin, consultant neurologist13,
  41. Wolfgang Ries, consultant nephrologist17,
  42. Axel Rohr, consultant of neuroradiology and assistant director3,
  43. Lars Christian Rump, professor of nephrology and director18,
  44. Ola Samuelsson, professor of nephrology19,
  45. Friedhelm Sayk, consultant of gastroenterology2,
  46. Bernhard M W Schmidt, consultant nephrologist1,
  47. Sabine Schnatter, consultant nephrologist20,
  48. Harald Schöcklmann, professor of nephrology3,
  49. Stefan Schreiber, professor of internal medicine and director3,
  50. Cay U von Seydewitz, consultant haematologist6,
  51. Jürgen Steinhoff, professor of nephrology and assistant director2,
  52. Sylvia Stracke, nephrologist and director21,
  53. Sebastian Suerbaum, professor of medical microbiology and director1,
  54. Andreas van de Loo, professor of internal medicine9,
  55. Martin Vischedyk, consultant nephrologist10,
  56. Karin Weissenborn, professor of neurology1,
  57. Peter Wellhöner, consultant gastroenterologist2,
  58. Monika Wiesner, consultant nephrologist22,
  59. Sebastian Zeissig, senior house officer3,
  60. Jürgen Büning, consultant gastroenterologist2,
  61. Mario Schiffer, professor of nephrology1,
  62. Tanja Kuehbacher, consultant gastroenterologist3
  63. on behalf of the EHEC-HUS consortium
  1. 1Medical School Hannover, Hannover, Germany
  2. 2University Hospital Schleswig-Holstein, Lübeck, Germany
  3. 3University Hospital Schleswig-Holstein, Kiel, Germany
  4. 4Evangelic Hospital Gilead, Bielefeld, Germany
  5. 5Bremerhaven Hospital Reinkenheide, Bremerhaven, Germany
  6. 6Asklepios Hospital Hamburg Altona, Hamburg
  7. 7Hospital Hannover, Hannover, Germany
  8. 8University Hospital Münster, Münster, Germany
  9. 9Marienhospital Hamburg, Hamburg
  10. 10St Vincenz Hospital, Paderborn, Germany
  11. 11Red Cross Hospital Bremen, Bremen, Germany
  12. 12Bremen Hospital-Mitte, Bremen, Germany
  13. 13Asklepios Hospital Hamburg Barmbek, Hamburg
  14. 14University of Hagen, Hagen, Germany
  15. 15Freeman Hospital, Newcastle upon Tyne, UK
  16. 16Helios Hospital Schwerin, Schwerin, Germany
  17. 17Diakonissenkrankenhaus Flensburg
  18. 18Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
  19. 19Sahlgrenska University Hospital, Göteborg, Sweden
  20. 20Ernst von Bergmann Clinic, Potsdam, Germany
  21. 21Clinical Centre, University of Greifswald, Greifswald, Germany
  22. 22Hospital Oldenburg, Oldenburg, Germany
  1. Correspondence to: J Menne Department of Nephrology and Hypertension, Medical School Hannover, Carl Neuberg Strasse 1, 30625 Hannover, Germany menne.jan{at}mh-hannover.de
  • Accepted 6 June 2012

Abstract

Objective To evaluate the effect of different treatment strategies on enterohaemorrhagic Escherichia coli O104:H4 induced haemolytic uraemic syndrome.

Design Multicentre retrospective case-control study.

Setting 23 hospitals in northern Germany.

Participants 298 adults with enterohaemorrhagic E coli induced haemolytic uraemic syndrome.

Main outcome measures Dialysis, seizures, mechanical ventilation, abdominal surgery owing to perforation of the bowel or bowel necrosis, and death.

Results 160 of the 298 patients (54%) temporarily required dialysis, with only three needing treatment long term. 37 patients (12%) had seizures, 54 (18%) required mechanical ventilation, and 12 (4%) died. No clear benefit was found from use of plasmapheresis or plasmapheresis with glucocorticoids. 67 of the patients were treated with eculizumab, a monoclonal antibody directed against the complement cascade. No short term benefit was detected that could be attributed to this treatment. 52 patients in one centre that used a strategy of aggressive treatment with combined antibiotics had fewer seizures (2% v 15%, P=0.03), fewer deaths (0% v 5%, p=0.029), required no abdominal surgery, and excreted E coli for a shorter duration.

Conclusions Enterohaemorrhagic E coli induced haemolytic uraemic syndrome is a severe self limiting acute condition. Our findings question the benefit of eculizumab and of plasmapheresis with or without glucocorticoids. Patients with established haemolytic uraemic syndrome seemed to benefit from antibiotic treatment and this should be investigated in a controlled trial.

Introduction

The recent outbreak of enterohaemorrhagic Escherichia coli associated haemolytic uraemic syndrome in northern Germany, 2011, the largest to date, underscored the threat to public health of such pathogens.1 Diarrhoea associated haemolytic uraemic syndrome is characterised by the triad of microangiopathic haemolytic anaemia, thrombocytopenia, and acute kidney injury. In the outbreak in Germany, 855 of 3842 people infected with enterohaemorrhagic E coli developed haemolytic uraemic syndrome.2 Adults, especially healthy middle aged women, were predominantly affected (90%), probably because of the source of the infection (fenugreek sprouts). The second largest outbreak was considerably smaller and occurred in Scotland, when 34 of 512 (279 confirmed) people infected with enterohaemorrhagic E coli O157:H7 developed haemolytic uraemic syndrome; 28 were adults.3 4 The 22% incidence rate of haemolytic uraemic syndrome was higher than the 1-15% typically reported in other outbreaks of enterohaemorrhagic E coli.3 5 6 This is probably because the outbreak in Germany was caused by an unusual strain: enterohaemorrhagic E coli O104:H4. Further analysis by several teams, including use of third generation sequencing methods, revealed that the strain carried properties of a shiga 2 toxin producing E coli and an enteroaggregative E coli as well as multidrug resistant plasmids.7 8 9 10 This new combination of genes seemed to enhance the occurrence of haemolytic uraemic syndrome in comparison with the O157:H7 strain encountered in most previous outbreaks.8 11 Importantly, data gathered by German paediatricians suggested that the clinical course and outcome of the O104:H4 induced disease was similar to that of infections with O157:H7.12 Data generated from the 2011 outbreak could therefore add valuable information to the treatment of patients with haemolytic uraemic syndrome.

Over the years different treatments comprising plasmapheresis, glucocorticoids, antibiotics, and eculizumab have been developed to treat enterohaemorrhagic E coli associated haemolytic uraemic syndrome.5 13 14 Owing to the sporadic nature of the disease, most of the published studies examining the efficacy of treatments only analysed small groups and lacked a comparison group. Currently used active interventions are therefore not based on formal evidence. The large number of patients affected and differences in treatment strategies between the hospitals enabled us to compare and analyse treatment options in an exploratory fashion. We were mainly interested in the outcomes after treatment with plasmapheresis, glucocorticoids, antibiotics, and eculizumab (for terminal complement blockade). In some hospitals patients received no more than 3-5 sessions of plasmapheresis (limited plasmapheresis) followed by alternative treatments if required. In contrast, in most hospitals plasmapheresis was continued until platelet counts increased to at least 100/nL. Furthermore, several hospitals did not use glucocorticoids, whereas the others administered at least 50 mg with plasmapheresis. Twenty two hospitals followed the recommendation not to use antibiotics and administered them only if medically required. One hospital, however, used double or triple antibiotic treatment with meropenem, ciprofloxacin, and rifaximin with the aim of speedy eradication. Eculizumab is a monoclonal antibody directed against the complement protein C5. It has been successfully used to treat patients with paroxysmal nocturnal haemoglobinuria15 and atypical haemolytic uraemic syndrome.16 In May 2011, just at the beginning of the Germany outbreak, it was reported that eculizumab seemed to be beneficial in three infants with severe shiga toxin associated haemolytic uraemic syndrome.14 In these cases eculizumab was given to block dysregulated complement activation, which is thought to play a part in the development of haemolytic uraemic syndrome.17 18 In light of this report, eculizumab was extensively used (>300 patients) as a therapeutic option during the Germany outbreak. The recommended treatment schedule was 900 mg at days 0, 7, 14, and 21, and 1200 mg at days 28, 42, and 56. Additionally, azithromycin was prescribed for 14 days to prevent meningococcal infections.

We evaluated the effectiveness of these various therapeutic strategies in 298 patients with enterohaemorrhagic E coli associated haemolytic uraemic syndrome (35% of the population affected). These patients were similar for age, sex distribution, and mortality as the 855 patients with the disease reported to the Robert Koch Institute2 and in line with data collected by the relevant German registry (JT Kielstein, personal communication, 2012).

Methods

During the first three weeks of the outbreak of enterohaemorrhagic E coli associated haemolytic uraemic syndrome in northern Germany a group of clinicians involved in the clinical care of patients with this condition assembled to learn as much as possible from the outbreak as it evolved. The group initially comprised nephrologists, gastroenterologists, neurologists, and others from three university hospitals (Hannover, Lübeck, and Kiel). Other hospitals then joined. The participating centres prospectively agreed on the data to be collected. These data were entered into a database that was available to all members of the consortium to ensure correct data entry. As the outbreak evolved, more centres joined the consortium. A dedicated study team used a standardised case history form to collect the data retrospectively from patients’ charts. To minimise the number of mistakes, two study team members in most cases entered the data on each patient. In 20 hospitals it was possible to export all laboratory values into an Excel file, significantly reducing the number of possible mistakes. In the other centres the laboratory values were entered manually into the database. We carried out several checks of the database and we reviewed the patients’ charts for a second time if data seemed to be incorrect or important data were missing.

Patient population

This report is on the patients treated in these centres who fulfilled the criteria for haemolytic uraemic syndrome. Some of the centres were participating in an industry sponsored open label trial of eculizumab in patients with haemolytic uraemic syndrome. These centres had signed a non-disclosure agreement with the manufacturer of eculizumab (Soliris; Alexion, Cheshire, CT). We included in the present study those patients who received eculizumab outside of the trial.

The consortium has treated a total of 395 adults with enterohaemorrhagic E coli associated haemolytic uraemic syndrome at 23 centres: six university hospitals and 17 district hospitals. Ninety seven patients receiving eculizumab were excluded from analysis as they were part of the industry sponsored trial (tables 1 and 2, fig 1). The remaining 298 (75%) patients were included in the present analysis. Part of the clinical data on 33 of those patients has been published previously.19 20

Fig 1 Therapeutic strategies used by hospitals during outbreak of enterohaemorrhagic Escherichia coli associated haemolytic uraemic syndrome in northern Germany, 2011. Percentage of patients available for present analysis are in brackets

View this table:
Table 1

 Characteristics of patients within participating centres. Usage of eculizumab is shown to highlight how participation in the industry sponsored trial of the drug affected the number of patients available for analysis. Values are numbers (percentages) of participants

View this table:
Table 2

 Effect of eculizumab trial on number of patients available to analyse from hospitals using different treatment strategies. Values are numbers (percentages) of patients unless stated otherwise

Definition of haemolytic uraemic syndrome

We defined haemolytic uraemic syndrome as a diagnosis of enterohaemorrhagic E coli associated haemolytic uraemic syndrome between May and July 2011. Patients with stools positive for enterohaemorrhagic E coli or shiga toxin 2 or a history of bloody diarrhoea needed to fulfill the following three diagnostic criteria for haemolytic uraemic syndrome: platelet count <150×109/L, haemolytic anaemia with haemoglobin level less than lower limit of normal range, and serum creatinine level greater than upper limit of normal range. Overall, 283 out of the 298 patients fulfilled these criteria. The remaining 15 patients did not fulfill the criterion for increased creatinine levels but haemolytic uraemic syndrome was diagnosed by the treating doctor and eight of the patients underwent plasmapheresis. We included these patients in our analysis.

Outcome variables

We chose the following as outcome variables as they were the most severe complications observed during the outbreak: requirement for dialysis, mechanical ventilation, abdominal surgery due to perforation or bowel necrosis, documented seizures, and death.

Statistical analysis

As treatment strategies varied between the centres (fig 1 and table 2) we compared the outcomes of the different approaches. The analysis was done using R, Prism5, or SPSS. We compared the two treatment groups using an unpaired two sided Student’s t test and categorical variables using a χ2 test. Multivariate analysis was done by logistic regression. The independent variables and covariables (platelet count, creatinine level, leucocyte count, age, and sex) were defined a priori, and in univariate analysis all of these variables were predictors for a worse outcome. We also used a linear model to analyse if differences in the mean laboratory values over time differed between the treatment strategies. We did not adjust for multiple testing as the intention of our analysis was exploratory. Data are presented as means (standard deviations) and medians (ranges).

Results

Tables 3 and 4 summarise the characteristics of the patients at the time haemolytic uraemic syndrome was diagnosed. Overall, 212 (71%) were women and the median age was 45.5 (range 18-86) years. In total, 99% of cases had diarrhoea that usually turned sanguineous within one day (fig 2). The earliest date diarrhoea started was 10 May 2011 and the latest 19 June, with 75% of cases occurring before 24 May. Diarrhoea lasted a mean 9.3 (SD 6.7) days (median 6.0 days). Most patients were admitted to hospital within seven days after the onset of diarrhoea (fig 2), with a mean of 4.1 (SD 4.4) days (median 3.0 days). The mean length of hospital stay was 22.6 (SD 14.3) days (median 19 days). A formal diagnosis of haemolytic uraemic syndrome was established after a median of 6 (range 4-8) days after the onset of diarrhoea (fig 2).

Fig 2 Onset of symptoms and complications in 298 patients, laboratory results, and diagnosis of haemolytic uraemic syndrome

View this table:
Table 3

 Baseline characteristics of 298 patients at time haemolytic uraemic syndrome was diagnosed, in relation to receipt of plasmapheresis and antibiotic treatment. Values are means (standard deviations) unless stated otherwise

View this table:
Table 4

 Baseline characteristics at day of first plasmapheresis in 251 patients treated in centres that used platelet guided plasmapheresis (until platelet counts were >100/nL), re-evaluated plasmapheresis after 3-5 sessions (limited), or used glucocorticoids with plasmapheresis. Patients receiving eculizumab were compared with a matched control group with similar severity of haemolytic uraemic syndrome. Values are means (standard deviations) unless stated otherwise

Haemolysis and thrombocytopenia

At the onset of diarrhoea the levels of lactate dehydrogenase were slightly raised and platelet counts were in normal range but both variables quickly reached parallel levels that were indicative of disease, peaking at days 6-8 (fig 2). On average the platelet counts normalised within one week. In contrast, lactate dehydrogenase levels declined only slowly and remained out of normal range for several weeks. At diarrhoea onset, the results for haemoglobin levels were still within normal range but dropped progressively and reached a nadir at day 12 (fig 2). One hundred and sixty seven (56%) patients underwent transfusion, with a mean 4.9 (SD 3.6) blood units used (fig 2).

Acute kidney injury

A sharp increase in serum creatinine levels was noted during the first days of diarrhoea onset, and by day 8 the mean values had reached peak levels (fig 2). One hundred and sixty (54%) patients required renal replacement therapy and in 50% of these treatment was started within the first eight days (fig 2). Dialysis was required for a mean 10.3 (SD 11.5) days, with a mean 7.7 (SD 8.1) sessions provided. In most patients renal function normalised within 4-6 weeks. Long term (>6 months) data are available for 238 patients. At discharge 13 of these patients were receiving dialysis; only three required long term renal replacement therapy.

Neurological and other complications

One hundred and forty three (48%) patients had neurological symptoms (fig 2). Symptoms included cerebellar dysfunction, myoclonus, memory disorders, focal neurological deficits, and seizures.

Overall, 156 (52%) of the patients were treated in the intensive care unit and 54 (18%) required mechanical ventilation a mean 9.1 (SD 5.3) days after disease onset (fig 2). Patients were ventilated for a mean 11.4 (SD 10.1) days (median 8 (range 6-13) days). Twelve (4%) patients died (mean age 68.7 (range 24-86, fig 2). More than 50% of the deaths were due to infective complications. Laparotomy was required in seven patients who developed an acute abdomen due to perforation or severe colitis, often within the first days of the disease.

Effect of plasmapheresis

Plasmapheresis was carried out in 251 (84%) patients. Treatment was started after a mean 6.8 (SD 3.0) days of disease onset (table 3 and fig 2). The treatment was continued for a mean 7.6 (SD 7.4) days (median 5 (range 3-9) days) and the patients received a mean 7.3 (SD 6.7) sessions (median 5 (range 4-9) sessions). Mean lactate dehydrogenase levels declined and platelet counts increased after the start of plasmapheresis (fig 3). However, similar patterns for lactate dehydrogenase and creatinine levels were observed in 47 patients who did not undergo plasmapheresis (fig 4). These patients had a milder form of haemolytic uraemic syndrome (tables 3-6 and fig 4). Mortality seemed to be higher in the group that did not undergo plasmapheresis (table 4), but all four cases had complications due to infection that were unlikely to have been influenced by plasmapheresis. The cause of death in two patients was severe colitis with perforation of the bowel within the first three days after onset of diarrhoea. A third patient had severe colitis and refused invasive therapy. These three patients died within the first nine days after diarrhoea onset. A fourth patient (aged 83 years) recovered from the haemolytic uraemic episode but developed an aspiration pneumonia and died on day 38.

Fig 3 Effect of plasmapheresis on platelet counts and levels of lactate dehydrogenase, creatinine, and haemoglobin

Fig 4 Clinical data in patients with enterohaemorrhagic Escherichia coli induced haemolytic uraemia syndrome treated with or without limited plasmapheresis (3-5 sessions) versus platelet guided plasmapheresis, with plasmapheresis with or without glucocorticoid therapy, and with or without antibiotics

View this table:
Table 5

 Complications and outcomes in patients treated with or without plasmapheresis stratified by treatment strategies and in patients treated with eculizumab compared with control group with similar severity of haemolytic uraemic syndrome. Values are numbers (percentages) of patients unless stated otherwise

View this table:
Table 6

 Outcomes in 298 patients with haemolytic uraemic syndrome according to treatment strategies. Values are means (standard deviations) unless stated otherwise

Three of the centres (total 54 patients) carried out limited plasmapheresis (3-5 sessions) and then re-evaluated the treatment, whereas the other 20 centres continued with plasmapheresis until the platelet count had increased to at least 100/nL (table 3). We therefore analysed whether a more intensive plasmapheresis regimen would change the main outcome variables. At the start of treatment the patient groups were well matched (table 3). Patients in the platelet guided arm required dialysis more often after the start of plasmapheresis than patients in the limited plasmapheresis arm (tables 5 and 6). Platelet counts and the levels of lactate dehydrogenase and creatinine were comparable (fig 4)

Effect of plasmapheresis associated glucocorticoid therapy

For plasmapheresis most of the centres used high dose (≥50 mg) prednisone (or prednisolone) as premedication before fresh frozen plasma was administered. However, seven centres (total 80 patients) did not administer glucocorticoids (table 4). No benefit of glucocorticoid treatment was observed and the time course suggested a delay in platelet recovery and normalisation of lactate dehydrogenase levels (table 5 and fig 4). No significant differences could be detected using a linear model.

Effect of treatment with antibiotics

One university hospital administered a combination of at least two antibiotics. In particular, meropenem and ciprofloxacin were given intravenously in dosages adapted to glomerular filtration rates to prevent complications such as intestinal damage and sepsis. Rifaximin 600 mg daily was given orally to patients on the intensive care unit. Both groups (with or without antibiotics) had comparable baseline variables at diagnosis of haemolytic uraemic syndrome (table 3). With antibiotic treatment the duration of enterohaemorrhagic E coli excretion in stools was significantly shortened, from a mean 22.6 (SD 11.3) days to 14.8 (10.6) days (P<0.001). The incidence of seizures was significantly (P=0.03) lower in the antibiotic treatment group (tables 3 and 6). Furthermore, death was lower (0% v 5.2%, P=0.029), the need for intestinal surgery was not significant (0 v 2.8%), and there were no signs of toxic shock (table 5).

Effect of treatment with eculizumab

In the present cohort 67 patients received eculizumab outside the industry sponsored trial (table 4), with treatment started a mean 10.2 (SD 4.6) days after the onset of diarrhoea and a mean total of 2700 mg administered. Twenty three (34%) of these 67 patients required ventilation, 51 (76%) required dialysis, and 16 (24%) had seizures (table 7). The severity of haemolytic uraemic syndrome was therefore worse than in the rest of the study population. Despite treatment with eculizumab, 27 (40%) patients needed further plasmapheresis (table 6). After the start of treatment seven patients (10%) required ventilation, 12 (18%) required dialysis, and four (6%) had seizures (table 6).

View this table:
Table 7

 Baseline characteristics of 251 patients with haemolytic uraemia syndrome in centres that used plasmapheresis until normalisation of platelets >100/nL (platelet guided), that re-evaluated plasmapheresis after 3-5 sessions (limited), or did or did not use glucocorticoids with plasmapheresis. Values are means (standard deviations) unless stated otherwise

To analyse the therapeutic effect of eculizumab a control group with a similar severity of haemolytic uraemic syndrome was formed who did not receive eculizumab but did undergo plasmapheresis and fulfilled the following criteria: lactate dehydrogenase concentrations >700 U/L at diagnosis of haemolytic uraemic syndrome, with maximal concentrations >1500 U/L or seizure or ventilation or death. These criteria were fulfilled by 65 out of 184 patients who underwent plasmapheresis. These patients had similar baseline characteristics and rates of complications to the group treated with eculizumab (table 4). No significant difference was noted between the groups for platelet recovery and levels of lactate dehydrogenase, creatinine, or haemoglobin (fig 5). The rate of complications was also similar between the groups (tables 6 and 7).

Fig 5 Platelet counts and levels of lactate dehydrogenase, creatinine, and haemoglobin in 67 patients who were treated with eculizumab and 65 control patients with a similar severity of haemolytic uraemic syndrome

Discussion

Current treatment recommendations for adults with haemolytic uraemic syndrome might need to be modified in light of the findings in this large cohort of patients from the outbreak of enterohaemorrhagic Escherichia coli associated haemolytic uraemic syndrome in northern German, 2011. Evidence of the benefits from plasmapheresis was not clear. Contrary to current notions, antibiotic treatment of established haemolytic uraemic syndrome is not harmful and might even improve the outcome.

Strengths and limitations of the study

The major strength of this study was the large number of patients and the extensive data available by database to describe the population. Another advantage was that not all the centres used the same treatment strategy. These differences allowed a retrospective analysis of treatment strategies. This non-randomised group assignment was also the major weakness of our study. Our comparisons involved imperfect controls and thus bias was introduced by indication. Such bias is obvious in the analysis of plasmapheresis, a treatment that was begun in patients with severe haemolytic uraemic syndrome but not in the less severely affected patients who served as controls. Whenever possible we used adjusted analyses to take into account the differences in baseline severity. Nevertheless, these results should be interpreted with caution. Comparisons are easier when baseline characteristics are similar, such as in patients who received limited plasmapheresis versus platelet guided plasmapheresis.

Importantly, differences in participation between the centres owing to the exclusion of patients taking part in the industry sponsored trial gives rise to bias. For example, none of the patients receiving the early antibiotic strategy were excluded, owing to participation in the eculizumab trial, but 28% were excluded in the respective control group when they received eculizumab (table 2). As patients receiving eculizumab were usually sicker than the rest of the study population this biases the analysis towards a smaller therapeutic effect of antibiotics. Limited plasmapheresis would be biased towards a more favourable outcome as only 49.1% of the patients in these centres were included in the present analysis.

Clinical picture of haemolytic uraemic syndrome in adults

The 298 patients with haemolytic uraemic syndrome in the present study had a similar sex and age distribution to the adults with haemolytic uraemic syndrome reported by the Robert Koch Institute.1 2 Our figures for mortality were also in line with the national data (4.0% v 4.1%).3 Most of the formerly healthy patients were severely ill during the acute phase of the disease: 54% required dialysis, many had neurological problems, and more than 12% had seizures. Over half of the patients (n=156, 52%) were treated on an intensive care ward and more than 18% (n=54) required ventilation for an average of 10 days. Despite the severity of the disease during the acute phase, most of the patients recovered completely and only five were still receiving renal replacement therapy after nine months. A previous study reported that the neurological symptoms resolved completely in most patients.21

Plasmapheresis

At the beginning of this outbreak the German Society of Nephrology recommended use of plasmapheresis, especially for cases of enterohaemorrhagic E coli associated haemolytic uraemic syndrome with neurological or severe renal involvement. This recommendation is supported by the American Society for Apheresis, which gives a low II-3 recommendation for the usage of plasmapheresis in patients with typical haemolytic uraemic syndrome.22 It is believed that plasmapheresis might remove the circulating shiga toxin or factors that damage the endothelium. Data to support such an assumption are, however, scarce. Firstly, shiga toxin has never been identified in the circulation. Secondly, the density of the infecting organism and concentration of toxin in stools diminish in the colon as haemolytic uraemic syndrome develops.23 Thirdly, there is ample evidence of vascular injury before haemolytic uraemic syndrome ensues,24 and the microvascular damage is possibly already manifested before the clinical manifestation of the disease. In addition, injected shiga toxin in animal models has shown a short half life in the circulation. Hence the evidence for plasmapheresis is based on empirical observations.25 In the Scottish outbreak overall mortality seemed better in the small number of patients treated with plasmapheresis, although statistically robust conclusions were not possible owing to lack of power.3 More recently, findings in five patients in the 2011 outbreak were published and it was suggested that plasmapheresis is beneficial.26 Furthermore, the authors of an accompanying editorial mentioned that plasmapheresis remains “the cornerstone of treatment.”27 We observed an improvement in the average platelet count and a drop in the levels of lactate dehydrogenase after initiation of plasmapheresis in 251 patients. However, our data also suggest that this might reflect the natural course of the disease, as plasmapheresis was usually started at the peak of disease activity, around days 6-8, misleading the observers to conclude that the improvement shortly after the start of therapy resulted from plasmapheresis. In our cohort, 47 patients were not treated with plasmapheresis. The time courses of platelet count and levels of lactate dehydrogenase, haemoglobin, and creatinine during recovery from the disease were similar to the group that received plasmapheresis. Secondly, the three centres that treated patients with limited plasmapheresis (3-5 sessions) had similar or better outcomes than centres that continued plasmapheresis until platelet counts had increased to more than 100/nL. These results question the current recommendation to use plasmapheresis as a standard treatment in adults with enterohaemorrhagic E coli associated haemolytic uraemic syndrome26 27 and are in agreement with the experience of paediatric nephrologists, who encounter shiga toxin induced haemolytic uraemic syndrome more often than doctors treating adults and use plasmapheresis only rarely. During the current epidemic, 17 of the 90 children were treated with plasmapheresis (data not shown) and the outcome was good.

According to the recommendation of the German Society of Nephrology, fresh frozen plasma should be used for plasmapheresis.28 No plasmapheresis was carried out with albumin in the adults, whereas albumin was mainly used in the children. We cannot exclude the possibility that fresh plasma enhanced the disease process—for example, through further complement activation. This needs further research.

We believe that a randomised trial analysing supportive treatment plus limited plasmapheresis (3-5 sessions using albumin) compared with no additional treatment is necessary for clarification. However, from our experience during the outbreak in Germany and the experience of the paediatricians in the routine treatment of children with typical haemolytic uraemic syndrome, we believe that no benefit or only a marginal benefit will be found.

Glucocorticoids with plasmapheresis

Despite our finding that glucocorticoids had an effect on recovery of lactate dehydrogenase and creatinine levels and platelet counts, it was not significant and led to a higher number of patients requiring dialysis. As enterohaemorrhagic E coli associated haemolytic uraemic syndrome is an infectious disease, concomitant treatment with glucocorticoids might be harmful—an observation in keeping with a randomised trial from Italy, in which no benefit was found in children.29

Antibiotic treatment

The use of antibiotics to treat enterohaemorrhagic E coli infection is controversial. In theory antibiotic treatment may lead to higher toxicity through an intestinal Jarisch-Herxheimer reaction, with a massive release of shiga toxin through bacterial death during the prodromal phase of diarrhoea.30 This concept is supported by a mouse model with shiga toxin producing E coli, showing that treatment with fluoroquinolone resulted in a higher release of toxin and mortality.31 A report in 71 children with enterohaemorrhagic E coli O157:H7 induced diarrhoea described that five of nine children receiving antibiotics developed haemolytic uraemic syndrome compared with five of 62 children not receiving antibiotics.32 In our patient cohort, all but one medical centre did not use antibiotic treatment. In this centre patients were treated with a combination of at least two antibiotics sensitive to enterohaemorrhagic E coli (meropenem and ciprofloxacin and additionally rifaximin in patients on intensive care ward) after a diagnosis of haemolytic uraemic syndrome. None of the patients treated with antibiotics developed signs of toxic shock. Enterohaemorrhagic E coli was eradicated about eight days earlier than in the other centres and the rates of seizures and mortality were improved. None of the patients required intestinal surgery. In this centre the lower incidence of seizures might also be explained by a more aggressive use of prophylactic antiepileptics in patients with neurological symptoms. Nevertheless, the results are encouraging as they suggest that the production and potential drug induced release of shiga toxin might be irrelevant and that pre-emptive antibiotics do not worsen the clinical course of established haemolytic uraemic syndrome. Moreover, in contrast with previous studies the antibiotic strategy was more aggressive. Patients simultaneously received at least two (many even three) antibiotics effective against enterohaemorrhagic E coli, which might have contributed to the beneficial outcome. As antibiotics seem to improve, but definitely do not worsen, the course of the infection we believe that they are beneficial in the later stages of the disease when the prodromal phase with diarrhoea has nearly subsided. Mice infected with different enterohaemorrhagic E coli O157:H7 strains have shown lower mortality and weight loss if treated with rifampicin compared with placebo.33 Therefore it could be speculated that a suitable antibiotic combination strategy at the onset of bloody diarrhoea in (enteroaggregative) enterohaemorrhagic E coli infections might even prevent the development of haemolytic uraemic syndrome.

We believe that a randomised trial should be carried out to assess whether antibiotic treatment is beneficial in patients with enterohaemorrhagic E coli associated haemolytic uraemic syndrome.

Complement 5 inhibition

Eculizumab was used widely as a compassionate treatment (that is, outside the currently accepted indication for atypical haemolytic uraemic syndrome and paroxysmal nocturnal haematuria) during the outbreak in Germany. Evaluations of treatment effects were prone to bias by indication, as most of the centres treated the least sick patients with supportive care only, sicker patients with plasmapheresis, and the sickest with eculizumab. The treatment patterns in the different centres were not uniform; one larger hospital did not give eculizumab at all and in other hospitals it was administered only later during the outbreak. This enabled us to identify a control group of similarly sick patients to compare eculizumab treatment with plasmapheresis. This evaluation cannot be a substitute for a randomised controlled trial, but we believe that this is the only way to get an impression of the true effect of eculizumab treatment. Based on the data presented, patients treated with eculizumab did not improve significantly compared with a control group of patients with the same severity of haemolytic uraemic syndrome. Patients treated with eculizumab still developed new complications, such as seizure or requirement for ventilation, and in more than 40% of the patients plasmapheresis was continued after eculizumab had been started. Data on long term (6-12 months) renal and neurological follow-up in all patients treated with eculizumab will be required to assess the effect of this treatment strategy. In addition, the effect of eculizumab might be confounded because more than 98% of patients who received the drug were simultaneously treated with the antibiotic azithromycin for meningococcal prophylaxis. Recently, this antibiotic has been shown to eradicate O104:H4 quickly34 and does not lead to shiga toxin release in vitro.35

Lessons for future outbreaks

A major shortcoming of our analysis was that we did not carry out a randomised study. The large number of patients affected would have been ideal to test some of the questions and hypotheses about treatment strategies that existed at the start of the outbreak. As the outbreak began at the end of May and had almost finished by July, it was impossible to design a randomised trial and get approval within that time.

Conclusions and policy implications

Haemolytic uraemic syndrome associated with E coli O104:H4 is an acute self limited disease with a high percentage of patients requiring dialysis and ventilation and having severe neurological impairment. Our retrospective analyses question the benefit of plasmapheresis and concomitant glucocorticoid treatment in adults with enterohaemorrhagic E coli associated haemolytic uraemic syndrome. Contrary to current belief, antibiotics do not seem to worsen the clinical course in patients with established haemolytic uraemic syndrome, but may be of clinical benefit. Further prospective, randomised investigations of antibiotic treatment and its timing in future cases of (enteroaggregative) enterohaemorrhagic E coli associated haemolytic uraemic syndrome and even outbreaks with the O157:H7 strain are required. We observed no significant short term benefit of eculizumab treatment.

What is already known on this topic

  • Recommendations for the treatment of patients with haemolytic uraemic syndrome are based on small cases series and include plasmapheresis with or without glucocorticoids

  • Antibiotics are believed to be harmful and to induce the development of haemolytic uraemic syndrome and are therefore withheld in patients with enterohaemorrhagic Escherichia coli related infection

  • In few reported cases eculizumab was helpful in the treatment of patients with shiga toxin induced haemolytic uraemic syndrome

What this study adds

  • No benefit of plasmapheresis or glucocorticoid treatment was found for patients with enterohaemorrhagic E coli O104:H4 induced haemolytic uraemic syndrome, and prolonged treatment may even do more harm than good

  • An aggressive antibiotic treatment strategy was not harmful in patients with established haemolytic uraemic syndrome and might even be beneficial

  • No benefit of eculizumab could be found when short term outcome was compared with a patient group with similar severity of disease

Notes

Cite this as: BMJ 2012;345:e4565

Footnotes

  • We thank the doctors, nurses, and technicians of the participating hospitals for their work during the enterohaemorrhagic E coli outbreak; the primary care physicians who provided follow-up data on their patients; and the patients, who were keen to share the information and their experience to understand better this rare disease.

  • Collaborators not listed as authors: Hannover: Study group: Marianne Bergmann, Katharina Berhorst-Ziadi, Robert Diedrich, Sang Hi K Emden, Annika Hampel, Annika Müller, Nadine Richters, Sengül Samiri, Aleksej Sartison, and Frank Vetter (medical students), Franz C Bange (professor of medical microbiology), Svenja K Bahte (registrar), Philip Bintaro (senior house officer), Susanne Fleig (senior house officer), Wilfried Gwinner (professor of nephrology), Nils Hanke (senior house officer), Meike Heeren (assistant physician of neurology), Markus Hiss (consultant nephrologist), Silvia Linnenweber-Held (senior house officer), Alexander H. Lukasz (senior house officer), Saskia Merkel (registrar), Ansgar Reising (registrar), Roland Schmitt (registrar), Anke Schwarz (professor of nephrology), Sibylle von Vietinghoff (senior house officer), Annette Wagner (professor of nephrology), Hans Worthmann (senior house officer). Databank programming: Stephan Kaminski (IT and database consultant), Sang Hi K Emden (medical student and software designer), Kiel: Alexander Arlt (consultant gastroenterologist), Johannes Bethge (senior house officer), Burkhard Bewig (professor of pulmonology), Rainer Guenther (consultant gastroenterologist), Helge Hellriegel (consultant neurologist), Holger Kristen (consultant nephrologist), Bettina Möller (senior house officer), Stephan J Ott (senior house officer), Sabine Schubert (consultant microbiologist), Markus Seeger (consultant for sonography); partial data management: Konrad Aden (house officer). Lübeck: Inge Derad (consultant nephrologist), Stefanie Duderstadt (house officer), Henriette Füllgraf (consultant neurologist), Christian Haas(consultant nephrologist), Carsten Jankowiak (consultant gastroenterologist), Annette Kodal (house officer), Jan-Christian Ketel (consultant nephrologist), Volkhard Kurowski (consultant cardiologist), Walter Lehne (senior house officer), Markus Meier (consultant nephrologist), Christian Meinhardt (senior house officer), Philip Muck (senior house officer), Arik Sauer (consultant gastroenterologist), Klaus Schmidt (consultant gastroenterologist), Werner Solbach (professor of microbiology and director), Sven Süfke (consultant nephrologist), Gunther Weitz (consultant gastroenterologist), Stephan Werth (house officer), Sebastian Wolfrum (consultant cardiologist), Helge Züllich (senior house officer). Bielefeld: Rainer Valentin (consultant nephrologist). Bremen: Uwe Kuhlmann (nephrologist and director). Düsseldorf: Karen Eschweiler (senior house officer), Seher Kücükköylü (senior house officer). Gütersloh: Jens Klempin (consultant nephrologist and director). Hildesheim: Burkhard Kreft (professor of nephrology and director). Lüneburg: Torsten Kucharzik (professor of gastroenterology and director). Greifswald: Sebastian Fussek (medical student). Stade: Karl W Kroencke (consultant nephrologist).

  • Contributors: JM, MN, RS, JBe, JBü, MS, and TK contributed equally and were involved in all aspects of this study, including patient care, data collection, data handling, and writing the manuscript. JBe and JM were responsible for the databank and generation of the first drafts of graphics and tables. TFM helped to write the manuscript. JöB, JBr, RB, VB, RD, GD, KF, CG, JG, CH, FH, HH, SHR, BH, ML, JTK, UCK, JK, UK, HL, MPM, TNM, CM, CNG, JN, HP, LeR, LuR, JR, WR, AR, LCR, OS, BMWS, SaS, StS, KUvS, JS, SyS, SeS, AvdL, MV, KW, MW, and SZ were dedicated to patient care and reviewed the manuscript. HF, CH, UCK, UK, CM, LeR, LuR, AR, FS, StS, PW, and SZ helped with patient care and data collection and reviewed the manuscript. AG and MK collected the data.

  • Funding: This study received partial funding from the Deutsche Forschungsgesellschaft excellence cluster “Inflammation at interfaces” (recipient SS, No EXC 306).

  • Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: doctors from the following 10 centers (Medical School Hannover, Hannover; University Hospital Schleswig Holstein, Lübeck; Evangelic Hospital Gilead, Bielefeld; Bremerhaven Hospital Reinkenheide, Bremerhaven; University Hospital Münster, Münster; Red Cross Hospital Bremen, Bremen; Bremen Hospital-Mitte, Bremen; Hospital Oldenburg, Oldenburg; Diakonissenkrankenhaus Flensburg; Asklepios Clinic Hamburg Barmbek, Hamburg) participated in a single arm multicentre trial, sponsored by Alexion Pharmaceuticals; no support from any organisation for the submitted work; and no other relationships or activities that could appear to have influenced the submitted work.

  • Ethical approval: This study was approved by the ethical committees of the Hannover Medical School (No 1123-2011), University of Lübeck (No 11-103), and University of Kiel (No AZ A156/03).

  • Data sharing: No additional data available.

This is an open-access article distributed under the terms of the Creative Commons Attribution Non-commercial License, which permits use, distribution, and reproduction in any medium, provided the original work is properly cited, the use is non commercial and is otherwise in compliance with the license. See: http://creativecommons.org/licenses/by-nc/2.0/ and http://creativecommons.org/licenses/by-nc/2.0/legalcode.

References

Free via Creative Commons: Non-commercial use only Open access

Citations

THIS WEEK'S POLL