Leukemoid reactions complicating colitis due to clostridium difficile.

Background: We sought to describe the characteristics of patients who had Clostridium difficile colitis complicated by leukemoid reactions (total leukocyte count greater than 35 X [10.sup.9]/L) and to determine whether this complication is associated with higher morbidity or mortality than C difficile colitis without leukemoid reactions.

Methods: We performed a retrospective case series analysis of patients with a positive fecal assay for C difficile toxin and a peak leukocyte count greater than 35 X [10.sup.9]/L during 1998 and 1999. Twenty cases that met these criteria were compared with 65 randomly selected control patients (patients with a positive C difficile toxin and a peak leukocyte count less than 35 X [10.sup.9]/L).

Results: The mean peak leukocyte count was 52 [+ or -] 18.2 X [10.sup.9]/L ([+ or -] SD) in the case group and 14.9 [+ or -] 6.5 X [10.sup.9]/L in the control group. Patients with a leukemoid reaction had a lower temperature, a lower serum albumin level, and a higher hematocrit value. Multivariable logistic regression showed respiratory tract infection and lower temperature to be independent predictors of a leukemoid reaction. There were 10 deaths (50%) in the leukemoid reaction group and 5 deaths (7.7%) in the control group. All seven patients with a peak leukocyte count greater than 50 X [10.sup.9]/L died, compared with eight deaths (10.3%) among the remaining 78 patients whose peak leukocyte count was less than 50 X [10.sup.9]/L.

Conclusion: Patients with C difficile colitis and a leukocyte count greater than 35 X [10.sup.9]/L have a poor prognosis with a much higher mortality rate than patients who have C difficile colitis without a leukemoid reaction.

***********

Colitis due to Clostridium difficile is a common problem in hospitals nationwide, with the clinical spectrum ranging from asymptomatic carriage to toxic megacolon and death. Most cases are associated with recent antimicrobial administration and subsequent nosocomial acquisition, especially in severely ill patients. (1) Common clinical manifestations of C difficile colitis include abdominal discomfort, diarrhea, and fever, as well as leukocytosis, electrolyte derangements, and hypoproteinemia. (2,3) Leukocyte counts are typically elevated but may be depressed as well, both of which likely represent a response to C difficile colitis itself or to underlying infection. Total leukocyte counts greater than 25 to 30 X [10.sup.9]/L are referred to as leukemoid reactions. (4) Leukemoid reactions are typically neutrophilic and may occur with various infections, malignancy, tissue necrosis, severe hemorrhage, hemolysis, and burns. Extreme leukocytosis is mentioned occasionally as a clinical manifestation of C difficile colitis (2-5) but little exists in the literature directly addressing the potential association with leukemoid reactions and C difficile colitis. We retrospectively studied patients with C difficile colitis and coexistent leukocyte counts greater than 35 X [10.sup.9]/L to assess clinical characteristics and outcome.

Methods

We retrospectively analyzed data on all patients with a positive fecal assay for C difficile toxin who were admitted during a 2-year period to Miami Valley Hospital, an 800-bed, acute care, regional referral facility that is a major teaching affiliate of Wright State University School of Medicine in Dayton, Ohio. Patients were classified as cases if their total leukocyte count was greater than 35 X [10.sup.9]/L at any time during the hospitalization. We randomly selected 65 patients from a total of 314 patients during the same 2-year period with a positive fecal assay for C difficile toxin and a white blood cell count less than 35 X [10.sup.9]/L to serve as a control group (approximately every fifth patient from the total list of 314 patients was manually selected). The presence of C difficile toxin was determined with an enzyme immunoassay (premiere C difficile toxin assay; Meridian Corp, Cincinnati, OH). Excluded from the study were patients with any type of leukemia or recent use of filgrastim (a human granulocyte colony-stimulating factor [G-CSF]). Patient demographics, underlying infections, and major medical disorders were recorded. Vital signs and symptoms were analyzed, as well as the following laboratory data: total leukocyte count, absolute neutrophil count, hematocrit, platelet count, albumin, potassium, sodium, and creatinine. These data were generally, but not invariably, recorded on the day of the peak leukocyte count. Recent or current antibiotic administration was recorded when possible. In addition, drug or surgical therapy, length of hospital stay, and in-hospital mortality rates were recorded.

Statistical Analysis

Means and standard deviations (SDs) are reported for continuous variables and percentages for categorical variables. The independent-samples t test was used for group comparisons where the dependent variable was continuous, and the [chi square] test or Fisher's exact test, where necessary, was used for categorical variables. For these univariate analyses, inferences were made at the 0.05 level. To identify independent predictors of leukemoid reaction (ie, characteristics that are statistically significant after other variables are controlled), variables related to the outcome (leukemoid reaction) at P < 0.05 on a univariate analysis were entered into a multivariable logistic regression (MLR) equation. Characteristics with P < 0.05 on MLR analysis were classified as independent predictors.

Results

Over the study period, there were 334 patients with positive fecal assays for C difficile. Of these, a total of 20 patients (6%) had a peak white blood cell count exceeding 35 X [10.sup.9]/L on at least one occasion. Antibiotic administration during the current hospitalization or the previous several weeks was documented in all 20 patients with leukemoid reaction (case group) and 59 (91%) of the 65 patients in the control group. Diarrhea was recorded in 19 (95%) of the case patients and in 58 (89%) of the control patients; abdominal pain was noted in 6 case patients (30%) and in 22 control patients (34%). There were no statistically significant differences between the two groups with regard to underlying medical illnesses, class of antibiotic exposure, use of tube feedings, mechanical ventilation, or history of recent surgery (Table 1). Patients with leukemoid reactions, however, were more likely to have had a respiratory tract infection than patients in the control group: 16 of 20 (80%) versus 23 of 65 (35%) (P = 0.0004).

The mean ([+ or -] SD) peak total leukocyte count for case patients was 52 [+ or -] 18.2 X [10.sup.9]/L versus 14.9 [+ or -] 6.5 X [10.sup.9]/L for the 65 control patients. The mean age was 73.1 [+ or -] 16.1 years for case patients and 64.1 [+ or -] 19 years for controls (P = 0.06). There were 10 men and 10 women in the case group and 25 men (39%) and 40 women (61%) in the control group (P = 0.36). The mean temperature was 98.8 [+ or -] 1.6[degrees]F for case patients and 100.1 [+ or -] 1.6[degrees]F for control patients (P = 0.003). The mean hematocrit value was 38.8% [+ or -] 6.1% for cases and 34.2% [+ or -] 5.5% for controls (P = 0.002). The mean serum albumin value was 2.65 [+ or -] 0.74 g/dL for case patients and 3.04 [+ or -] 0.68 g/dl for control patients (P = 0.043).

To determine independent predictors of leukemoid reactions, the five characteristics with a P value of 0.05 or below (as well as age, P = 0.06) on univariate analysis were entered into a multivariable logistic regression equation: temperature, hematocrit, albumin, diabetes mellitus, and respiratory infection. Respiratory infection (P = 0.01) and temperature (P = 0.04) were independent predictors of leukemoid reactions. Patients with a leukemoid reaction were more likely to have had a respiratory tract infection at some time during their hospitalization and a lower temperature than control patients.

Treatment was similar in both groups: metronidazole was administered to 16 case patients (80%) and to 59 control patients (91%) and oral vancomycin was administered to 9 case patients (45%) and to 16 control patients (25%). No patients in either group had surgical intervention.

There were 10 deaths among the 20 patients with leukemoid reactions (50%) compared with 5 deaths among the 65 patients in the control group (7.7%) (P < 0.001). All 7 patients with a peak leukocyte count greater than 50 X [10.sup.9]/L died (100%), compared with 8 (10%) of the remaining 78 patients and controls whose peak leukocyte count was less than 50 X [10.sup.9]/L (P < 0.001). Table 2 shows the causes of death obtained from the medical records of the 10 case patients who died. The mean length of hospital stay was 24.3 [+ or -] 22.7 days for surviving case patients versus 18.6 [+ or -] 22.6 days for surviving control patients (P = 0.47).

Discussion

Leukocytosis is a common manifestation of colitis due to C difficile, (6-9) but leukemoid reactions are uncommon, occurring in only 6% of patients with a positive fecal assay. Leukemoid reactions have been reported to occur in a variety of situations such as systemic infection, malignancy, hemorrhage, diabetic ketoacidosis, cirrhosis, burns, and tissue necrosis. (10-12) The release of neutrophils and their precursors from the marrow is largely due to interleukin-3 and G-CSF, which leads to enhanced neutrophil production. (4) Conditions that cause systemic inflammation (eg, infection, neoplasm, inflammatory disorders) are often associated with elevated levels of G-CSF, and leukocytosis can be significant. (4) In this study, patients in the control group manifested leukocytosis, which could have been due to underlying acute illness as well as C difficile colitis. Patients with leukemoid reactions were similar to control patients with regard to underlying illness, infection (except for respiratory tract infection), and recent antibiotic use (Table 1). The reasons for leukemoid reactions in patients with C difficile colitis are unclear but are likely related to a significant systemic inflammatory response causing stimulation of the marrow. Patients with severe pseudomembranous colitis such as toxic megacolon and patients who have surgical intervention are more likely to have a leukemoid reaction. (1, 13)

Leukemoid reactions have been reported in postpartum women with Clostridium sordellii sepsis and are characterized by striking degrees of leukocytosis (66-200 X [10.sup.9]/L), in association with hypoalbuminemia and hemoconcentration. (14) The marked inflammatory response leading to leukemoid reactions in these patients may have been due to toxins antigenically similar to the A and B toxins produced by C difficile. (14, 15) A case of fatal C difficile cellulitis in an infant characterized by refractory sepsis and a leukocyte count of 61.8 X [10.sup.9]/L was reported and postulated to be due partly to the effects of toxins. (16) Arnon et al (17) administered purified C difficile toxins A and B through the intravenous and intraperitoneal route to rhesus monkeys to assess for systemic inflammatory effects. The animals had a significant systemic inflammatory response that included marked leukemoid reactions, hypoalbuminemia, elevated hematocrit, and a decrease in body temperature, eventually leading to death in all the subjects. Autopsy studies revealed capillary endothelial cell swelling as well as ascites and pleural effusions in most animals. The authors concluded that C difficile toxins may impair cellular integrity, which could account for vascular permeability changes and cell death.

In our study, comparison with the control group showed that patients with leukemoid reactions and statistically significant alterations in several clinical parameters, including lower body temperature, lower serum albumin concentration, and higher hematocrit values (Table 1). These findings are similar to the findings reported by Arnon et al (17) as well as the other case reports mentioned. (14, 16) Multivariable analysis in our study revealed lower body temperature to be an independent predictor of a leukemoid reaction in patients with C difficile colitis. The reason for this association is unclear, but it may be a more vigorous systemic inflammatory response leading to temperature dysregulation. Hypothermia is more common in patients with severe sepsis and is associated with a higher mortality rate than fever. (18) Capillary leak (17) as well as diarrhea and increased insensible fluid losses may account for the elevated hematocrit of case patients in our study. Hypoalbuminemia has been associated with higher in-hospital mortality rates in general medical patients (19) and is a prognostic factor for increased mortality in patients with C difficile colitis. (20) Hypoalbuminemia is a "negative acute phase reactant," and serum albumin typically declines in the setting of serious illness due to decreased hepatic production. (19) In addition, poor nutritional intake and diarrhea leading to protein loss may contribute to hypoalbuminemia. Protein-losing enteropathy due to colonic inflammation and diarrhea may lead to hypoalbuminemia in patients with severe C difficile colitis. (21)

An interesting finding in this study is that respiratory infection was noted to be an independent predictor of the presence of a leukemoid reaction in patients with C difficile colitis. The reasons for this are unclear, since we found no statistically significant differences among those in either group with regard to class of antibiotic exposure. It is possible that patients with respiratory tract infection had lower arterial oxygen levels, which could have been associated with more severe acute illness; however, we were unable to address this in our study.

Patients with leukemoid reactions had an average length of stay approximately 6 days longer than patients without leukemoid reactions. This was not significant from a statistical standpoint, but is potentially significant from an economic perspective, given the costs incurred for 1 day in the hospital. Also, the incidence of nosocomial infection increases with prolonged hospitalization, which may lead to further morbidity and elevated costs. (21,22) Perhaps the most important finding of our study is the much higher mortality rate among patients with leukemoid reactions, especially those with a peak leukocyte count greater than 50 X [10.sup.9]/L; all of them died, typically of sepsis (Table 2). The higher death rate among patients with leukemoid reactions may be due to refractory sepsis leading to cardiovascular collapse. Also, case patients were older than control patients by a mean of 9 years, which may also account for the higher mortality rate in the leukemoid reaction group. Another potential reason for the increased mortality rates in the leukemoid reaction group is the development of fulminant colitis or "toxic megacolon," which is associated with a high mortality rate and occasional need for surgical intervention. (13) However, patients with fulminant pseudomembranous colitis are often severely ill and have an operative mortality rate of up to 60%. (13) Of note, no patients with leukemoid reactions in this series had operative intervention. Whether this would have altered the mortality rate is unclear, since this group of patients may not be optimal surgical candidates because of comorbid conditions. However, it is difficult to definitively know which patients in this series may have had "toxic megacolon," since radiographic and endoscopic studies were not consistently available and no patients had surgery to document this condition.

There are limitations of our study. The relatively small sample size may limit the precision of our data. In addition, the retrospective nature of our study may have allowed us to overlook other variables that could have led to leukemoid reactions, such as undocumented nosocomial infections or hemorrhage. Also, some of the patients in the control group, despite a positive fecal assay for C difficile, may have been carriers without symptomatic colitis, thus accounting for less severe illness and lower mortality rates. Approximately 90% of patients in each group were reported to have diarrhea, however, which makes asymptomatic carriage less likely. Overall, however, our data suggest that patients with a positive fecal assay for C difficile complicated by a leukemoid reaction have a more severe systemic illness that leads to prolonged hospitalization and significantly higher mortality rates. Patients with C difficile colitis and a leukemoid reaction should be monitored for evidence of sepsis and development of toxic megacolon. In addition to aggressive, supportive medical care, surgical intervention should be considered in an attempt to improve outcome. More studies are needed to further characterize the clinical features of patients with C difficile colitis and leukemoid reactions and to develop optimal treatment strategies.

Table 1. Comparison of patients with and without leukemoid reaction (WBC
Count >35,000/[mm.sup.3])

                           Leukemoid reaction      No leukemoid reaction
Characteristic                group (n = 20)           group (n = 65)

Age, mean [+ or -]            73.1 [+ or -] 16.2        64.1 [+ or -]
  SD[dagger] (years)                                      19.0
Female (%)                    50.0                      61.5
Medical history (%)
  Diabetes mellitus           10.0                      32.3
  Renal failure               25.0                      15.4
  Cancer                      20.0                      12.3
  Congestive heart            35.0                      24.6
    failure
  Cerebrovascular              5.0                      13.8
    accident
Recent surgery (%)            30.0                      40.0
Admitted to ICU (%)*          50.0                      39.1
Mechanical ventilation        27.8                      29.7
  (%)*
Recent antibiotic use        100.0                      90.8
  (%)
  Cephalosporin               30.0                      29.2
  Quinolone                   30.0                      49.2
  Aminoglycoside               5.0                       0.0
  Clindamycin                  5.0                       4.6
  Penicillin                  25.0                      33.8
  Macrolide                    0.0                       9.2
  Vancomycin                  15.0                      20.0
  Tetracycline                 0.0                       1.5
Underlying infection
  (%)
  Respiratory tract           80.0                      35.4
  Abdominal                    5.0                       3.1
  Urinary tract                5.0                      13.8
  Soft tissue                  0.0                      12.3
  Osteomyelitis                0.0                       1.5
  Bacteremia                   5.0                      10.8
Surgical prophylaxis          30.0                      26.2
  (%)
Tube feeding (%)              25.0                      21.5
Symptoms (%)
  Diarrhea                    95.0                      89.2
  Abdominal pain              30.0                      33.8
  Vomiting                     5.0                      23.1
  Bloody stool                 5.0                       7.7
Temperature, mean             98.8 [+ or -] 1.6        100.1 [+ or -]
  [+ or -] SD                                            1.6
  (Farenheit)[dagger]
Platelet count, mean     332,000 [+ or -]          325,000 [+ or -]
  [+ or -] SD              144,000/[mm.sup.3]        168,000/[mm.sup.3]
Hematocrit, mean              38.8 [+ or -] 6.1         34.2 [+ or -]
  [+ or -] SD (%)                                        5.5
Creatinine, mean               1.93 [+ or -] 1.68        1.33 [+ or -]
  [+ or -] SD (mg/dl)                                    1.44
Albumin, mean [+ or -]         2.65 [+ or -] 0.74        3.04 [+ or -]
  S[D.sup.||] (mg/dl)                                    0.68
Potassium, mean                4.47 [+ or -] 1.39        4.00 [+ or -]
  [+ or -] SD (mEq/L)                                    0.82
Sodium, mean [+ or -]        136 [+ or -] 6            138 [+ or -] 6
  SD (mEq/L)
Treatment (%)
  Metronidazole               80.0                      90.8
  Vancomycin                  45.0                      24.6
Endoscopic intervention       15.0                       9.5
  (%)[double dagger]
Length of hospital            24.3 [+ or -] 22.7        18.7 [+ or -]
  stay, mean [+ or -]                                     22.6
  SD (days)
Survival (%)                  50.0                      92.3

                                 P value
Characteristic           Univariate  Multivariable

Age, mean [+ or -]        0.06          0.50
  SD[dagger] (years)
Female (%)                0.36
Medical history (%)
  Diabetes mellitus       0.05          0.45
  Renal failure           0.33
  Cancer                  0.46
  Congestive heart        0.36
    failure
  Cerebrovascular         0.44
    accident
Recent surgery (%)        0.42
Admitted to ICU (%)*      0.41
Mechanical ventilation    0.87
  (%)*
Recent antibiotic use     0.33
  (%)
  Cephalosporin           0.95
  Quinolone               0.13
  Aminoglycoside          0.24
  Clindamycin             1.00
  Penicillin              0.46
  Macrolide               0.33
  Vancomycin              0.75
  Tetracycline            1.00
Underlying infection
  (%)
  Respiratory tract       <.001         0.01
  Abdominal               0.56
  Urinary tract           0.44
  Soft tissue             0.19
  Osteomyelitis           1.00
  Bacteremia              0.67
Surgical prophylaxis      0.73
  (%)
Tube feeding (%)          0.76
Symptoms (%)
  Diarrhea                0.67
  Abdominal pain          0.75
  Vomiting                0.10
  Bloody stool            1.00
Temperature, mean         0.003         0.04
  [+ or -] SD
  (Farenheit)[dagger]
Platelet count, mean      0.85
  [+ or -] SD
Hematocrit, mean          0.002         0.07
  [+ or -] SD (%)
Creatinine, mean          0.12
  [+ or -] SD (mg/dl)
Albumin, mean [+ or -]    0.043         0.12
  S[D.sup.||] (mg/dl)
Potassium, mean           0.16
  [+ or -] SD (mEq/L)
Sodium, mean [+ or -]     0.46
  SD (mEq/L)
Treatment (%)
  Metronidazole           0.24
  Vancomycin              0.08
Endoscopic intervention   0.44
  (%)[double dagger]
Length of hospital        0.47
  stay, mean [+ or -]
  SD (days)
Survival (%)             <0.0001

* Data available for 18 patients with leukemoid reaction and 64 of those
without.
([dagger]) Data available for 18 patients with leukemoid reaction and 60
without.
** Data available for 18 patients with leukemoid reaction and 51
without.
([double dagger]) Data available for 63 patients without leukemoid
reaction.
([section]) Data for survivors only: 10 with leukemoid reaction and 60
without.

Table 2. Causes of death in patients with leukemoid reactions

          Peak               WBC Before
Patient   WBC                Death*             Cause of death

 1        50,800/[mm.sup.3]  50,800/[mm.sup.3]  Sepsis
 2       105,300             66,000             Sepsis
 3        53,100             29,500             Sepsis
 4        38,600             38,600             Respiratory failure
 5        69,600             17,100             Liver failure
 6        75,200             56,100             Renal failure
 7        41,500             15,700             Sepsis
 8        83,900             81,000             Squamous cell cancer
 9        48,100             44,300             Sepsis
10        53,900             52,800             Stroke

*The last WBC count available before death (range 0 to 3 days).

Accepted July 8, 2002.

References

1. Grundfest-Broniatowski S, Quader M, Alexander F, et al. Clostridium difficile colitis in the critically ill. Dis Colon Rectum 1996;39:619-623.

2. Kelly CP, LaMont JT. Clostridium difficile infection. Annu Rev Med 1998;49:375-390.

3. Fekety R, Shah AB. Diagnosis and treatment of Clostridium difficile colitis. JAMA 1993;269:71-79.

4. Bagby GC. Leukopenia and leukocytosis, in Goldman L, Bennett JC (eds): Cecil's Textbook of Medicine. Philadelphia, W.B. Saunders Co., 2000, ed 21, pp 919-933.

5. Counihan TC, Roberts PL. Pseudomembranous colitis. Surg Clin North Am 1993;73:1063-1075.

6. Kelly CP, Pothoulakis C, LaMont JT. Clostridium difficile colitis. N Engl J Med 1994;330:257-262.

7. Anand A, Bashey B, Mir T, et al. Epidemiology, clinical manifestations, and outcome of Clostridium difficile colitis. Am J Gastroenterol 1994;89:519-523.

8. Taege AJ, Adal KA. Clostridium difficile diarrhea and colitis: A clinical review. Cleve Clin J Med 1999;66:503-507.

9. Bartlett JG. How to identify the cause of antibiotic-associated diarrhea. J Crit Illn 1994;9:1063-1067.

10. McKee LC. Excess leukocytosis (leukemoid reactions) associated with malignant diseases. South Med J 1985;78:1475-1482.

11. Burris AS. Leukemoid reaction associated with severe diabetic ketoacidosis. South Med J 1986;79:647-648.

12. Marinella MA. Extreme leukemoid reaction associated with retroperitoneal hemorrhage. Arch Intern Med 1998;158:300-301.

13. Medich DS, Lee KK, Simmons RL, et al. Laparotomy for fulminant pseudomembranous colitis. Arch Surg 1992;127:847-853.

14. McGregor JA, Soper DE, Lovell G, et al. Maternal deaths associated with Clostridium sordelli infection. Am J Obstet Gynecol 1989;161:987-994.

15. Popoff MR. Purification and characterization of Clostridium sordellii lethal toxin and cross-reactivity with Clostridium difficile cytotoxin. Infect Immun 1987;55:35-43.

16. Katner HP, Pankey GA, Bonis SL. Fatal Clostridium difficile cellulitis. Pediatr Infect Dis J 1987;6:294-295.

17. Amon SS, Mills DC, Day PA, et al. Rapid death of infant rhesus monkeys injected with Clostridium difficile toxins A and B: Physiologic and pathologic basis. J Pediatr 1984;104:34-40.

18. Zimmerman JL, Taylor RW. Sepsis and septic shock, in Civetta JM, Taylor RW, Kirby RR (eds): Critical Care. Philadelphia, Lippincott-Raven, 1997, ed 3, pp 405-412.

19. Marinella MA, Markert RJ. Admission serum albumin and length of hospitalization in elderly patients. South Med J 1998;91:851-854.

20. Ramaswamy R. Grover H, Corpuz M, et al. Prognostic criteria in Clostridium difficile colitis. Am J Gastroenterol 1996;91:460-464.

21. Rybolt AH, Bennett RG, Laughon BE, et al. Protein-losing enteropathy associated with Clostridium difficile infection. Lancet 1989;1:1353-1355.

22. Marinella MA, Pearson C, Chenoweth C. The stethoscope: A potential source of nosocomial infection? Arch Intern Med 1997;157:786-789.

RELATED ARTICLE: Key Points

* Leukemoid reactions in patients with Clostridium difficile colitis are associated with lower body temperature and hypoalbuminemia.

* A respiratory tract infection in patients with C difficile colitis is an independent predictor of the presence of a leukemoid reaction.

* Patients with leukemoid reactions in association with C difficile colitis have a higher mortality rate than patients who have C difficile colitis without leukemoid reactions.

* The mortality rate for C difficile colitis with a peak leukocyte count greater than 50 X [10.sup.9]/L is exceedingly high, being 100% in this series.

Mark A. Marinella, MD, Steven D. Burdette, MD, Roger Bedimo, MD, and Ronald J. Markert, PHD

From the Department of Internal Medicine, Wright State University School of Medicine, Dayton, OH.

Reprint requests to Mark A. Marinella, MD, 33 W. Rahn Road #201, Dayton, OH 45429. Email: [e-mail]