Original Studies
Factors Affecting Time to Death From Start of Treatment Among Children Succumbing to Bacterial Meningitis Irmeli Roine, MD, PhD,* Tuula Pelkonen, MD, PhD,†‡§ Luis Bernardino, MD,‡ Manuel Leite, MD,‡ Matti Kataja, PhD,¶ Anne Pitkäranta, MD, PhD,†‖ and Heikki Peltola, MD, PhD§ Background: Many risks of death in childhood bacterial meningitis are well-identified, but factors influencing survival time have received less attention. Better understanding of this issue could help explain why adjuvant medications have performed unevenly in different trials. Methods: In a post hoc analysis of prospectively collected data from a large bacterial meningitis treatment trial in Luanda, Angola, we compared time to death after initiation of antimicrobial treatment among 206 children with etiology and other patient characteristics. The risks of dying very quickly (0–4 hours), quickly (4–8 hours) or after longer periods were analyzed by logistic regression. Results: Median time to death was 18.5 hours, half the time in Streptococcus pneumoniae (11.8 hours) compared with Haemophilus influenzae (26.8 hours) meningitis. Of all deaths caused by pneumococcal or H.influenzae meningitis, 42% versus 16%, respectively, occurred within the first 8 hours. In addition, patients who succumbed within 8 hours, unlike those dying later, had a short disease history, shock, hypoglycemia and poor cerebrospinal fluid white cell response. Conclusions: Time to death in Angola is so short that hardly anything, except perhaps modern intensive care, is likely to improve outcome in a patient with meningitis, especially the pneumococcal disease. Key Words: bacterial meningitis, outcome, mortality, time to death, time of survival (Pediatr Infect Dis J 2014;33:789–792)
C
hildhood bacterial meningitis is still prevalent in areas where vaccines are unavailable or underused because of economic constraints.1 Worse outcomes, death, hearing loss and/or permanent neurologic sequelae are associated with greater severity of illness on admission and often late arrival for treatment.2 In countries like Angola, mortality can approach 50% despite adequate antibiotic treatment.3 Although many risks of death in meningitis have been well-identified,4,5 factors influencing time to death have received little attention. This is an important issue because it may partly explain why adjuvant medications that perform well in 1 setting
fail in others where patients die shortly after the start of treatment. Identifying factors associated with early deaths would also help determine whether management was optimal. For example, in a pediatric study from Taiwan in 1998,6 septic shock was the cause of death in two-thirds of the patients who succumbed within the first 3 days, suggesting that more attention should be paid to the active prevention and management of this deadly condition. In this study, we investigated factors associated with the time interval to death after institution of antimicrobial treatment in a large series of pediatric bacterial meningitis in Luanda, Angola. There was special focus on deaths that occurred within the first 4 to 8 hours, when a treatment has the shortest time, and thus probably the smallest chance, of improving outcome of the patient.
METHODS This was a post hoc, descriptive analysis of prospectively collected data from a randomized, double-blind clinical trial on 723 children with bacterial meningitis, which examined the effects of cefotaxime (250 mg/kg/d) infusion and high-dose oral acetaminophen in the Pediatric Hospital of Luanda, Angola.7 It was approved by the hospital’s Ethics Committee and registered (ISRCTN62824827). The Ethics Committee declined the use of adjuvant corticosteroids; instead all patients received oral glycerol. Fluids were not restricted. All patients aged 2 months to 12 years with probable bacterial meningitis7 and their guardian’s informed consent were enrolled during July 2005 to June 2008. The present study included the patients with confirmed bacterial meningitis7 (n = 553). When a child died, the time in hours between the initiation of treatment and moment of death was registered in the patient’s chart. On admission, severe respiratory distress (= visible chest indrawing) was found in 67/553 (12%) patients, severe malnutrition (= middle upper arm circumference below 2 standard deviations) in 138/550 (25%) and some underlying condition (mostly frequently sickle cell disease and chronic suppurative otitis media) in 60/553 (11%) of the patients.
Statistical Analysis Accepted for publication January 9, 2014. From the *Faculty of Medicine, University Diego Portales, Santiago, Chile; †Institute of Clinical Medicine, University of Helsinki, Helsinki, Finland; ‡David Bernardino Children’s Hospital, Luanda, Angola; §Children’s Hospital, Helsinki University Central Hospital, and Helsinki University; ¶National Institute for Health and Welfare; and ‖Department of Otorhinolaryngology, Helsinki University Central Hospital, and Helsinki University, Helsinki, Finland. This work was supported by the Sigrid Juselius and the Paediatric Research Foundations, Helsinki, Finland. H.P. is a scientific consultant of the Serum Institute of India. The authors have no other funding or conflicts of interest to disclose. Address for correspondence: Irmeli Roine, MD, PhD, Los Misioneros 2237, 7520179 Santiago, Chile. E-mail: [email protected]. Copyright © 2014 by Lippincott Williams & Wilkins ISSN: 0891-3668/14/3308-0789 DOI: 10.1097/INF.0000000000000350
Time to death was assessed in relation to several other clinical and laboratory variables using Spearman correlation, the contingency table and the Kruskall-Wallis test, whichever was appropriate. It was tested both as a continuous variable, and in periods of 0–4, 5–8, 9–24, 25–120 (approximately days 2–5 of treatment) and >120 hours (after day 5). Logistic regression analysis was used to examine whether the known risks of death (independent variables) were specifically related to a certain period to death (dependent variable, each period compared with the survivors). The strength of associations is expressed as odds ratios (OR) with 95% confidence intervals (CI). The quantitative data are presented as medians with interquartile range (IQR). P 0.05) was observed with age, plasma hemoglobin, blood white cell count or several other patient and laboratory variables listed in Table, Supplemental Digital Content 1, http://links.lww.com/INF/B874. The treatment alternatives in the original trial7 also influenced time to death. Administration of the antibiotic as an infusion together with high-dose oral paracetamol tended (P = 0.06) to prolong time to death (median, 58 hours; IQR, 111; n = 47) compared with administration of the antibiotic as a bolus with an
TABLE 1. Differences in Admission Findings According to Time to Death in 206 Non-survivors of Childhood Bacterial Meningitis
Variable Age, months Underlying illness‡ Preadmission illness, days¶ Seizures during admission Glasgow Coma Score║ Systolic blood pressure, mm Hg Severe respiratory distress Severe dehydration MUAC**, cm CSF-glucose mg/dL§§ Etiology‡‡ S. pneumoniae H. influenzae N. meningitidis Other‡‡
0–4 h*
5–8 h
9–24 h
25–120 h
> 120 h
N
n = 28
n = 30
n = 56
n = 65
n = 27
P Value
206 206 206 206 205 195 206 206 203 202 170 82 62 7 19
15 [59]† 1 (4)§ 3 [4] 21 (75) 9 [4] 90 [30] 9 (32) 2 (7) 13 [3] 9 [13]
11 [9] 2 (7) 3 [2] 21 (70) 8 [3] 90 [25] 12 (40) 1 (3) 13.5 [3] 9 [9]
9·5 [14] 3 (5) 6·5 [6] 39 (70) 8 [5] 90 [23] 8 (14) 2 (4) 13 [2] 9 [10]
12 [19] 12 (18) 6 [7] 41 (63) 9 [5] 90 [18] 11 (17) 4 (6) 13 [3] 10 [20]
32 [72] 7 (26) 7 [4] 11 (41) 9 [6] 100 [10] 3 (11) 0 (−) 14 [2] 17 [16]
19 (68) 2 (7) 3 (11) 1 (4)
16 (53) 8 (27) 1 (3) 1 (3)
22 (39) 21 (38) 2 (4) 5 (9)
18 (28) 24 (40) 0 (0) 10 (15)
7 (26) 7 (26) 1 (4) 2 (7)
0.08 0.01 0.005 0.06 0.45 0.39 0.02 0.66 0.27 0.09 0.009 0.007 0.02 0.18 0.19
Contingency table and Kruskall-Wallis test. *Hours after initiation of cefotaxime treatment. †Median value with IQR in brackets. ‡Sickle cell disease (n = 12), prior episode of bacterial meningitis (n = 3), chronic otitis media (n = 3), prior head trauma or surgery (n = 3), prolonged cough and mother with tuberculosis (n = 1), fever for 1 year (n = 1), sick for 1 month (n = 1), cleft lip and palate (n = 1). §In parentheses, percentages. ¶Symptoms and signs suggestive of bacterial meningitis before admission. ║Range 3–15 (best). **Middle upper arm circumference. ‡‡Of the 51 bacteria tested for cefotaxime sensitivity, 49 (96%) were sensitive and 2 (4%) resistant: 1 H. influenzae patient who died after 123 hours and 1 Klebsiella pneumoniae patient who died after 92 hours. ‡‡K. pneumoniae, n = 4; Streptococcus Gr. C, n = 3; Proteus mirabilis, n = 3; Esherichia.coli, n = 2; Salmonella sp., n = 1; Pseudomonas aeruginosa, n = 1; P. vulgaris, n = 1; Enterobacter aglomerans, n = 1; Citrobacter freundii, n = 1. §§Cerebrospinal fluid from diagnostic tap on admission.
FIGURE 1. Time to death according to etiology of 206 children with bacterial meningitis.
790 | www.pidj.com
© 2014 Lippincott Williams & Wilkins
© 2014 Lippincott Williams & Wilkins
Logistic regression model. *Compared with 347 survivors of childhood bacterial meningitis. † Range 3–15 (best) ‡A blood pressure 120 hours and in 0/347 (0%) of survivors (P = 0.03). §Middle upper arm circumference below –2SD indicates severe malnutrition. ¶CSF from diagnostic tap on admission. ║Confirmed in 452 patients. In addition to etiologies mentioned in the Table, there were 49 patients with meningococcal meningitis. ** K. pneumoniae 5, Streptococcus Gr. B 1, C 4, D 1, G 1, Proteus mirabilis 5, Salmonella sp. 4, E. coli 3, Stapylococcus aureus 2, P. aeruginosa 1, E. aglomerans 1, C. freundii 1 and not identified but with growth 3.
0.68 0.80 0.21 0.46–3.35 0.33–2.38 0.57–13.4 0.50–1.74 0.60–1.95 2.03–11.9 2.82 0.56 0.83 2.16–6.45 0.03–0.47 0.10–6.88 5.59 0.11 0.86
0.0004 0.003 0.89
1.24–6.45 0.24–1.33 0.10–6.59
0.01 0.19 0.86
1.39 0.91 2.30
0.75–2.55 0.50–1.67 07.8–6.76
0.29 0.76 0.13
0.93 1.08 4.93
0.83 0.81 0.0004
1.24 0.88 2.76
0.005 0.94 0.42 0.92 0.39 0.93 0.49 1.44–7.64 0.44–2.15 0.47–5.60 0.35–2.62 0.55–4.71 0.24–4.79 0.06–3.77 1.48–4.89 1.39–4.16 1.27–5.92 0.73–2.55 1.22–4.82 0.10–1.84 0.51–3.30 3.26 3.39 8.97 2.82 0.55 0.96 1.96 1.44–7.64 1.75–10.2 2.61–15.6 1.29–6.20 1.97–13.9 1.72–11.5 1.72–11.6
Glasgow Coma Score < 8† Seizures during admission Severe respiratory distress Blood pressure < 90 mm Hg‡ MUAC < -2 standard deviations§ CSF¶ white cell count < 100/μL Blood glucose < 40 g/dL Etiology‖ S. pneumoniae H. influenzae Other**
550 553 553 525 550 551 548 452 184 187 32
3.31 4.23 6.38 3.08 5.23 4.46 4.47
0.005 0.003
Factors Affecting Time to Death From Start of Treatment Among Children Succumbing to Bacterial Meningitis Irmeli Roine, MD, PhD,* Tuula Pelkonen, MD, PhD,†‡§ Luis Bernardino, MD,‡ Manuel Leite, MD,‡ Matti Kataja, PhD,¶ Anne Pitkäranta, MD, PhD,†‖ and Heikki Peltola, MD, PhD§ Background: Many risks of death in childhood bacterial meningitis are well-identified, but factors influencing survival time have received less attention. Better understanding of this issue could help explain why adjuvant medications have performed unevenly in different trials. Methods: In a post hoc analysis of prospectively collected data from a large bacterial meningitis treatment trial in Luanda, Angola, we compared time to death after initiation of antimicrobial treatment among 206 children with etiology and other patient characteristics. The risks of dying very quickly (0–4 hours), quickly (4–8 hours) or after longer periods were analyzed by logistic regression. Results: Median time to death was 18.5 hours, half the time in Streptococcus pneumoniae (11.8 hours) compared with Haemophilus influenzae (26.8 hours) meningitis. Of all deaths caused by pneumococcal or H.influenzae meningitis, 42% versus 16%, respectively, occurred within the first 8 hours. In addition, patients who succumbed within 8 hours, unlike those dying later, had a short disease history, shock, hypoglycemia and poor cerebrospinal fluid white cell response. Conclusions: Time to death in Angola is so short that hardly anything, except perhaps modern intensive care, is likely to improve outcome in a patient with meningitis, especially the pneumococcal disease. Key Words: bacterial meningitis, outcome, mortality, time to death, time of survival (Pediatr Infect Dis J 2014;33:789–792)
C
hildhood bacterial meningitis is still prevalent in areas where vaccines are unavailable or underused because of economic constraints.1 Worse outcomes, death, hearing loss and/or permanent neurologic sequelae are associated with greater severity of illness on admission and often late arrival for treatment.2 In countries like Angola, mortality can approach 50% despite adequate antibiotic treatment.3 Although many risks of death in meningitis have been well-identified,4,5 factors influencing time to death have received little attention. This is an important issue because it may partly explain why adjuvant medications that perform well in 1 setting
fail in others where patients die shortly after the start of treatment. Identifying factors associated with early deaths would also help determine whether management was optimal. For example, in a pediatric study from Taiwan in 1998,6 septic shock was the cause of death in two-thirds of the patients who succumbed within the first 3 days, suggesting that more attention should be paid to the active prevention and management of this deadly condition. In this study, we investigated factors associated with the time interval to death after institution of antimicrobial treatment in a large series of pediatric bacterial meningitis in Luanda, Angola. There was special focus on deaths that occurred within the first 4 to 8 hours, when a treatment has the shortest time, and thus probably the smallest chance, of improving outcome of the patient.
METHODS This was a post hoc, descriptive analysis of prospectively collected data from a randomized, double-blind clinical trial on 723 children with bacterial meningitis, which examined the effects of cefotaxime (250 mg/kg/d) infusion and high-dose oral acetaminophen in the Pediatric Hospital of Luanda, Angola.7 It was approved by the hospital’s Ethics Committee and registered (ISRCTN62824827). The Ethics Committee declined the use of adjuvant corticosteroids; instead all patients received oral glycerol. Fluids were not restricted. All patients aged 2 months to 12 years with probable bacterial meningitis7 and their guardian’s informed consent were enrolled during July 2005 to June 2008. The present study included the patients with confirmed bacterial meningitis7 (n = 553). When a child died, the time in hours between the initiation of treatment and moment of death was registered in the patient’s chart. On admission, severe respiratory distress (= visible chest indrawing) was found in 67/553 (12%) patients, severe malnutrition (= middle upper arm circumference below 2 standard deviations) in 138/550 (25%) and some underlying condition (mostly frequently sickle cell disease and chronic suppurative otitis media) in 60/553 (11%) of the patients.
Statistical Analysis Accepted for publication January 9, 2014. From the *Faculty of Medicine, University Diego Portales, Santiago, Chile; †Institute of Clinical Medicine, University of Helsinki, Helsinki, Finland; ‡David Bernardino Children’s Hospital, Luanda, Angola; §Children’s Hospital, Helsinki University Central Hospital, and Helsinki University; ¶National Institute for Health and Welfare; and ‖Department of Otorhinolaryngology, Helsinki University Central Hospital, and Helsinki University, Helsinki, Finland. This work was supported by the Sigrid Juselius and the Paediatric Research Foundations, Helsinki, Finland. H.P. is a scientific consultant of the Serum Institute of India. The authors have no other funding or conflicts of interest to disclose. Address for correspondence: Irmeli Roine, MD, PhD, Los Misioneros 2237, 7520179 Santiago, Chile. E-mail: [email protected]. Copyright © 2014 by Lippincott Williams & Wilkins ISSN: 0891-3668/14/3308-0789 DOI: 10.1097/INF.0000000000000350
Time to death was assessed in relation to several other clinical and laboratory variables using Spearman correlation, the contingency table and the Kruskall-Wallis test, whichever was appropriate. It was tested both as a continuous variable, and in periods of 0–4, 5–8, 9–24, 25–120 (approximately days 2–5 of treatment) and >120 hours (after day 5). Logistic regression analysis was used to examine whether the known risks of death (independent variables) were specifically related to a certain period to death (dependent variable, each period compared with the survivors). The strength of associations is expressed as odds ratios (OR) with 95% confidence intervals (CI). The quantitative data are presented as medians with interquartile range (IQR). P 0.05) was observed with age, plasma hemoglobin, blood white cell count or several other patient and laboratory variables listed in Table, Supplemental Digital Content 1, http://links.lww.com/INF/B874. The treatment alternatives in the original trial7 also influenced time to death. Administration of the antibiotic as an infusion together with high-dose oral paracetamol tended (P = 0.06) to prolong time to death (median, 58 hours; IQR, 111; n = 47) compared with administration of the antibiotic as a bolus with an
TABLE 1. Differences in Admission Findings According to Time to Death in 206 Non-survivors of Childhood Bacterial Meningitis
Variable Age, months Underlying illness‡ Preadmission illness, days¶ Seizures during admission Glasgow Coma Score║ Systolic blood pressure, mm Hg Severe respiratory distress Severe dehydration MUAC**, cm CSF-glucose mg/dL§§ Etiology‡‡ S. pneumoniae H. influenzae N. meningitidis Other‡‡
0–4 h*
5–8 h
9–24 h
25–120 h
> 120 h
N
n = 28
n = 30
n = 56
n = 65
n = 27
P Value
206 206 206 206 205 195 206 206 203 202 170 82 62 7 19
15 [59]† 1 (4)§ 3 [4] 21 (75) 9 [4] 90 [30] 9 (32) 2 (7) 13 [3] 9 [13]
11 [9] 2 (7) 3 [2] 21 (70) 8 [3] 90 [25] 12 (40) 1 (3) 13.5 [3] 9 [9]
9·5 [14] 3 (5) 6·5 [6] 39 (70) 8 [5] 90 [23] 8 (14) 2 (4) 13 [2] 9 [10]
12 [19] 12 (18) 6 [7] 41 (63) 9 [5] 90 [18] 11 (17) 4 (6) 13 [3] 10 [20]
32 [72] 7 (26) 7 [4] 11 (41) 9 [6] 100 [10] 3 (11) 0 (−) 14 [2] 17 [16]
19 (68) 2 (7) 3 (11) 1 (4)
16 (53) 8 (27) 1 (3) 1 (3)
22 (39) 21 (38) 2 (4) 5 (9)
18 (28) 24 (40) 0 (0) 10 (15)
7 (26) 7 (26) 1 (4) 2 (7)
0.08 0.01 0.005 0.06 0.45 0.39 0.02 0.66 0.27 0.09 0.009 0.007 0.02 0.18 0.19
Contingency table and Kruskall-Wallis test. *Hours after initiation of cefotaxime treatment. †Median value with IQR in brackets. ‡Sickle cell disease (n = 12), prior episode of bacterial meningitis (n = 3), chronic otitis media (n = 3), prior head trauma or surgery (n = 3), prolonged cough and mother with tuberculosis (n = 1), fever for 1 year (n = 1), sick for 1 month (n = 1), cleft lip and palate (n = 1). §In parentheses, percentages. ¶Symptoms and signs suggestive of bacterial meningitis before admission. ║Range 3–15 (best). **Middle upper arm circumference. ‡‡Of the 51 bacteria tested for cefotaxime sensitivity, 49 (96%) were sensitive and 2 (4%) resistant: 1 H. influenzae patient who died after 123 hours and 1 Klebsiella pneumoniae patient who died after 92 hours. ‡‡K. pneumoniae, n = 4; Streptococcus Gr. C, n = 3; Proteus mirabilis, n = 3; Esherichia.coli, n = 2; Salmonella sp., n = 1; Pseudomonas aeruginosa, n = 1; P. vulgaris, n = 1; Enterobacter aglomerans, n = 1; Citrobacter freundii, n = 1. §§Cerebrospinal fluid from diagnostic tap on admission.
FIGURE 1. Time to death according to etiology of 206 children with bacterial meningitis.
790 | www.pidj.com
© 2014 Lippincott Williams & Wilkins
© 2014 Lippincott Williams & Wilkins
Logistic regression model. *Compared with 347 survivors of childhood bacterial meningitis. † Range 3–15 (best) ‡A blood pressure 120 hours and in 0/347 (0%) of survivors (P = 0.03). §Middle upper arm circumference below –2SD indicates severe malnutrition. ¶CSF from diagnostic tap on admission. ║Confirmed in 452 patients. In addition to etiologies mentioned in the Table, there were 49 patients with meningococcal meningitis. ** K. pneumoniae 5, Streptococcus Gr. B 1, C 4, D 1, G 1, Proteus mirabilis 5, Salmonella sp. 4, E. coli 3, Stapylococcus aureus 2, P. aeruginosa 1, E. aglomerans 1, C. freundii 1 and not identified but with growth 3.
0.68 0.80 0.21 0.46–3.35 0.33–2.38 0.57–13.4 0.50–1.74 0.60–1.95 2.03–11.9 2.82 0.56 0.83 2.16–6.45 0.03–0.47 0.10–6.88 5.59 0.11 0.86
0.0004 0.003 0.89
1.24–6.45 0.24–1.33 0.10–6.59
0.01 0.19 0.86
1.39 0.91 2.30
0.75–2.55 0.50–1.67 07.8–6.76
0.29 0.76 0.13
0.93 1.08 4.93
0.83 0.81 0.0004
1.24 0.88 2.76
0.005 0.94 0.42 0.92 0.39 0.93 0.49 1.44–7.64 0.44–2.15 0.47–5.60 0.35–2.62 0.55–4.71 0.24–4.79 0.06–3.77 1.48–4.89 1.39–4.16 1.27–5.92 0.73–2.55 1.22–4.82 0.10–1.84 0.51–3.30 3.26 3.39 8.97 2.82 0.55 0.96 1.96 1.44–7.64 1.75–10.2 2.61–15.6 1.29–6.20 1.97–13.9 1.72–11.5 1.72–11.6
Glasgow Coma Score < 8† Seizures during admission Severe respiratory distress Blood pressure < 90 mm Hg‡ MUAC < -2 standard deviations§ CSF¶ white cell count < 100/μL Blood glucose < 40 g/dL Etiology‖ S. pneumoniae H. influenzae Other**
550 553 553 525 550 551 548 452 184 187 32
3.31 4.23 6.38 3.08 5.23 4.46 4.47
0.005 0.003
