Antibiotics are one of the most frequently prescribed medications in children with acute infectious diseases.1–8 In Japan, the most commonly prescribed antibiotics were macrolides and third-generation cephalosporins.1–8 Inappropriate use and overprescriptions of broad-spectrum third-generation cephalosporins are not warranted because of their low bioavailability and concerns about the development of resistant strains, rising healthcare expenditure and potential side effects. Additionally, there are several cases that reported a potential association between pivalate-conjugated antibiotics (eg, cefcapene pivoxil, cefditoren pivoxil, cefteram pivoxil and tebipenem pivoxil) and hypoglycemia.9–16 The underlying mechanisms are considered that pivalate-conjugated antimicrobials may induce the secondary carnitine deficiency under the conditions of decreased oral intake, fever and/or infections, resulting in hypoglycemia and skeletal muscle symptoms. However, these case series and case reports were limited to assess the potential associations, and there is no observational study that investigated the association between pivalate-conjugated antimicrobials and hypoglycemia.
To address these gaps in knowledge, we conducted a retrospective observational study and systematic review of the literature. We hypothesized that the use of pivalate-conjugated antibiotics was associated with decreased blood glucose levels in children who have respiratory infections and that it mediated and/or moderated the effects of respiratory symptom duration on blood glucose levels.
MATERIALS AND METHODS
Study Population and Participation
A retrospective observational study was conducted on children with lower respiratory infections who were admitted to the Department of Pediatrics at Maebashi Red Cross Hospital between June 2015 and March 2017. Maebashi Red Cross Hospital is a tertiary hospital that provides health care by specialists for patients with mild to severe disorders. Lower respiratory infections were determined by the information of patient history, symptoms and physical examination. We did not include patients who had symptoms of acute infectious diarrhea. The patient data were identified using electronic medical records. We included a total of 814 consecutive inpatients less than 15 years of age with the diagnosis of respiratory infections. The study was approved by the institutional review board at Maebashi Red Cross Hospital.
Measurements of Variables
Patient pre-exposure/treatment characteristics included age, sex (male or female), duration of respiratory symptoms and fever in days, use of antibiotics and corticosteroids, blood glucose levels at admission and chronic medical conditions (congenital heart, neurologic, hematologic, pulmonary, endocrine, gastrointestinal, and genetic disorders), as either continuous or categorical variables appropriately. Types of antibiotics were categorized into 2 groups based on the presence of pivalic acid conjugations. All children received only one type of antibiotics, and no children received more than 1 type of antibiotics. Antibiotic use data was based on prehospital information only. All blood glucose levels were measured at the time of admissions.
Descriptive statistics, including means and frequencies for continuous and categorical data, were analyzed for all patients stratified by use and types of antibiotics. To estimate the potential effects of antibiotic use on blood glucose levels and ascertain whether antibiotic use had mediating and moderating effects for the pathway from the duration of respiratory symptoms to blood glucose levels, we performed 2 different statistical approaches: traditional regression (Figure, Supplemental Digital Content 1, http://links.lww.com/INF/D608) and causal mediation analyses (Figure, Supplemental Digital Content 2, http://links.lww.com/INF/D609 and Figure, Supplemental Digital Content 3, http://links.lww.com/INF/D610).
First, we constructed linear regression models that contained blood glucose levels as the dependent variables and baseline patient characteristics and duration of antibiotic use in days as independent variables (Figure, Supplemental Digital Content 1, http://links.lww.com/INF/D608). Model 1 examined the crude association between duration of antibiotic use and blood glucose levels. Models 2–4 investigated the association after adjusting for patient age and sex (model 2), steroid and chronic disease (model 3) and duration of fever and respiratory symptoms (model 4). Model 5 ascertained the association after adjusting for all covariates. We assigned duration of respiratory symptoms and fever as variables because they could be associated with decreased levels of blood glucose levels through decreased appetite and elevated energy demands.
Second, we performed causal mediation analyses to investigate the effects of duration of respiratory symptom and antibiotic use on blood glucose levels and determine whether durations of antibiotic use had mediating and/or moderating effects for the pathway from duration of respiratory symptoms to blood glucose levels, after adjusting for potential confounders with multiple linear regression models. Causal mediation analyses were developed by VanderWeele17,18 and allow for exposure-mediator interactions and their causal interpretations using 4-way decompositions. These decompositions provide the maximum insight for clarifying the contribution of interactive and mediating mechanisms in a given observed effect.17,18 The controlled direct effect (CDE) reflects the effect of exposure (duration of respiratory symptoms) on the outcome (blood glucose levels) that does not require the mediator (duration of antibiotic use). The pure indirect effects represent the effect of a mediator on the outcome in the absence of the exposure. The reference interaction effect and mediated interaction effect showed the additive interaction effects of a mediator on the outcome in the absence and presence of the effects of exposure on a mediator (Figure, Supplemental Digital Content 2, http://links.lww.com/INF/D609). For these analyses, we considered CDE as the direct effects of symptom duration on blood glucose levels and portion eliminated (PE, which is reference interaction effect + pure indirect effects + mediated interaction effect) as the moderating and mediating effects of antibiotics on blood glucose levels (Figure, Supplemental Digital Content 3, http://links.lww.com/INF/D610). SEs and confidence limits were estimated by bootstrapping. All results were reported with 95% confidence intervals, and statistical significance was set at 2-sided P < 0.05. All data were analyzed using STATA software version 14.2 (StataCorp LP, TX).
Finally, we conducted a systematic review of the literature for the association between pivalate-conjugated antibiotics and hypoglycemia. We searched 2 medical literature electronic database (PubMed and Embase) from their inception to January 2019. The search terms were “(pivalate or pivalic or pivaloyl) and (hypoglycemia or hypocarnitinemia or carnitine).” We determined to include all clinical and observational studies, case series and case reports that reported the association between pivalate-conjugated antibiotics and hypoglycemia. The authors screened all articles independently. When the disagreements occurred, we discussed their inclusion or exclusion until the agreements and summarized the results.
We identified a total of consecutive 814 patients who were hospitalized with lower respiratory infections over the study period. The patient characteristics are summarized in Table 1. Compared with no antibiotic group, children who received antibiotics were more likely to be older, less likely to be male, more likely to use corticosteroids and had a longer duration of respiratory symptoms and fever.
Table 2 summarizes the multivariable linear regression for the effects of types and duration of antibiotic use as well as duration of respiratory symptoms after adjusting for baseline characteristics. Model 1 indicated that 1-day increase in antibiotic use was associated with 1.1–1.2 mg/dL reduction in blood glucose levels. These trends did not change when we adjusted for age and sex (model 2), chronic disease and corticosteroid use (model 3). However, the effects of antibiotic use on blood glucose levels were mitigated after adjusting for duration of respiratory symptoms and fever (model 4). The final model (model 5) showed that age, corticosteroid use and duration of respiratory symptoms and fever had independent relationships to blood glucose levels.
To investigate the direct effects of respiratory symptom duration and mediating/moderating effects of antibiotic use on blood glucose levels, we conducted causal mediation analyses with 4-way decompositions (Table 3). Among patients who received pivalate-conjugated antibiotics, CDE was associated with reductions in blood glucose levels (−4.5 mg/dL for the 1-day increase in antibiotic use (95% confidence intervals, −1.6 to −7.3 mg/dL), whereas PE was not (−0.3 to 2.7 mg/dL). These findings indicate that the only direct effects of duration of respiratory symptoms on blood glucose levels in the absence of the mediator (duration of antibiotic use) had contributed to the reduction in blood glucose levels. Similar patterns were observed among those who received antibiotics that did not contain pivalic acid.
Table 4 summarized the results of the systematic review of the literature. We found a total of 7 articles (2 case series and 5 case reports) for the association between pivalate-conjugated antibiotics and hypoglycemia. The cases were 2 months to 42 years old (most cases are children < 4 years of age) and were received antibiotics for 3 days to 12 months. As most cases were reported in Japan, we also searched the relevant articles written in Japanese using Ichushi-Web. A total of 14 cases of patients were reported with the age range of 11 months to 85 years and duration of antibiotic use of 1 day to 4 months. These results are summarized in Table, Supplemental Digital Content 1, http://links.lww.com/INF/D607.
The present study investigated the effects of duration of respiratory symptoms and antibiotic use on the blood glucose levels. In the traditional regression approach, we observed that the duration of fever and respiratory symptoms had an independent relationship to blood glucose levels, but the duration of antibiotic use did not. Causal mediation analyses also indicated that CDE of the duration of respiratory symptoms contributed to the reductions in blood glucose levels, whereas PE did not. Despite our prior hypotheses, we failed to demonstrate the effects of antibiotic use on reduction in blood glucose levels as the previous reports showed for patients who received pivalate-conjugated antibiotics. To the best of our knowledge, this is the first study that investigated the potential effects of antibiotic use on blood glucose levels.
Pivalate is conjugated with antimicrobials, such as third-generation cephalosporins, to improve their bioavailability.10–16 After absorption from the intestinal tract, pivalate acid undergoes conjugation with carnitine in the body, and almost all of the carnitine-conjugated pivalate acid is excreted into the urine.10–16 As a result, the use of pivalate-conjugated antimicrobials increases carnitine consumption in the body, resulting in secondary carnitine deficiency and subsequent clinical manifestations, such as hypoglycemia, hyperammonemia, hepatic disorder and skeletal muscle symptoms.
Our literature review found that cases of hypoglycemia potentially due to pivalate-conjugated antibiotics were more likely to be younger children with prolonged use of antibiotics.10–16 These findings could be explained by the relatively small amounts of glycogen and carnitine storage in younger children. However, hypoglycemia in some cases occurred even after a few days of antibiotic use, suggesting that genetic variations of metabolism for the carnitine circuit may have affected the development of hypoglycemia with secondary carnitine deficiency. Additionally, case reports were clustered in Japan,10–13,15,16 which supports the possibility of genetic predispositions of carnitine metabolisms in Japanese children.
However, our study failed to prove the potential effects of pivalate-conjugated antimicrobials on blood glucose levels. Our findings indicate that the blood glucose levels were directly affected by the duration of fever and respiratory symptoms (CDE), rather than by the mediating/moderating effects of pivalate-conjugated antimicrobial use (PE). However, these results should be interpreted with cautions. First, we were not able to measure serum carnitine levels due to the retrospective nature of our study design. Moreover, the serum carnitine levels may have been lower in those who received pivalate-conjugated antimicrobials than in those who did not. Second, our patients experienced relatively shorter durations of antimicrobial use than those in previously reported papers. Therefore, the duration of antimicrobials may have been insufficient to induce secondary carnitine deficiency and subsequent hypoglycemia.
There are additional limitations to our study. As our study was conducted in a single hospital, generalizability would be uncertain. Unmeasured factors such as serum carnitine levels and their genetic variations for carnitine metabolisms could not be obtained, and further studies are needed to account for these factors. Although we assumed the durations of antimicrobials based on the combinations of physician interview with parents and reviewing pharmacy records, there may have been measurement errors in durations of antimicrobial use due to patient noncompliance. Furthermore, effect measure modification could be present for the association between antibiotic use and hypoglycemia. This means that the pathogenesis in the relatively small fraction of patients who develop hypoglycemia could be different (such as metabolisms of carnitine pathways) from others.
In conclusion, we failed to prove the potential effects of pivalate-conjugated antimicrobial use on blood glucose levels, and our findings indicate that durations of respiratory symptoms and fever were a strong risk factor for reducing blood glucose levels, but the effects of pivalate-conjugated antimicrobials were unclear. Further studies are needed to illuminate these findings.
1. Okubo Y, Miyairi I, Michihata N, et al. Recent prescription patterns for children with acute infectious diarrhea. J Pediatr Gastroenterol Nutr. 2019;68:13–16.
2. Okubo Y, Michihata N, Morisaki N, et al. Recent patterns in antibiotic use for children with group A streptococcal infections in Japan. J Glob Antimicrob Resist. 2018;13:55–59.
3. Okubo Y, Michihata N, Morisaki N, et al. Recent trends in practice patterns and impact of corticosteroid use on pediatric Mycoplasma pneumoniae-related respiratory infections. Respir Investig. 2018;56:158–165.
4. Uda K, Okubo Y, Shoji K, et al. Trends of neuraminidase inhibitors use in children with influenza related respiratory infections. Pediatr Pulmonol. 2018;53:802–808.
5. Muraki Y, Yagi T, Tsuji Y, et al. Japanese antimicrobial consumption surveillance: first report on oral and parenteral antimicrobial consumption in Japan (2009-2013). J Glob Antimicrob Resist. 2016;7:19–23.
6. Kinoshita N, Morisaki N, Uda K, et al. Nationwide study of outpatient oral antimicrobial utilization patterns for children in Japan (2013-2016). J Infect Chemother. 2019;25:22–27.
7. Uda K, Okubo Y, Kinoshita N, et al. Nationwide survey of indications for oral antimicrobial prescription for pediatric patients from 2013 to 2016 in Japan. J Infect Chemother. 2019;25:758–763.
8. Okubo Y, Hayakawa I, Nariai H, et al. Recent practice patterns in diagnostic procedures anticonvulsants, and antibiotics for children hospitalized with febrile seizure. Seizure. 2019;67:52–56.
9. Pharmaceuticals and Medical Devices Agency. PMDA alert for proper use of drugs. Serious hypocarnitinemia and hypoglycemia
in children treated with antibacterials with a pivoxil group. Available at https://www.pmda.go.jp/files/000153551.pdf
. Accessed June 30, 2019.
10. Kobayashi H, Fukuda S, Yamada K, et al. Clinical features of carnitine deficiency secondary to pivalate-conjugated antibiotic therapy. J Pediatr. 2016;173:183–187.
11. Nakajima Y, Ito T, Maeda Y, et al. Detection of pivaloylcarnitine in pediatric patients with hypocarnitinemia after long-term administration of pivalate-containing antibiotics. Tohoku J Exp Med. 2010;221:309–313.
12. Nakazaki K, Ogawa E, Ishige M, et al. Hypocarnitinemia observed in an infant treated with short-term administration of antibiotic containing pivalic acid. Tohoku J Exp Med. 2018;244:279–282.
13. Ito M, Fukuda M, Suzuki Y, et al. Carnitine-related hypoglycemia
caused by 3 days of pivalate antibiotic therapy in a patient with severe muscular dystrophy: a case report. BMC Pediatr. 2017;17:73.
14. Jun JS, Lee EJ, Park HD, et al. Systemic primary carnitine deficiency with hypoglycemic encephalopathy. Ann Pediatr Endocrinol Metab. 2016;21:226–229.
15. Okumura A, Morita M, Ikeno M, et al. Acute encephalopathy in a child with secondary carnitine deficiency due to pivalate-conjugated antibiotics
. Pediatr Infect Dis J. 2011;30:92.
16. Makino Y, Sugiura T, Ito T, et al. Carnitine-associated encephalopathy caused by long-term treatment with an antibiotic containing pivalic acid. Pediatrics. 2007;120:e739–e741.
17. VanderWeele TJ. A unification of mediation and interaction: a 4-way decomposition. Epidemiology. 2014;25:749–761.
18. Discacciati A, Bellavia A, Lee JJ, et al. Med4way: a stata command to investigate mediating and interactive mechanisms using the four-way effect decomposition. Int J Epidemiol. 2018. doi: 10.1093/ije/dyy236. [Epub ahead of print]