INTRODUCTION

In recent decades, due to indiscriminate antimicrobial use and the selective pressure exerted by them, infections related to multi-resistant microorganisms have emerged. These microorganisms can range from bacteria, fungi, viruses and other parasites, which have the ability to render antimicrobials ineffective¹.

Antimicrobial resistance (AMR), which occurs when a pathogen shows resistance to one or more AMR from three or more tested categories,^1,2,3 is a situation of worldwide concern and impact, so much so that the World Health Organization estimates that, by 2050, resistance to AMR will lead to the death of around one person every three seconds (more than ten million people per year). This prediction was made in 2019, before the emergence of the coronavirus pandemic, COVID-19.^4-7 COVID-19, one of the greatest pandemics in history, originated in China at the end of 2019,^7-8 and its causal agent is SARS-CoV-2, a strain of coronavirus that causes an upper respiratory tract infection, which can progress to pneumonia as a secondary infection and severe acute respiratory syndrome. With so many patients suffering from severe respiratory conditions, hospital admissions have increased, mainly in an intensive care environment, with a significant increase and indiscriminate antibiotic use throughout the country, resulting in antimicrobial resistance as a public health consequence.^5-6

With increasing rates of antimicrobial resistance due to indiscriminate AMR use (such as azithromycin) to combat healthcare-associated infections, one of the antibiotics used against organisms resistant to carbapenems was polymyxin B. Due to the high demand for this drug, a national shortage of the drug occurred during 2021⁸.

Based on previous studies to restrict broad-spectrum antibiotic use, replacing polymyxin B with aminoglycosides, the strategy adopted by many institutions to overcome the national shortage crisis was to use aminoglycosides, mainly amikacin and gentamicin, as an alternative. However, these studies were carried out in ideal situations, where drug replacement was a choice and not a necessity, as observed during the pandemic.^9-14

Taking into account this historical moment and the public health crisis that has taken place, the hypothesis of this work is that the replacement of polymyxin B with aminoglycosides did not lead to greater mortality among critically ill admitted to hospital patients, provided that the multidrug-resistant bacteria were sensitive to amikacin and/or gentamicin. Regarding the outcome of renal failure, using both drugs is expected to harm the organ, as both classes are known to be nephrotoxic.¹⁵ However, there is still no comparison in the literature between polymyxin B and aminoglycoside and the greater possibility of using one or the other leading to renal replacement therapy.

Therefore, the study’s main objective was to analyze the impact that replacing polymyxin B with amikacin and gentamicin had on patient outcomes. To this end, the mortality of patients who received aminoglycoside versus polymyxin B and the recovery duration from the infectious condition for each of the antibiotics used were assessed. Data on progression to renal replacement therapy and length of stay in a critical environment were also collected.

METHOD

This is an analytical study with an observational, cross-sectional design, with a quantitative approach, through a retrospective analysis, which was carried out based on data obtained from medical records of critically ill patients admitted to a small tertiary hospital in the Metropolitan Region of Curitiba, in the drug shortage period, from August and September 2020 and August and September 2021.^16-17

Critical patients who required polymyxin B use during 2020 and those who used amikacin or gentamicin during 2021 were included. Patients who used AMR concomitantly and who were only over 18 years of age were also included. Incomplete medical records and those of patients who died within 24 hours of using antibiotics or being admitted to hospital were excluded.

Data were collected and tabulated in spreadsheets using Excell^®. For descriptive analysis, categorical variables were presented according to their frequencies, prevalence and percentages, while quantitative variables were described according to means, deviation and standard error. Participants were divided into groups according to drug use. In most analyses, the sample was composed of two groups: 1) aminoglycosides (AM); and 2) polymyxin (PL). In other analyses, a subdivision of the drug was carried out, with four groups: 1) sensitive aminoglycoside (SAM); 2) aminoglycoside resistance (AMR); 3) PL; and 4) PL + AM (PLAM).^18-19

For inferential analyses, data tabulated in Excell^® were transferred to the Statistical Package for the Social Sciences (SPSS, IBM Statistics, v.23, 2015). The Mann-Whitney U test was used to compare, between AM and PL groups, age, total length of stay, length of stay in a critical environment, duration of use and number of antibiotic cycles. With the aim of investigating the primary outcome, which is equivalent to the patient being discharged from hospital or death, and the secondary outcome, which is equivalent to progressing to dialysis, the chi-square test was used to compare the frequency of occurrence of these variables between groups.^18-19

The Spearman correlation test was used to correlate Polymerase Chain Reaction (PCR), creatinine and urea values on D3, total length of stay and length of stay in a critical environment, duration of use and number of antibiotic cycles with hospital discharge/death outcomes. The correlation between palliative care and age was also analyzed.^18-19

Generalized Estimating Equation (GEE) models with appropriate link function (identity or log) and (linear, gamma or tweedie) distribution assumed were used to compare the quantitative variables (PCR, creatinine and urea) between groups (AM and PL) and length (D0, D3, D7, D10 and D14). Thus, group and duration were the effects analyzed as well as their interaction. The distribution model was selected (best overall fit) based on the lowest value of Quasi-likelihood under Independence Model Criterion (QIC, 2007). AR1 structure matrix was used, and, when necessary, Bonferroni post-hoc for subsequent comparisons. The significance of all analyzes was set at 5% (p < 0.05).^18-19

As it involves reviewing medical records, the project was previously submitted to the Research Ethics Committee of São José dos Pinhais, under Consubstantiated Opinion 5,549,515 (CAAE (Certificadode Apresentação para Apreciação Ética - Certificate of Presentation for Ethical Consideration): 58850522.9.0000.9587), in compliance with the Standards for Conducting Research on Human Beings of the Brazilian National Health Council, in accordance with Resolution 466/12.

RESULTS

Thus, 22 patients comprised the sample of this study. Depending on drug use, the AM group had 10 patients (69.9 ± 7.82 years, 8 men, 2 women), and the PL group had 12 (70.4 ± 8.26 years, 8 men, 4 women). There was no significant difference in age between groups (U = 56.000; p = 0.791).

The main point researched, which was to compare the outcome of discharge or death between groups, is presented in Tables 1 and 2. For this result, two analyzes are presented, dividing patients into 2 (Table 1 – PL and AM) or 4 groups (Table 2 - SAM, AMR, PL and PLAM), in order to try to remove the influence of the low sample number. The chi-square test found no difference in the frequency of death and discharge in either of the two analyzes.

However, it is noteworthy that, when observing the groups of patients who received only AM, mortality was higher in those who were sensitive to the drug (75%) than in those who were not resistant to it (33%).

Table 3 presents different outcomes. It was observed that there was no difference between length of stay, number of antibiotic cycles, number of care-related infections or number of antibiotics between AM and PL groups.

According to Table 3, it can be seen that there was a statistical difference in relation to length of stay in a critical environment, which was longer for patients in the PL group as well as a longer period of antibiotic use. In order to justify these results, the duration between the start of treatment and the outcome of death was also compared between groups in the hypothesis that there was a positive correlation in relation to the duration of antibiotic use with greater survival. However, the Mann-Whitney test did not observe a significant difference in this variable (U = 21.00; p = 0.147), possibly due to the variability of values and the reduced number of patients, since the mean number of days is descriptively longer in the PL group (AM = 8.5 ± 5.75 vs. PL = 15.2 ± 11.2 days).

Mortality was high in both groups. A complementary analysis to assess whether older age would be a predisposing factor for a worse outcome was performed. The Mann-Whitney test did not identify a difference in age (U = 28.00; p = 0.254) nor in the number of infections (U = 35.50; p = 0.567) among patients who were discharged or died.

When comparing the frequency of need for dialysis (Table 4), the chi-square test identified a significant difference between groups, such that patients in the PL group needed dialysis more than those in the AM group. When analyzing patients divided into 4 groups (Table 5 - SAM, AMR, PL and PLAM), SAM, AMR and PL groups were very similar in terms of the need for dialysis, however, when there was a combination of two drugs (PLAM), the frequency of dialysis was much higher, showing that using two nephrotoxic medications had a more significant impact on the dialysis outcome.

Another analysis was carried out with patients divided between the discharge and death groups, regardless of the medication used, to assess how much comorbidities influenced the outcomes. The comorbidities assessed were hypertension (HT), type 2 diabetes mellitus, coronary artery disease, heart failure and chronic obstructive pulmonary disease (COPD). The chi-square test found no difference in the frequency of occurrence of any comorbidity both when comparing drug groups and outcome groups (X² < 2.426; p > 0.119). As a complementary analysis, patients were classified into those who had 3 comorbidities or more and 2 comorbidities or less. The chi-square test was used to investigate whether those with more than 3 comorbidities had a higher frequency of death in the sample. The results showed that 100% of patients with more than 3 comorbidities (n = 7) died, compared to 67% of patients with 2 comorbidities or less (n = 10). Despite the descriptive difference, the test did not find a significant difference (X² = 3.020; p = 0.082).

The drug and outcome groups were also compared in relation to the reason for hospital admission. It was analyzed whether patients who were admitted to hospital due to COVID-19 died more often than due to other causes (all others grouped into a single category), but there was no statistical difference. Descriptively, 30% (n = 3) of patients in the AM group were admitted to hospital due to COVID-19, and in the PL group, 58% (n = 7).

In the AM group, 70% (n = 7) used the antibiotic for ventilator-associated pneumonia (VAP), and 30% (n = 3) for urinary tract infection (UTI). In the PL group, 84% (n = 10) used it for VAP, 1% (n = 1) for UTI, and only 1% for other causes (1 patient with osteomyelitis). No differences were found in comparisons between groups in any of these variables either.

Finally, one of the secondary objectives of this study was to assess the recovery duration from the infectious condition for each of the antibiotics used through PCR analysis, in addition to the progression of renal dysfunction with urea and creatinine data on D0, D3, D7, D10 and D14. However, this analysis was hampered due to the low sample number, and it was not possible to obtain reliable data due to the high death rate within the first 5 days of hospital admission.

DISCUSSION

The work’s main objective was to assess the primary outcome – death or discharge – of patients. The tests showed no difference in the frequency of death and discharge in either of the two analyses. In other words, polymyxin B use did not protect patients from the final negative outcome – death, even when only AM was used in organisms resistant to these. This result is consistent with the two studies that used AM to replace polymyxin B without an increase in mortality.^9-13Once again, it is highlighted that the replacement of polymyxin B with gentamicin or amikacin was not a strategic option for antimicrobial control, but rather a necessity due to the current restriction and shortage of polymyxin B. However, in terms of public health, considering costs and antimicrobial resistance, this result is consistent with the possibility of strategies to restrict broad-spectrum antibiotic use.

However, the results showed that, contradictorily, the group of patients who had an infection with AM-resistant bacteria (and were treated with this drug due to the scarcity of polymyxin B) had lower mortality (33%) than the group who had an infection with AM-sensitive bacteria.

This contradiction allows us to raise two hypotheses for the results obtained. The first would be that replacing polymyxin B with AM did not impact death. The second would be that patients treated during the pandemic had a high risk of death, and this would occur regardless of the class of antimicrobial used.

Although polymyxin B is a broad-spectrum antibiotic for gram-negatives, especially for Acinetobacter spp, Pseudomonas aeruginosa and Klebsiella pneumoniae carbapenemase (KPC),^15,20 its use was not superior to AM use, as it did not lead to fewer infections and did not reduce the need to use other antibiotic cycles or total hospital stay. In other words, once again, in terms of public health, antimicrobial resistance and global development, using AM instead of polymyxin B would be justified.

The results shown in Table 3 showed that patients who used polymyxin B remained admitted to hospital in a critical environment longer than the group that used AM. The attempt to justify these results was a complementary analysis to assess whether these patients remained alive longer because they had used the antibiotic. However, no statistical difference was observed between these variables, i.e., the hypothesis raised did not justify the difference in days of hospital admission in a critical environment. However, it is worth noting that, possibly, different and more enlightening results would be obtained with a larger sample, since the average number of days is descriptively longer in the PL group (AM = 8.5 ± 5.75 vs. PL = 15.2 ± 11.2 days). It is worth noting that, even though they remained in a critical environment for longer and used antibiotics for longer, this did not influence the primary outcome (death or discharge).

The secondary outcome assessed was in relation to progression to dialysis. The first result obtained showed that patients in the PL group needed dialysis more than those in the AM group. Although both classes of drugs are known to be nephrotoxic,^15,20 it was expected that the group that used AM would have a worse renal outcome. The complementary analysis carried out to clarify the result obtained demonstrated that, in reality, when two drugs were used in the same patient (i.e., the sum of two nephrotoxic drugs), there was a greater progression to dialysis.

Regarding comorbidities, there was no statistical difference between groups, i.e., having HT, type 2 diabetes mellitus, coronary artery disease, heart failure or COPD alone. did not cause more patients to die, even when assessing whether having 3 or more comorbidities (e.g., the impact of the sum of diseases) also did not demonstrate a worse outcome. However, the results showed that 100% of patients with more than 3 comorbidities died, demonstrating the need for a larger sample group (with a possible result different from that obtained). It is assumed that this result would be different due to the fact that the impact of comorbidities on the death outcome has already been analyzed in previous studies, in which it was seen that patients with comorbidities (mainly cardiovascular) had worse outcomes compared to patients without comorbidities.^21,22,23

Regarding age or number of infections, there was no difference between groups. The reason that led to antibiotic use and hospital admission also did not impact the outcome.

Although, descriptively, more patients admitted to hospital due to COVID-19 died than due to other causes, there was no statistical difference in the tests performed. In other words, hospital admission for COVID-19 was not what impacted the death of patients.

The evolutionary comparison of PCR from treatment was hampered, as most patients died within the first 5 days after starting antibiotic therapy.

The study has restrictions, mainly due to the small sample size, which did not allow for a more meaningful analysis of the impact of some variables (such as age and comorbidities). The high number of patients who died within the first 5 days after identifying the infectious process was also a limiting factor (impairing the evolutionary comparison of laboratory tests, such as PCR, urea and creatinine). Despite the limitations, the study contributes to clinical practice by reinforcing the importance of carrying out more studies comparing broad-spectrum AMR use, helping the medical community when making decisions related to the prescription and rational use of these drugs. The study also helps professionals assess the results of drug substitution measures in scenarios of national shortages of these drugs.

CONCLUSION

The temporary shortage of polymyxin B led to its therapeutic replacement, with AM use. The results of this action showed that there are two viable hypotheses: that this replacement did not change the primary outcome (death); or regardless of the drug used, the outcome would be the same. No differences were observed between comorbidities and age between groups. Although both drugs are nephrotoxic, it was seen that the greatest need for dialysis occurred when there was concomitant use of the two classes of AMR. The study demonstrates the importance, in terms of public health, of using broad-spectrum AMR in a rational manner, aiming at the possibility of replacement by smaller-spectrum AMR. However, the work has limitations due to the low number of patients, and these data require further studies.

REFERENCES

1. Schrader SM, Vaubourgeix J, Nathan C. Biology of antimicrobial resistance and approaches to combat it. Science Translational Medicine 2020; 12(549). doi:10.1126/scitranslmed.aaz6992.

2. Mancuso G, Midiri A, Gerace E, et al. Bacterial antibiotic resistance: The most critical pathogens. Pathogens 2021; 10(10):1310-2021. https://doi:org/10.3390/pathogens10101310.

3. Chokshi A, Sifri Z, Cennino D, et al. Global contributors to antibiotic resistance. Journal of Global Infectious Diseases 2019, 11(1):1-36. doi:10.4103/jgid.jgid_110_18.

4. World Healf Organization. New report calls for urgent action to avert antimicrobial resistance crisis, 2022. https://www.who.int/news/item/29-04-2019-new-report-calls-for-urgent-action-to-avert-antimicrobial-resistance-crisis

5. Minante BI, Ticotosti GG, Barros IS, et al. Resistência antimicribiana pós- COVID-19: Revisão de Literatura. Revista Científica Multidisciplinar 2022; 3(3): 1-14. doi.org/10.47820/recima21.v3i3.1266.

6. V’kovski P, Kratzel A, Steiner S, et al. Coronavirus biology and replication: implications for SARS-CoV-2. Nature Reviews Microbiology 2021; 19:155–170. https://doi.org/10.1038/s41579-020-00468-6.

7. Huttner B, Catho G, Pano-Pardo JR, et al. COVID-19: don’t neglect antimicrobial stewardship principles! Clinical Microbiology and Infection 2020; 26:808–810. doi: 10.1016/j.cmi.2020.04.024.

8. Jornal O Globo. Pacientes que venceram a Covid-19 morrem por infecção hospitalar por desabastecimento de antibiótico, aponta levantamento 2022. https://g1.globo.com/jornal-hoje/noticia/2021/07/23/pacientes-que-venceram-a-covid-19-morrem-por-infeccao-hospitalar-por-desabastecimento-de-antibiotico-aponta-levantamento.ghtml.

9. Beirao EM, Consentino TB, Urtado P, et al. Avaliação impacto em microorganismos colonizantes após restrição de polimixina em UTI Hospital Terciário. The Brazilian Journal of Infectious Diseases 2022; EP 006. https://doi.org/10.1016/j.bjid.2021.101742.

10. Yusef D, Hayajneh WA, Bani-Issa A, et al. Impact of an antimicrobial stewardship programme on reducing broad-spectrum antibiotic use and its effect on carbapenem-resistant Acinetobacter Baumannii (CRAb) in hospitals in Jordan. Journal of Antimicrobial Chemotherapy 2021; 76(2):516-523. https://doi.org/10.1093/jac/dkaa464.

11. Tuon FF, Yamada C, Telles J, et al. Doxiciclina oral para infecções por Acinetobacter Baumanni resistente a Carbapenem como uma estratégia de preservação de polimixina: resultados de uma corte retrospectiva. The Brazilian Journal of Infectious Diseases. 2022; 26(1): PI5. https://doi.org/10.1016/j.bjid.2021.10200.

12. Altunsoy A, Aypak C, Azap A, et al. The impact of a nationwide antibiotic restriction program on antibiotic usage and resistance against nosocomial pathogens in Turkey. International Journal of Medical Sciences 2011; 8(4):339-344. doi: 107150/ijms.8.339.

13. Curcio DJ. Antibiotic prescription in intensive care units in Latin America. Revista Argentina de Microbiología 2011; 43(3):203-211. doi:10.1590/S0325-75412011000300007.

14. Souza-Pinheiro LF, Martins CP, Martins CF, et al. Fatores de risco e mortalidade em pacientes criticamente enfermos com infecções por microrganismos multirresistentes. Revista Eletrônica Acervo Saúde 2021; 24;13(4):1:7. https://doi.org/10.25248/reas.e7319.2021.

15. Yamada T, Fujii S, Shigemi A, et al. A meta-analysis of the target trough concentration of gentamicin and amikacin for reducing the risk of nephrotoxicity. Journal of Infection and Chemotherapy 2021, 27(2):256-261. doi:10.1016/j.jiac.2020.09.033

16. Cataldo R, Arancibia M, Stojanocam J, et al. General concepts in biostatistics and clinical epidemiology: Observational studies with cross-sectional and ecological designs. Medwave 2019. v. 19, n. 8: 1-7. doi: 10.5867/medwave.2019.08.7698.

17. Rego A, Pina M, Meyer-Jr V. Quantos participantes são necessários para um estudo qualitativo? Linhas práticas de orientação. Revista de Gestão dos Países de Língua Portuguesa 2018; 17(2):43-57. doi.org/10.12660/rgplp.v17n2.2018.78224.

18. Ballinger GA. Using generalized estimating equations for longitudinal data analysis. Organizational Research Methods 2004, 7(2):127–150. https://doi: 10.1177/1094428104263672.

19. Cui J. QIC Program and model selection in GEE analyses. The Stata Journal 2007, 7 (2):209–220, https://doi:10.1177/1536867X0700700205.

20. Mendes CAC, Burdmann EA. Polimixinas. Revisão com ênfase na sua nefrotoxicidade. Revista Associação Médica Brasil 2009; 55(6):752–758. doi.org/10.1590/S0104-42302009000600023.

21. Benito LAO, Lima RC, Karnikowski MGO, Silva ICR. Comorbidades e fatores de risco identificados em pessoas que vieram a óbito por Síndrome Respiratória Aguda Grave por Covid-19. Revista de Divulgação Científica Sena Aires 2021, 10(3): 607-623.

22. Rosa RRPA; Lavareda R, Linhares JEBS. Influência das comorbidades para a ocorrência de óbitos por COVID-19 em 2020: razão de chances no estado do Amazonas. HU Revista 2022, 48:1-8.

23. Sousa-Munoz RL, Ronconi DE, Dantas GC, Lucena DMS. Impacto de multimorbidade sobre mortalidade em idosos: estudo de coorte pós-hospitalização. Revista Brasileira de Geriatria e Gerontologia 2013, 16:579-589.