ABSTRACT
Clindamycin is a common β‐lactam substitute for surgical prophylaxis. Whether peri‐operative disturbances in consciousness are disproportionately reported with clindamycin in global pharmacovigilance data remains unclear. Using WHO VigiBase (up to January 27, 2025), we performed a de‐duplicated case/non‐case disproportionality analysis of Individual Case Safety Reports (ICSRs) co‐reporting ≥ 1 anesthetic or sedative. Cases were defined by the MedDRA HLT “Disturbances in consciousness NEC.” We estimated RORs and aRORs for clindamycin versus cefazolin and other peri‐operative comparators (metronidazole, vancomycin, ampicillin), adjusting for age group, sex, and co‐reported medications. K‐modes clustering characterized co‐reporting patterns among clindamycin reports. Sensitivity analyses excluded sepsis‐related terms/vasopressors and applied a narrower outcome definition excluding syncope/presyncope. Of 3623 clindamycin ICSRs, 159 (4.4%) included a consciousness‐disturbance term. Versus cefazolin, clindamycin generated a disproportionality signal (ROR 2.59, 95% CI 2.07–3.24; aROR 2.23, 95% CI 1.70–2.91), with similar estimates after excluding sepsis/vasopressors (aROR 2.21, 95% CI 1.64–2.98) and syncope/presyncope (aROR 2.11, 95% CI 1.60–2.79). Clustering suggested co‐reporting patterns consistent with lidocaine‐enriched older‐age reports, sedative/anesthetic‐enriched younger‐age reports, and opioid‐enriched reports with high‐acuity markers. In VigiBase, disproportionality analyses identified a reporting signal for disturbances in consciousness with clindamycin relative to cefazolin in surgical settings. Because spontaneous reports lack exposure denominators and are susceptible to residual confounding, these findings are hypothesis‐generating and do not estimate incidence or comparative risk. Evaluation in data sources with denominators and prospective clinical characterization is warranted.
Keywords: cefazolin, clindamycin, consciousness disorders, pharmacovigilance, surgical prophylaxis
Study Highlights.
- What is the current knowledge on the topic?
-
○Clindamycin is widely used for surgical prophylaxis when β‐lactams are avoided, yet peri‐operative neurologic adverse event reporting patterns have not been well characterized in global pharmacovigilance data.
-
○
- What question did this study address?
-
○Whether a disproportionality reporting signal for peri‐operative disturbances in consciousness is observed for clindamycin compared with common comparator antibiotics in VigiBase.
-
○
- What does this study add to our knowledge?
-
○In WHO VigiBase analyses, clindamycin showed an adjusted disproportionality signal for disturbances in consciousness versus cefazolin (aROR 2.23, 95% CI 1.70–2.91) and clustering summarized heterogeneous co‐reporting patterns.
-
○
- How might this change clinical pharmacology or translational science?
-
○These hypothesis‐generating findings highlight the need for evaluation in data sources with exposure denominators and richer peri‐operative detail, rather than informing changes to prophylaxis selection or monitoring practices based on spontaneous reports alone.
-
○
1. Introduction
Clindamycin, a lincosamide antibiotic, is commonly used for surgical prophylaxis in patients with β‐lactam allergy and for the treatment of infections caused by Gram‐positive cocci and anaerobic bacteria, including serious soft tissue, bone, and oral infections. Despite its broad clinical application and well‐documented antimicrobial efficacy, neurological adverse events have been reported, including perioperative settings where polypharmacy is common.
Neurological adverse drug reactions (ADRs) associated with clindamycin—including reports of neuromuscular blockade—have been reported in case studies and pharmacovigilance databases [1]. Neuromuscular blockade may contribute to severe complications, including respiratory failure and altered levels of consciousness [2], which are clinically relevant considerations during anesthesia and sedation. However, disturbances of consciousness (e.g., somnolence, loss of consciousness, syncope) have not been systematically evaluated as a potential safety signal in global pharmacovigilance data for clindamycin. This gap is particularly relevant in perioperative polypharmacy, where co‐reported CNS depressants (e.g., opioids, benzodiazepines, anesthetics) may influence the occurrence, recognition, and reporting of such events [3].
The perioperative pharmacological environment is complex, involving multiple agents with overlapping sedative and neuromuscular effects. In this context, spontaneous‐report databases can be used to screen for disproportionality signals and generate hypotheses regarding drug–event co‐reporting patterns, although they cannot isolate causal effects or quantify comparative risks.
Large‐scale pharmacovigilance resources like VigiBase, the World Health Organization's global database of Individual Case Safety Reports (ICSRs), support signal detection for rare ADRs in diverse real‐world populations. Using disproportionality analysis, these databases can generate hypotheses about drug‐event reporting associations and highlight strata with higher reporting disproportionality, although they lack exposure denominators and cannot estimate incidence or comparative risk. Moreover, advanced techniques such as cluster analysis can describe distinct co‐reporting patterns within ADR reports, facilitating a more precise understanding of reporting risk factors and co‐medication patterns and potential interactions in reporting contexts.
In this study, we leveraged VigiBase data to assess whether a disproportionality signal exists for reported disturbances in consciousness in a structured manner, focusing on cases where CNS depressants were co‐reported. By comparing clindamycin to other commonly used antibiotics (cefazolin, metronidazole, vancomycin, and ampicillin) in reports involving anesthesia or sedation, we estimated reporting odds ratios (RORs) for clindamycin versus these comparators. Furthermore, we applied multivariable adjustments and unsupervised clustering to characterize report characteristics and co‐medication patterns associated with differential ADR reporting. Our findings are intended to be hypothesis‐generating and to motivate further evaluation in data sources with exposure denominators and richer clinical detail.
2. Materials and Methods
This study is reported in accordance with the STROBE Statement.
2.1. Population/Data Source
Our analyses used VigiBase, the World Health Organization's global database of Individual Case Safety Reports (ICSRs) [4]. Since its inception in 1968, VigiBase has accumulated over 41 million reports from more than 120 countries, making it one of the most comprehensive sources of real‐world pharmacovigilance data. Each ICSR includes detailed information on administrative aspects (e.g., reporting country, report type, reporter qualifications), patient demographics (e.g., sex, age), the timing of adverse reaction onset, and clinical outcomes coded using the Medical Dictionary for Regulatory Activities (MedDRA). Reports may include WHO causality assessments and drug‐related fields (e.g., drug name, indication, dates, dosage, and dechallenge/rechallenge information), although completeness varies across reports. Adverse drug reactions (ADRs) are categorized as serious or non‐serious according to WHO criteria, with serious ADRs defined as those resulting in death, life‐threatening events, hospitalization or its prolongation, persistent or significant disability/incapacity, or other clinically important outcomes, as determined by the reporting healthcare professional [5].
2.2. Case/Non‐Case Study in VigiBase
We restricted the analysis to deduplicated ICSRs that listed either clindamycin or an antibiotic comparator, identified using drug names in accordance with the Anatomical Therapeutic Chemical (ATC) classification system, from VigiBase inception until 27 January 2025. Antibiotic comparators used for the analyses were cefazolin, metronidazole, vancomycin, and ampicillin. Cases where both clindamycin and the comparator antibiotic were reported together were excluded. To focus on reports co‐listing central nervous system depressants, we restricted the dataset to reports involving at least one drug commonly used in anesthesia or sedation, including midazolam, fentanyl, propofol, lidocaine, suxamethonium, rocuronium, remifentanil, ketamine, dexmedetomidine, etomidate, or thiopental. Within this subset, we assessed whether reports listing clindamycin, compared with reports listing comparator antibiotics, showed disproportionate reporting of consciousness disturbances.
Cases were defined as ICSRs reporting “Disturbances in consciousness NEC”, based on the High‐Level Term (HLT) classification from MedDRA version 27.1. Non‐cases comprised all other ICSRs not coded with this HLT. We conducted a disproportionality analysis using a case/non‐case approach to assess reporting disproportionality of consciousness disturbances for clindamycin versus cefazolin in peri‐operative reports [6, 7, 8]. This enabled the estimation of reporting odds ratios (RORs) and corresponding 95% confidence intervals (CIs).
2.3. Comparator Rationale
We selected cefazolin, metronidazole, vancomycin, and ampicillin as comparators to contextualize the clindamycin signal against antibiotics with overlapping peri‐operative indications. Cefazolin is commonly used for prophylaxis in clean and clean‐contaminated procedures; metronidazole is used for anaerobic coverage; and vancomycin is used when MRSA coverage is desired or when β‐lactams are avoided. However, we acknowledge that VigiBase lacks exposure denominators (total number of patients treated). Therefore, direct comparisons of RORs between drugs reflect differences in reporting probability rather than incidence or comparative risk. Variability in regional guidelines and availability (e.g., preference for cefazolin in the US vs. other agents globally) may influence reporting patterns.
2.4. Statistical Analysis
Data were expressed as means ± SD or percentage. Age and sex variables were cleaned and categorized, with age grouped into four categories: under 18, 18–45, 45–64, and 65 years or older. Sex was treated as missing when coded as unknown or ambiguous. For the case/non‐case analysis, we estimated the ROR with 95% CI for each ADR under study, and a disproportionality signal was defined as an ROR > 1 and a lower 95% CI bound > 1. Furthermore, multivariable analyses were performed, adjusting for age and sex, and each of the CNS depressant drugs defined in the above section to account for measured covariates using adjusted reporting odds ratio (aRORs). A main sensitivity analysis was performed by expanding the perioperative drug list to include quaternary ammonium neuromuscular blockers (e.g., pancuronium, vecuronium, atracurium defined as the ATC M03AC class), which were commonly used in earlier decades. This allowed us to assess whether broader drug inclusion affected the magnitude or significance of the disproportionality signal. To assess the robustness of the findings, two additional sensitivity analyses were performed. First, to address potential confounding by indication where clindamycin might be preferentially used in septic or hemodynamically unstable patients, we repeated the analysis excluding reports containing MedDRA Preferred Terms for sepsis (Sepsis, Septic shock, Bacteraemia, Viraemia, Fungaemia) or reporting the use of vasopressors (norepinephrine, epinephrine, vasopressin, dopamine, dobutamine, phenylephrine). Second, to reduce potential influence of cardiovascular events on the outcome definition, we repeated the analysis using a “narrow” definition of consciousness disturbances that excluded the Preferred Terms Syncope and Presyncope.
To examine distinct report characteristics among reports with disturbances in consciousness in clindamycin ICSRs, we performed a cluster analysis using the K‐modes algorithm, which is specifically suited for categorical data. The analysis was conducted on variables such as sex, age group, and co‐reported medications. To determine the optimal number of clusters, we applied the elbow method, evaluating within‐cluster dissimilarity across models with 1 to 10 clusters. Once clusters were fixed, comparisons were made between them to assess potential differences in characteristics. Categorical variables were compared using the chi‐squared test without continuity correction when all expected cell counts were ≥ 5. When expected counts were < 5 in any cell, Fisher's exact test was used instead. Continuous variables were compared using the Wilcoxon rank‐sum test for two‐group comparisons or the Kruskal–Wallis test when more than two groups were present. The significance level for two‐sided tests was set at 0.05.
The analyses adhered to the READUS‐PV guidelines with an execution in R 4.4.3 [9, 10].
3. Results
3.1. Descriptive Characteristics
VigiBase ICSRs listing clindamycin in the selected population totaled 3623. Adhering to the selection criteria, we identified 159 ICSRs coded as “Disturbances in consciousness NEC” (Figure 1). Most reported consciousness disturbance terms included somnolence (n = 51), loss of consciousness (n = 37) and syncope (n = 37) (Table 1). Consciousness‐disturbance reports listing clindamycin predominantly involved female patients (62%), relatively consistent with other antibiotics (47% to 63%). They were most reported in the 45–64 years old age group (38%) followed by the 65+. In all antibiotics, reports most commonly involved patients aged ≥ 45 years. Eighty‐four per cent of the reports were considered serious, relatively similar to other antibiotics (from 80% to 92%). However, the pattern of seriousness criteria differed between antibiotics, with caused/prolonged hospitalization being the most reported with clindamycin (39%). The most frequently co‐reported drug was fentanyl for all antibiotics. Lidocaine, midazolam, paracetamol and furosemide were also frequently co‐reported.
FIGURE 1.
Flow chart of ICSR selection from WHO VigiBase for clindamycin and the outcome definition (“Disturbances in consciousness NEC”).
TABLE 1.
Characteristics of individual case safety reports (ICSRs) coded as “Disturbances in consciousness NEC” and listing clindamycin or comparator antibiotics in VigiBase.
| Clindamycin b | Cefazolin | Metronidazole | Vancomycin | Ampicillin | ||
|---|---|---|---|---|---|---|
| Number of reports | 148 | 168 | 193 | 252 | 68 | |
| 5 Most reported consciousness disturbances terms (details) | Somnolence N = 51 (32%) | Somnolence N = 39 (23%) | Somnolence N = 75 (39%) | Somnolence N = 74 (29%) | Somnolence N = 16 (24%) | |
| Loss of consciousness N = 37 (23%) | Loss of consciousness N = 32 (19%) | Syncope N = 39 (20%) | Lethargy N = 45 (18%) | Altered state of consciousness N = 16 (24%) | ||
| Syncope N = 37 (23%) | Depressed level of consciousness N = 28 (17%) | Loss of consciousness N = 37 (19%) | Depressed level of consciousness N = 39 (15%) | Depressed level of consciousness N = 11 (16%) | ||
| Depressed level of consciousness N = 18 (11%) | Lethargy N = 17 (10%) | Lethargy N = 31 (16%) | Loss of consciousness N = 39 (15%) | Loss of consciousness N = 10 (15%) | ||
| Lethargy N = 17 (11%) | Altered state of consciousness N = 15 (9%) | Depressed level of consciousness N = 17 (9%) | Syncope N = 39 (15%) | Lethargy N = 6 (9%) | ||
| Sex | Not specified | 145 | 164 | 188 | 244 | 66 |
| Female | 90 (62%) | 85 (52%) | 118 (63%) | 131 (54%) | 31 (47%) | |
| Age | Not specified | 117 | 150 | 155 | 203 | 60 |
| 0–18 | 10 (8.5%) | 11 (7.3%) | 14 (9.0%) | 20 (9.9%) | 14 (23%) | |
| 18–45 | 26 (22%) | 35 (23%) | 42 (27%) | 30 (15%) | 10 (17%) | |
| 45–64 | 45 (38%) | 46 (31%) | 53 (34%) | 79 (39%) | 10 (17%) | |
| 65+ | 36 (31%) | 58 (39%) | 46 (30%) | 74 (36%) | 26 (43%) | |
| Serious | N available | 144 | 142 | 172 | 238 | 58 |
| Yes | 121 (84%) | 121 (85%) | 138 (80%) | 219 (92%) | 50 (86%) | |
| Seriousness criteria a | Not specified | 29 (20%) | 40 (28%) | 36 (21%) | 33 (14%) | 29 (50%) |
| Death | 18 (13%) | 24 (17%) | 42 (24%) | 83 (35%) | 8 (14%) | |
| Life threatening | 10 (6.9%) | 22 (15%) | 19 (11%) | 28 (12%) | 5 (8.6%) | |
| Caused/Prolonged Hospitalization | 56 (39%) | 35 (25%) | 56 (33%) | 73 (31%) | 13 (22%) | |
| Disabling/Incapacitating | 3 (2.1%) | 3 (2.1%) | 0 (0%) | 1 (0.4%) | 0 (0%) | |
| Other | 28 (19%) | 18 (13%) | 19 (11%) | 19 (8.0%) | 3 (5.2%) | |
| 5 Most co‐reported drugs | Fentanyl N = 73 (49%) | Fentanyl N = 105 (63%) | Fentanyl N = 108 (56%) | Fentanyl N = 147 (58%) | Fentanyl N = 24 (35%) | |
| Lidocaine N = 64 (43%) | Midazolam N = 78 (46%) | Paracetamol N = 76 (40%) | Paracetamol N = 91 (36%) | Midazolam N = 24 (35%) | ||
| Paracetamol N = 47 (32%) | Propofol N = 63 (38%) | Lidocaine N = 66 (34%) | Midazolam N = 87 (34%) | Furosemide N = 23 (34%) | ||
| Furosemide N = 47 (32%) | Paracetamol N = 45 (27%) | Furosemide N = 57 (30%) | Furosemide N = 82 (33%) | Paracetamol N = 20 (29%) | ||
| Gabapentin N = 45 (30%) | Lidocaine N = 44 (26%) | Morphine N = 57 (30%) | Pantoprazole N = 71 (28%) | Propofol N = 14 (21%) | ||
Several seriousness criteria can be added at the same time.
The number of reports presented here corresponds to comparisons with cefazolin. This number may vary slightly when using other antibiotics as comparators, as reports including both clindamycin and the comparator antibiotic are excluded from the analysis.
3.2. Univariate and Multivariable Case/Non‐Case Analysis in VigiBase
Table 2 illustrates that compared to each of the 4 comparator antibiotics, a disproportionality signal for consciousness disturbances was observed for clindamycin in univariate analyses (RORs ranged from 1.22 to 2.59) and multivariable analysis (aRORs ranged from 1.23 to 2.23) post‐adjustment for age group, sex, co‐reported central nervous system depressant drugs commonly used in anesthesia or sedation, and remained significant only versus cefazolin and metronidazole. Similar results were observed in the main sensitivity analysis (Table S1). To assess the robustness of the primary signal, two additional sensitivity analyses were performed (Table S2). First, in the “low‐acuity” subset excluding sepsis and vasopressor use, the disproportionality signal for clindamycin versus cefazolin remained (aROR 2.21, 95% CI 1.64–2.98). Second, when applying a “narrow” definition excluding syncope to reduce potential influence of cardiovascular etiologies, the disproportionality signal remained significant (aROR 2.11, 95% CI 1.60–2.79).
TABLE 2.
Disproportionality analysis assessing a reporting signal for disturbances in consciousness for clindamycin versus four comparator antibiotics in VigiBase.
| Comparator | Univariate analysis | Multivariable analysis a | ||
|---|---|---|---|---|
| ROR b | 95% CI | aROR | 95% CI | |
| Cefazolin | 2.59 | (2.07–3.24) | 2.23 | (1.70–2.91) |
| Metronidazole | 1.27 | (1.02–1.59) | 1.28 | (1.00–1.65) |
| Vancomycin | 1.22 | (0.99–1.51) | 1.23 | (0.95–1.58) |
| Ampicillin | 1.55 | (1.16–2.07) | 1.37 | (0.97–1.92) |
Adjusted for age category, sex, and a list of central nervous system depressant drugs commonly used in anesthesia or sedation (see methods section).
Reporting odds ratio.
3.3. Cluster Analysis
Among 2502 clindamycin ICSRs, the optimized number of clusters was three: Cluster 1 (n = 1162), Cluster 2 (n = 446), and Cluster 3 (n = 894) (Table 3). Cluster 2 had the lowest proportion of reports coded with “Disturbances in consciousness NEC” (2.7%), compared to 5.0% in Cluster 1 and 4.7% in Cluster 3 (p = 0.13).
TABLE 3.
Report characteristics by K‐modes cluster among clindamycin ICSRs.
| By Cluster | N total 2502 | |||
|---|---|---|---|---|
| 1 N = 1,162 a | 2 N = 446 a | 3 N = 894 a | p‐value b | |
| Consciousness disturbances, yes | 58 (5.0%) | 12 (2.7%) | 42 (4.7%) | 0.13 |
| Type | ||||
| Not available to sender (unknown) | 1 (< 0.1%) | 10 (2.2%) | 7 (0.8%) | |
| Other | 13 (1.1%) | 12 (2.7%) | 15 (1.7%) | |
| Report from study | 287 (25%) | 43 (9.6%) | 168 (19%) | |
| Spontaneous | 861 (74%) | 381 (85%) | 704 (79%) | |
| Region | < 0.001 | |||
| African Region | 0 (0%) | 2 (0.4%) | 0 (0%) | |
| Eastern Mediterranean Region | 6 (0.5%) | 5 (1.1%) | 8 (0.9%) | |
| European Region | 219 (19%) | 93 (21%) | 181 (20%) | |
| Region of the Americas | 878 (76%) | 257 (58%) | 579 (65%) | |
| South‐East Asia Region | 4 (0.3%) | 1 (0.2%) | 6 (0.7%) | |
| Western Pacific Region | 55 (4.7%) | 88 (20%) | 120 (13%) | |
| Age category | < 0.001 | |||
| 0–18 | 69 (5.9%) | 45 (10%) | 76 (8.5%) | |
| 18–45 | 246 (21%) | 269 (60%) | 117 (13%) | |
| 45–64 | 482 (41%) | 41 (9.2%) | 484 (54%) | |
| 65+ | 365 (31%) | 91 (20%) | 217 (24%) | |
| Sex, female | 742 (64%) | 281 (63%) | 490 (55%) | < 0.001 |
| Seriousness | < 0.001 | |||
| Caused/Prolonged Hospitalization | 452 (39%) | 11 (2.5%) | 205 (23%) | |
| Death | 91 (7.8%) | 13 (2.9%) | 130 (15%) | |
| Disabling/Incapacitating | 18 (1.5%) | 1 (0.2%) | 11 (1.2%) | |
| Life threatening | 70 (6.0%) | 26 (5.8%) | 85 (9.5%) | |
| Other | 302 (26%) | 31 (7.0%) | 186 (21%) | |
| Serious, yes | 962 (83%) | 100 (22%) | 693 (78%) | < 0.001 |
| Midazolam, yes | 132 (11%) | 176 (39%) | 266 (30%) | < 0.001 |
| Fentanyl, yes | 96 (8.3%) | 313 (70%) | 731 (82%) | < 0.001 |
| Propofol, yes | 172 (15%) | 249 (56%) | 104 (12%) | < 0.001 |
| Lidocaine, yes | 962 (83%) | 56 (13%) | 75 (8.4%) | < 0.001 |
| Suxamethonium, yes | 32 (2.8%) | 13 (2.9%) | 20 (2.2%) | 0.7 |
| Rocuronium, yes | 44 (3.8%) | 78 (17%) | 47 (5.3%) | < 0.001 |
| Remifentanil, yes | 34 (2.9%) | 37 (8.3%) | 21 (2.3%) | < 0.001 |
| Ketamine, yes | 49 (4.2%) | 26 (5.8%) | 36 (4.0%) | 0.3 |
| Dexmedetomidine, yes | 25 (2.2%) | 31 (7.0%) | 48 (5.4%) | < 0.001 |
| Etomidate, yes | 9 (0.8%) | 12 (2.7%) | 10 (1.1%) | 0.007 |
| Thiopental, yes | 7 (0.6%) | 4 (0.9%) | 11 (1.2%) | 0.3 |
n (%).
Pearson's Chi‐squared test; Fisher's exact test.
Reports in Cluster 2 more frequently involved younger age groups, with 60% aged 18–45 years, while only 21% and 13% of patients in Clusters 1 and 3 fell within this age group, respectively. In contrast, reports involving patients aged 45 years or older represented 29% in Cluster 2, compared to 72% and 76% in Clusters 1 and 3, respectively. Sex distributions differed across clusters, with 63% females compared to 64% for Cluster 1 and 55% for Cluster 3 (p < 0.001).
The region of origin also differed. Reports in Cluster 2 came more frequently from the Western Pacific Region (20% vs. 4.7% and 13%) and less frequently from the Region of the Americas (58% vs. 76% and 65%), with statistically significant differences (p < 0.001). Regarding the type of report, Cluster 2 included fewer reports from studies (9.6% vs. 25% and 22%) and more spontaneous reports (85% vs. 74% and 79%), again with a significant difference between groups.
Reports in Cluster 2 were less frequently classified as serious. Only 22% were considered serious reports, compared to 83% and 78% in Clusters 1 and 3, respectively. Hospitalization or prolonged hospitalization was reported in just 2.5% of Cluster 2 reports, whereas 39% and 23% of reports in Clusters 1 and 3, respectively. Other seriousness criteria such as life‐threatening events or disabling/incapacitating outcomes were also less frequent.
Cluster 2 was characterized by a markedly higher co‐reporting of several anesthetic or sedative agents. Midazolam was reported in 39% of Cluster 2 reports compared to 11% in Clusters 1 and 3. Fentanyl was reported in 70% vs. 13%, propofol in 56% vs. 14%, rocuronium in 17% vs. 5%, remifentanil in 8.3% vs. 2.6%, and dexmedetomidine in 7.0% vs. 3.4%. Conversely, lidocaine was less frequently reported in Cluster 2 (13%) than in Clusters 1 and 3 (83%). The frequency of ketamine, suxamethonium, thiopental, and etomidate was similar across clusters, with no statistically significant differences in these reports.
Cluster 1, which showed a comparable proportion of consciousness disturbance reports to Cluster 3, was characterized by notably lower co‐reporting of midazolam and fentanyl. Midazolam, a benzodiazepine, appeared in only 11% of Cluster 1 reports, markedly less than in Cluster 2, where it was present in 39% of reports and Cluster 3 (30%). Fentanyl, a potent opioid analgesic, followed the same pattern—reported in only 8.3% of Cluster 1 reports, but sharply higher in Cluster 2 at 70% and Cluster 3 at 82%. In contrast, lidocaine—a local anesthetic also used in antiarrhythmic therapy—was more frequently co‐reported in Cluster 1, appearing in 83% of reports, versus only 13% in Cluster 2 and 8.4% in Cluster 3.
Cluster 2 was characterized by frequent co‐reporting of agents typically used in anesthetic induction and intensive care. Propofol, a fast‐acting general anesthetic, was co‐reported in 56% of Cluster 2 reports, in contrast to only 15% in Cluster 1 and 12% in Cluster 3. Rocuronium, a neuromuscular blocker, was reported in 17% of Cluster 2 reports, compared to 3.8% and 5.3% in Clusters 1 and 3, respectively. Remifentanil, an ultra‐short‐acting opioid, appeared in 8.3% of Cluster 2 reports, versus 2.9% in Cluster 1 and 2.3% in Cluster 3. Similarly, dexmedetomidine, a sedative, was more frequent in Cluster 2 (7.0%) than in Cluster 1 (2.2%) or Cluster 3 (5.4%). Etomidate, a general anesthetic agent, was also reported more in Cluster 2 (2.7%) compared to the other clusters (0.8% and 1.1% in Clusters 1 and 3, respectively).
4. Discussion
Using > 41 million ICSRs from VigiBase we identified a consistent disproportionality signal for disturbances in consciousness reported with clindamycin compared with cefazolin (aROR 2.23, 95% CI 1.70–2.91). Over 80% of reports met WHO seriousness criteria [5] and 13% were fatal. The signal persisted after adjustment for ten commonly co‐administered sedatives and neuromuscular blockers, and remained robust in sensitivity analyses excluding sepsis/vasopressors and using a narrower outcome definition [7]. Notably, the absolute number of reports was similar for clindamycin (n = 148) and cefazolin (n = 168), despite differences in utilization and reporting contexts between drugs. In a 2011 multicenter study of U.S. hospitals, cefazolin accounted for 9.3% of antibiotic prescriptions (59.6% for surgical prophylaxis), while clindamycin represented only 2.9% (6.4% for surgical prophylaxis) [11]. These differences in utilization underscore that spontaneous reports lack aligned exposure denominators; therefore, absolute report counts and cross‐drug comparisons should not be interpreted as incidence or comparative risk.
Unsupervised K‐modes clustering identified three reproducible co‐reporting patterns that further describe the primary signal. Although the proportion of consciousness disturbances did not differ significantly between clusters (p = 0.13), the clustering analysis revealed distinct report characteristics and co‐reported medication patterns. The following proposed interpretations are exploratory and intended to generate hypotheses for future investigation. While causality cannot be inferred from spontaneous report data, the clustering analysis highlights clinical contexts where reports that include clindamycin and disturbances in consciousness may be differentially reported, motivating further evaluation in data sources with richer clinical detail.
Cluster 1 – “traditional Operating Room practice.” Reports in older age groups with frequent lidocaine co‐reporting and less frequent midazolam or fentanyl co‐reporting had the highest proportion classified as serious (39% prolonged hospitalization). This pattern may reflect peri‐operative polypharmacy contexts in which disturbances in consciousness are reported alongside clindamycin, motivating further evaluation; mechanistic interpretation cannot be established from spontaneous reports [12].
Cluster 2 – “balanced intravenous anaesthesia.” Despite the highest co‐reporting of propofol, midazolam and remifentanil, this cluster, which was enriched for younger age groups and fewer seriousness classifications, had the lowest absolute reporting rate (2.7%). Two non‐mutually exclusive possibilities are: masking by expected deep sedation [13] and a healthy‐user effect. Both imply under‐recognition rather than absence of toxicity and highlight a context in which disturbances in consciousness may be differentially reported, motivating prospective evaluation, including whether monitoring strategies influence detection and reporting [14].
Cluster 3 – “opioid‐intensive high‐acuity.” Reports in Cluster 3 more frequently involved older age groups and fentanyl co‐reporting (> 80%) and more often included fatal outcomes (15%). This co‐reporting pattern is hypothesis‐generating; pharmacodynamic interactions between clindamycin and opioids have been proposed [15], but spontaneous reports cannot establish mechanisms or causal pathways.
Across clusters, lidocaine probably behaved as a marker of infiltration technique rather than a neuro‐protective factor, whereas benzodiazepines/opioids may have influenced the likelihood of recognition and reporting of disturbances in consciousness. These patterns suggest possible avenues for future investigation into hypotheses regarding reporting heterogeneity. Specifically, elderly patients receiving infiltration anesthesia, those undergoing high‐sedation ambulatory procedures, and patients in opioid‐intensive surgical contexts represent clinical contexts in which reporting patterns differed and may be prioritized for prospective characterization. Prospective studies with exposure denominators and richer clinical detail will be necessary to evaluate whether these patterns reflect modifiable clinical risks. Several pharmacodynamic mechanisms have been proposed that could be consistent with this reporting signal. First, clindamycin has been reported to attenuate presynaptic calcium influx and enhance the effect of non‐depolarising neuromuscular blockers, prolonging paralysis and delaying emergence [16]. Second, clindamycin is known to interfere with neuromuscular transmission through peripheral mechanisms, primarily by inhibiting presynaptic acetylcholine release and possibly affecting postsynaptic receptor sensitivity [17]. Under conditions such as surgical stress or systemic inflammation, its lipophilic nature may enhance central nervous system penetration [18, 19, 20]. While clindamycin's primary pharmacological actions are not centrally mediated, CNS access raises the possibility of central effects that could potentiate or unmask interactions with other agents acting on arousal, vigilance, or respiratory control pathways. These potential central contributions, in conjunction with its peripheral neuromuscular effects, may influence the recognition and reporting of events in the presence of sedative or depressant co‐medications. Further investigation is warranted to clarify the extent and clinical significance of these possible interactions. Third, pharmacodynamic interactions involving clindamycin have been proposed to displace free calcium and exacerbate opioid‐induced chest‐wall rigidity, a phenomenon described in prior literature and potentially relevant to the high‐mortality, fentanyl‐rich Cluster 3. These pathways are mechanistically coherent with older pharmacodynamic studies and are hypothesis‐generating considerations that should be evaluated in studies with richer clinical detail.
Cefazolin is commonly recommended as first‐line prophylaxis for clean and clean‐contaminated surgery [21], yet up to 10% of patients labeled “penicillin‐allergic” receive clindamycin [22]. Two areas are relevant for peri‐operative antibiotic decision‐making and warrant further study. First, routine allergy verification or a graded cephalosporin challenge [23] may allow safe use of cefazolin in many cases, potentially reducing reliance on alternatives for which neurological reporting signals have been raised in spontaneous‐report data. Moreover, clindamycin has been associated with a higher rate of surgical site infections compared to other antibiotics [24, 25], which is relevant to antibiotic selection and should be considered alongside local guidelines and patient‐level factors. Second, in situations where clindamycin use is necessary, the role of co‐administered sedatives and opioids, and how co‐medication patterns relate to reporting of disturbances in consciousness should be evaluated in prospective studies with richer peri‐operative detail. Prospective studies with standardized peri‐operative data elements would help characterize timing, alternative explanations, and co‐medication patterns. These considerations could complement the β‐lactam stewardship goals outlined in the 2013 IDSA/ASHP prophylaxis update [21] and can inform future work that integrates pharmacovigilance signals with guideline development.
This study shares the intrinsic limitations of spontaneous‐report analyses: under‐reporting, variable data quality, absence of exposure denominators and potential confounding by indication or frailty [26]. Additionally, we could not fully adjust for disease severity or rule out stimulated reporting, where awareness of a drug's safety profile influences reporting rates. We lacked granular timing data to distinguish intra‐operative from post‐operative events and cannot establish causality. Consequently, these findings should be interpreted strictly as hypothesis‐generating and should not be used to change peri‐operative prophylaxis policies in isolation [27]. Nevertheless, the analysis benefits from the large global pharmacovigilance resource, strict deduplication, comparator‐controlled disproportionality [9], and consistent findings across comparator antibiotics and clustering analyses. The strength and consistency of the signal, together with proposed pharmacodynamic mechanisms, support prioritizing further evaluation in data sources with exposure denominators before drawing practice‐changing conclusions.
In summary, among > 3600 clindamycin ICSRs in WHO VigiBase, disproportionality analyses identified a consistent reporting signal for disturbances in consciousness compared with comparator antibiotics.
Data‐driven clustering described heterogeneous co‐reporting patterns across age groups and co‐medication contexts, including lidocaine‐enriched reports and opioid‐enriched reports. These descriptive patterns are hypothesis‐generating and motivate evaluation in studies with exposure denominators and richer peri‐operative detail, rather than changes to prophylaxis policies or monitoring practices based on spontaneous reports alone. Prospective studies, including registry or EHR‐based comparative cohorts, are needed to evaluate whether this reporting signal corresponds to clinically meaningful differences when exposure denominators and key confounders can be addressed. To evaluate this signal, future research should utilize electronic health record (EHR) datasets or prospective registries where denominators are known. Such studies could account for surgical duration, depth of anesthesia, and specific surgical indications, thereby better accounting for the confounding factors that cannot be fully controlled in a spontaneous reporting analysis.
Author Contributions
K.N., B.C., H.M., D.T., K.H., T.N., S.M., M.K., and Y.H. wrote the manuscript; K.N. and Y.H. designed the research; K.N., B.C., and Y.H. performed the research; K.N., B.C., and Y.H. analyzed the data.
Funding
This work was supported by the Japan Society for the Promotion of Science (JSPS) KAKENHI (Grant Number JP26K20575).
Ethics Statement
This study analyzed de‐identified Individual Case Safety Reports from WHO VigiBase.
Consent
Individual consent was not applicable because the study used de‐identified Individual Case Safety Reports (ICSRs) from the WHO VigiBase database.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Table S1: Disproportionality analysis with clindamycin compared to four different antibiotics in VigiBase to search for a signal of consciousness disturbances.
Table S2: Sensitivity analyses addressing confounding by indication (sepsis/vasopressors) and event definition specificity (excluding syncope).
Acknowledgments
We gratefully acknowledge the World Health Organization (WHO) Programme for International Drug Monitoring and the Uppsala Monitoring Centre (UMC) for providing access to data from VigiBase, the WHO global database of individual case safety reports. The information supplied by VigiBase does not represent the opinion of the UMC or the WHO, and the authors are solely responsible for the content of this publication. We used an AI language model (ChatGPT, OpenAI) for language editing and to assist with programming (e.g., code suggestions and debugging) during data analysis; all analytic decisions, final code, and results were written, verified, and approved by the authors.
Data Availability Statement
Individual Case Safety Reports (ICSRs) analyzed in this study are held in the WHO VigiBase database and are not publicly available owing to licensing restrictions. Aggregated data supporting the findings of this study are available from the corresponding author upon reasonable request.
References
- 1. Mattappalil A. and Mergenhagen K. A., “Neurotoxicity With Antimicrobials in the Elderly: A Review,” Clinical Therapeutics 36 (2014): 1489–1511.e4, 10.1016/j.clinthera.2014.09.020. [DOI] [PubMed] [Google Scholar]
- 2. Shah M. U., “Chapter 11 ‐ Neuromuscular Blocking Agents and Skeletal Muscle Relaxants,” in Side Effects of Drugs Annual (Elsevier, 2016), 105–113, 10.1016/bs.seda.2016.07.012. [DOI] [Google Scholar]
- 3. Prielipp R. C., Coursin D. B., Wood K. E., and Murray M. J., “Complications Associated With Sedative and Neuromuscular Blocking Drugs in Critically Ill Patients,” Critical Care Clinics 11 (1995): 983–1003. [PubMed] [Google Scholar]
- 4. Bate A., Lindquist M., and Edwards I. R., “The Application of Knowledge Discovery in Databases to Post‐Marketing Drug Safety: Example of the WHO Database,” Fundamental & Clinical Pharmacology 22 (2008): 127–140, 10.1111/j.1472-8206.2007.00552.x. [DOI] [PubMed] [Google Scholar]
- 5. Edwards I. R. and Aronson J. K., “Adverse Drug Reactions: Definitions, Diagnosis, and Management,” Lancet 356 (2000): 1255–1259, 10.1016/S0140-6736(00)02799-9. [DOI] [PubMed] [Google Scholar]
- 6. Egberts A. C. G., Meyboom R. H. B., and van Puijenbroek E. P., “Use of Measures of Disproportionality in Pharmacovigilance: Three Dutch Examples,” Drug Safety 25 (2002): 453–458, 10.2165/00002018-200225060-00010. [DOI] [PubMed] [Google Scholar]
- 7. Montastruc J.‐L., Sommet A., Bagheri H., and Lapeyre‐Mestre M., “Benefits and Strengths of the Disproportionality Analysis for Identification of Adverse Drug Reactions in a Pharmacovigilance Database,” British Journal of Clinical Pharmacology 72 (2011): 905–908, 10.1111/j.1365-2125.2011.04037.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Faillie J.‐L., “Case‐Non‐Case Studies: Principle, Methods, Bias and Interpretation,” Thérapie 74 (2019): 225–232, 10.1016/j.therap.2019.01.006. [DOI] [PubMed] [Google Scholar]
- 9. Fusaroli M., Salvo F., Begaud B., et al., “The REporting of A Disproportionality Analysis for DrUg Safety Signal Detection Using Individual Case Safety Reports in PharmacoVigilance (READUS‐PV): Explanation and Elaboration,” Drug Safety 47 (2024): 585–599, 10.1007/s40264-024-01423-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Fusaroli M., Salvo F., Begaud B., et al., “The Reporting of a Disproportionality Analysis for Drug Safety Signal Detection Using Individual Case Safety Reports in PharmacoVigilance (READUS‐PV): Development and Statement,” Drug Safety 47, no. 6 (2024): 575–584, 10.1007/s40264-024-01421-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Magill S. S., Edwards J. R., Beldavs Z. G., et al., “Prevalence of Antimicrobial Use in U.S. Acute Care Hospitals, May–September 2011,” Journal of the American Medical Association 312 (2014): 1438–1446, 10.1001/jama.2014.12923. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Wu G., Wu G., and Wu H., “A Costly Lesson: Fatal Respiratory Depression Induced by Clindamycin During Postoperative Patient Controlled Analgesia,” Pain Physician 18 (2015): E429–E431. [PubMed] [Google Scholar]
- 13. Godoroja‐Diarto D., Constantin A., Moldovan C., Rusu E., and Sorbello M., “Efficacy and Safety of Deep Sedation and Anaesthesia for Complex Endoscopic Procedures‐A Narrative Review,” Diagnostics (Basel) 12 (2022): 1523, 10.3390/diagnostics12071523. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Myles P. S., Leslie K., McNeil J., Forbes A., and Chan M. T. V., “Bispectral Index Monitoring to Prevent Awareness During Anaesthesia: The B‐Aware Randomised Controlled Trial,” Lancet 363 (2004): 1757–1763, 10.1016/S0140-6736(04)16300-9. [DOI] [PubMed] [Google Scholar]
- 15. Rosal N. R., Thelmo F. L., Tzarnas S., DiCalvo L., Tariq S., and Grossman C., “Wooden Chest Syndrome: A Case Report of Fentanyl‐Induced Chest Wall Rigidity,” Journal of Investigative Medicine High Impact Case Reports 9 (2021): 23247096211034036, 10.1177/23247096211034036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Lee J. H., Lee S.‐I., Chung C. J., et al., “The Synergistic Effect of Gentamicin and Clindamycin on Rocuronium‐Induced Neuromuscular Blockade,” Korean Journal of Anesthesiology 64 (2013): 143–151, 10.4097/kjae.2013.64.2.143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Wright J. M. and Collier B., “Characterization of the Neuromuscular Block Produced by Clindamycin and Lincomycin,” Canadian Journal of Physiology and Pharmacology 54 (1976): 937–944, 10.1139/y76-130. [DOI] [PubMed] [Google Scholar]
- 18. Gatti G., Malena M., Casazza R., Borin M., Bassetti M., and Cruciani M., “Penetration of Clindamycin and Its Metabolite N‐Demethylclindamycin Into Cerebrospinal Fluid Following Intravenous Infusion of Clindamycin Phosphate in Patients With AIDS,” Antimicrobial Agents and Chemotherapy 42 (1998): 3014–3017, 10.1128/AAC.42.11.3014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Smith M. J., Gonzalez D., Goldman J. L., et al., “Pharmacokinetics of Clindamycin in Obese and Nonobese Children,” Antimicrobial Agents and Chemotherapy 61 (2017): e02014‐16, 10.1128/AAC.02014-16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Alam El‐Din L. W., El‐Bassiouny N. A., Khedr W. M., and Werida R. H., “The Potential Impact of Clindamycin on Neurosurgery Patients: A Randomized Controlled Trial,” Inflammopharmacology 33, no. 7 (2025): 4051–4061, 10.1007/s10787-025-01810-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Bratzler D. W., Dellinger E. P., Olsen K. M., et al., “Clinical Practice Guidelines for Antimicrobial Prophylaxis in Surgery,” American Journal of Health‐System Pharmacy 70 (2013): 195–283, 10.2146/ajhp120568. [DOI] [PubMed] [Google Scholar]
- 22. Blumenthal K. G., Ryan E. E., Li Y., Lee H., Kuhlen J. L., and Shenoy E. S., “The Impact of a Reported Penicillin Allergy on Surgical Site Infection Risk,” Clinical Infectious Diseases 66 (2018): 329–336, 10.1093/cid/cix794. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Blumenthal K. G., Li Y., Hsu J. T., et al., “Outcomes From an Inpatient Beta‐Lactam Allergy Guideline Across a Large US Health System,” Infection Control and Hospital Epidemiology 40 (2019): 528–535, 10.1017/ice.2019.50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Lam P. W., Tarighi P., Elligsen M., et al., “Self‐Reported Beta‐Lactam Allergy and the Risk of Surgical Site Infection: A Retrospective Cohort Study,” Infection Control and Hospital Epidemiology 41 (2020): 438–443, 10.1017/ice.2019.374. [DOI] [PubMed] [Google Scholar]
- 25. Pool C., Kass J., Spivack J., et al., “Increased Surgical Site Infection Rates Following Clindamycin Use in Head and Neck Free Tissue Transfer,” Otolaryngology and Head and Neck Surgery 154 (2016): 272–278, 10.1177/0194599815617129. [DOI] [PubMed] [Google Scholar]
- 26. Palleria C., Leporini C., Chimirri S., et al., “Limitations and Obstacles of the Spontaneous Adverse Drugs Reactions Reporting: Two “Challenging” Case Reports,” Journal of Pharmacology and Pharmacotherapeutics 4 (2013): S66–S72, 10.4103/0976-500X.120955. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Potter E., Reyes M., Naples J., and Dal Pan G., “FDA Adverse Event Reporting System (FAERS) Essentials: A Guide to Understanding, Applying, and Interpreting Adverse Event Data Reported to FAERS,” Clinical Pharmacology and Therapeutics 118 (2025): 567–582, 10.1002/cpt.3701. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1: Disproportionality analysis with clindamycin compared to four different antibiotics in VigiBase to search for a signal of consciousness disturbances.
Table S2: Sensitivity analyses addressing confounding by indication (sepsis/vasopressors) and event definition specificity (excluding syncope).
Data Availability Statement
Individual Case Safety Reports (ICSRs) analyzed in this study are held in the WHO VigiBase database and are not publicly available owing to licensing restrictions. Aggregated data supporting the findings of this study are available from the corresponding author upon reasonable request.