Surgical frequency analysis of patients clustered according to postoperative pain trajectory: a retrospective study (2025)

The growing prevalence of oropharyngeal cancer and orthognathic surgeries has drawn attention to the need for effective postoperative pain management strategies. These procedures, due to their invasive nature and extensive surgical areas, are associated with significantly higher postoperative pain compared to dentofacial surgeries that have traditionally been performed. Furthermore, even within the same category, variations in surgical techniques and procedural details can result in markedly different pain experiences^1,2,3. This variability underscores the importance of considering specific surgical details when developing tailored pain management strategies.

According to established methods of postoperative pain management, in cases of severe pain, adjunct analgesics such as antidepressants or anticonvulsants can be utilized to reduce opioid usage while enhancing the descending inhibitory pathway and suppressing the excitatory pathway, thereby providing analgesic effects. Furthermore, the local anesthetic infiltration into the surgical area intra-and postoperatively facilitates postoperative pain control^4,5,6. Recognition of the need for multimodal and preemptive analgesia to control postoperative pain is growing^{4,5,6,7,8,9,10}. However, current research lacks sufficient analysis of postoperative pain patterns for dental surgeries, making it challenging for clinicians to explain postoperative pain to patients and manage it preemptively.

In this study, we aimed to assess the Visual Analog Scale (VAS) scores at 0, 6, 12, 24, and 36h postoperatively, cluster patients with similar changing trends in postoperative VAS scores, identify differences in pain changes among clusters and analyze the frequency of surgeries corresponding to each cluster. Beyond previous analyses, this study will contribute to postoperative pain prediction based on the type of surgery, ultimately aiding clinicians in managing postoperative pain in patients.

A total of 1,019 patients undergoing elective surgeries were screened. Among them, 920 patients were aged 18 years or older. After excluding 130 patients based on the exclusion criteria, 790 patients were included in the final analysis. The excluded patients comprised 30 with chronic obstructive pulmonary disease, 6 with history of head injury, 53 with adverse reactions to analgesics, and 41 with psychiatric disorders (Fig.1).

Changes in postoperative VAS scores were classified into three clusters using k-means clustering. The elbow method was used to determine the number of clusters (Fig.2a), and the three clusters were visualized in two dimensions (Fig.2b) using a principal component analysis.

As seen in Fig.3; Table1, examining the VAS pattern changes at the five postoperative time points revealed a general decrease in pain intensity over time (p < 0.001). After examining the pain changes in each cluster (Table1), the 790 patients were divided into three clusters comprising 109, 313, and 368 patients. From clusters 1 to 3, progressively lower VAS scores were observed at the same time points. Cluster-wise analysis showed that Cluster 1 exhibited no significant change in severe pain at 0h, 6h, and 12h (p = 1.000), with a mean VAS of more than 7.4, gradually decreasing to 5.18 ± 1.94 by 36h postoperatively. Cluster 2 presented moderate pain with a VAS of 5.95 ± 1.44 immediately postoperative, decreasing over time to 3.05 ± 1.21 by 36h. Cluster 3 initially reported moderate pain with a VAS of 5.21 ± 1.49, which significantly decreased over time to 0.92 ± 0.83 by 36h postoperatively. Clusters 2 and 3 showed a decrease in pain over time, with Cluster 2 exhibiting a steady decline in VAS across the investigated time points, whereas Cluster 3 showed a rapid decrease until 12h, followed by a slow decline from 12 to 36h. Except for VAS comparisons between 0 and 6h, 0–12h, and 6–12h in Cluster 1 (p = 1.000) and 0–24h in Cluster 1 (p = 0.002), all other VAS comparisons within the same cluster at different times and between different clusters at the same time showed p < 0.001.

Upon reviewing the demographic information in Table2, we found that the proportion of male patients was higher than that of female patients. Clusters 1 and 2 had a slightly higher number of female patients, whereas Cluster 3 had 70 or more male patients. The average age in Cluster 1 was 40.26 years, which was higher than that in Cluster 3 at 35.62 years (p = 0.043). Regarding the ASA classification, most patients in all three clusters were classified as ASA 1; however, Cluster 1 had a lower proportion of ASA 1 patients than the other clusters, with a relative increase in ASA classifications 2 and 3. Regarding anesthesia type, inhalational anesthesia was used more frequently across all clusters.

Table3 provides insights into surgical duration, category, and PCA usage for each cluster. Cluster 1 had a longer surgical duration than clusters 2 and 3 (p < 0.001). Regarding postoperative PCA usage, 84.9% of the patients in the study utilized PCA after surgery. By cluster, 96.3% of patients in Cluster 1 used PCA postoperatively, which was significantly more frequent than in the other clusters (p = 0.001). For the PCA, fentanyl (1400 mcg for cancer surgeries, 700 mcg for non-cancer surgeries) and 150mg of ketorolac were mixed in normal saline to a total volume of 120ml for PCA, with a uniform setting of continuous dose at 1.0ml/hr, bolus dose of 1.0ml, and lock-out time of 10min. Regarding surgical categories, the proportion of surgeries differed significantly across clusters (p < 0.001). In Cluster 1, the number of cancer and orthognathic surgeries was similar, with other dentofacial surgeries being the least frequent. In Cluster 2, the number of orthognathic and other dentofacial surgeries was similar, with cancer surgery being the least frequent. Finally, in Cluster 3, other dentofacial surgery was the most frequent, followed by orthognathic surgery, with cancer surgery being the least frequent. Throughout the study period, the number of cancer surgeries in each cluster was nearly identical, with 45 cases in Cluster 1, 50 in Cluster 2, and 45 in Cluster 3.

Table4 offers additional detailed analysis based on surgical procedures. In reconstructive surgery, procedures such as plate removal, flap, reconstruction, Caldwell–Luc operation, bone graft, torus removal, alveoloplasty, re-fixation, and sequestrectomy are included. Dental surgery encompasses surgical extraction with and without cyst enucleation, as well as implant procedures. Excisional surgery refers to procedures that include excision of tissues surrounding the teeth or excisional biopsy of soft tissues in the oral and facial area. Notably, cancer-related surgeries are excluded from this category. Additionally, temporomandibular joint (TMJ) surgery includes coronoidectomy, gap arthroplasty, and condylectomy.

Accurate assessment and management of pain after surgery are crucial for reducing patient morbidity and mortality⁸. Previous studies have analyzed postoperative pain score trends and opioid use trajectories among hospitalized pain management patients using k-means clustering and have shown promise for segmenting subgroups, which could subsequently aid in personalized pain management^11,12,13. These findings align with those of our research, which focused on the necessity of sophisticated pain management.

Table4 provides a detailed analysis of surgeries by type, revealing significant differences even within the same category. The proportion of cancer surgeries was notably higher in cluster 1, accounting for 41.3% of the cases. The age of patients undergoing cancer surgery was significantly higher than that of those undergoing other types of surgery (56.3 ± 14.3 years for cancer surgery, 24.4 ± 6.27 years for orthognathic surgery, and 40.9 ± 15.7 years for other dentofacial surgery; p < 0.001). This suggests that the statistically significant differences in age and surgical duration observed in Cluster 1 could be compared to Cluster 3, in which only 12.2% of surgeries were cancer-related, predominantly within 46.7% of simple mass excisions. Although the cancer surgeries included in the study were almost evenly distributed across the three clusters, examination of the detailed surgical procedures revealed distinct differences. The cancer surgeries in Cluster 1 showed a higher frequency of mass excision combined with neck dissection and reconstruction, whereas in Cluster 3, the surgeries were predominantly simple mass excisions. Thus, while both are classified as cancer surgeries, patients undergoing procedures involving neck dissection and reconstruction are approximately 50% more likely to fall into Cluster 1. Conversely, those who underwent only mass excision had a 55.3% likelihood of following the pain trajectory of Cluster 3.

In orthognathic surgeries across all clusters, the most common procedures involved osteotomies of the maxilla and mandible, followed by concurrently performed genioplasty. Specifically, the pain trajectory for surgeries involving osteotomies of the maxilla and mandible was most frequently associated with Cluster 2 (49.3% of cases) and least frequently with Cluster 1 (13.7% of cases). When only osteotomy of the maxilla or mandible was performed, the pain patterns were most likely to align with Cluster 3. Conversely, although one might expect the pain to worsen with the addition of genioplasty owing to the increased surgical field, the frequency of being categorized in Cluster 3 was higher than that in Cluster 2. However, from another perspective, among patients undergoing osteotomies of the maxilla and mandible, 13.7% fell into Cluster 1 for postoperative pain, whereas for those who also underwent genioplasty, the rate was 17.2%, which was approximately 3.5% higher. Further analysis and examination of surgical differences are necessary for future research to understand the reasons for this finding.

Among the other dentofacial surgeries, reconstructive surgery was the most common across all clusters, followed by dental surgery. Overall, 53.2% of the other dentofacial surgeries followed the pain trajectory of Cluster 3. Except for TMJ surgery, which mostly follows the pain trajectory of Cluster 2 (although the number of cases is too small to make generalizations), all other types of other dentofacial surgery predominantly show the pain trajectory of Cluster 3.

Owing to the higher frequencies of cancer surgery and orthognathic surgery compared to other dentofacial surgery, the extensive surgical field and consequent increase in incision areas and tissue inflammation likely contributed to the moderate-to-severe postoperative pain experiences observed, particularly in Cluster 1^5,14,15,16. These findings suggest that not only the category of surgical procedure but also the specific surgical details and extent should be considered when managing postoperative pain. This study provides insights into the patterns of postoperative pain complaints based on the surgical details.

In addition to surgery type, the duration of surgery is a critical factor in shaping postoperative pain patterns. Longer surgeries often involve more extensive tissue manipulation, prolonged anesthesia, and heightened inflammatory responses, all of which can contribute to greater pain intensity. As observed in this study, Cluster 1, characterized by the highest pain levels, also had significantly longer surgery durations compared to other clusters (p < 0.001). This finding aligns with the higher proportion of cancer surgeries in this cluster, as these procedures typically involve a broader surgical scope. However, the isolated impact of surgery duration on pain could not be fully analyzed in this study due to the presence of multiple confounders. Future research should aim to control for such factors to provide a more detailed understanding of the interplay between surgery duration and postoperative pain.

Additionally, this study includes its focus on hospitalized patients, which does not account for the growing number of outpatient surgeries. Previous studies on outpatient versus inpatient surgeries for procedures such as hysterectomies, thyroidectomies, and spine surgeries have shown controversial results; however, no results for dental surgeries indicate that further research is necessary to understand outpatient dental surgeries^17,18,19.

Another limitation lies in the categorization of surgeries into three broad groups. While this grouping was intended to capture broad trends in postoperative pain trajectories, it may introduce variability due to the heterogeneity of procedures within each category. For example, cancer surgeries include both complex reconstructions with neck dissection and simpler mass excisions, which likely result in distinct pain patterns. Similarly, orthognathic surgeries encompass single-jaw osteotomies as well as combined procedures with genioplasty, each with varying degrees of tissue manipulation and postoperative pain. This heterogeneity could influence the clarity of the results and may limit the generalizability of the findings. Future prospective studies should address these limitations by adopting more granular categorizations of surgeries to better capture the specific effects of surgical details on pain trajectories. For example, stratifying cancer surgeries into ‘reconstruction with neck dissection’ versus ‘mass excision only,’ or subdividing orthognathic surgeries into ‘single-jaw osteotomy’ versus ‘combined osteotomy with genioplasty’ could provide more nuanced insights.

Additionally, this study focused on pain intensity as measured by the VAS, without capturing detailed descriptors of pain types (e.g., sharp, dull, or throbbing). These characteristics, commonly reported by dental surgery patients, could provide valuable insights into pain patterns.

Moreover, the intermittent nature of VAS assessments, which are conducted only at specific postoperative times, may overlook instances of breakthrough pain and additional analgesic requirements. While most surgeries use PCA, the lack of continuous pain recording makes it challenging to assess the influence of analgesic interventions on the VAS. Additionally, the administration of additional analgesics in the general ward was undocumented in the previously collected data. Furthermore, the preoperative assessments did not account for chronic pain. Prior use of pain medication and preoperative pain levels are known predictors of poor postoperative pain control^8,20,21. Due to the retrospective nature of the study, there were missing data. Future prospective studies should consider these factors, as well as incorporating multidimensional pain assessment tools to evaluate both the quantity and quality of pain, for a more comprehensive understanding of postoperative pain management.

Despite these limitations, the study benefited from minimal follow-up loss due to hospitalization for > 36h postoperatively. Additionally, face-to-face assessments using the VAS allowed for reliable pain evaluation at predetermined times. Various types of surgeries have been evaluated, in contrast to other studies that primarily focused on specific types of surgeries, such as tooth extraction, and a small number of studies involving a variety of procedures have focused on the concepts of pain medication rather than detailed pain profiles^{22,23,24,25,26}. However, what sets our study apart is the classification of patients based on changing patterns of postoperative pain and the subsequent analysis of surgical frequencies within each cluster, providing a novel perspective on postoperative pain. Furthermore, we meticulously analyzed the detailed surgical names according to the extent of the procedure, recognizing that merely categorizing surgeries based on their categories may not fully capture their complexity and invasiveness. By delving into the detailed surgical names and analyzing the extent of surgery, we aimed to understand the varying degrees of postoperative pain. While it is challenging to assess surgical difficulty and invasiveness based solely on surgical categories, our approach to understanding the scope of surgery and conducting an analysis accordingly aimed to minimize errors. The advantage of this analytical approach lies in highlighting the importance of sophisticated pain management techniques. Understanding how the degree of postoperative pain varies even within the same surgical category is crucial. As observed in our findings, patients undergoing similar oral cancer surgeries exhibited different postoperative pain patterns, depending on whether they underwent simple mass excision or more complex procedures, such as reconstruction or neck dissection. Recognizing these differences in postoperative pain patterns can aid clinicians in predicting postoperative pain, preemptively controlling pain, and explaining pain to patients. Furthermore, our institution’s prominence in the field of oral and maxillofacial surgery, with the highest number of general anesthesia surgeries nationwide, ranging from simple surgical extractions to cancer surgeries, provides a unique opportunity to analyze postoperative pain profiles for various dental surgeries.

By addressing the limitations and controlling variables, future prospective studies in this field can offer valuable insights into postoperative pain management for patients undergoing oral and maxillofacial surgery and aid anesthesiologists and oral and maxillofacial surgeons in clinical decision-making processes.

This study was driven by the observation that the postoperative pain patterns of the recently increasing cancer surgeries and orthognathic surgeries differ significantly from those of other traditional dentofacial surgeries. Moreover, it was noted that even within the same category of cancer surgeries, patients reported varying pain patterns depending on the extent and scope of the procedure, highlighting substantial variability based on specific surgical details. To quantitatively investigate these differences, postoperative VAS changes were analyzed to identify clusters of patients with similar pain trajectories. Within these clusters, the distribution of various surgical categories and specific procedures was examined. The findings confirmed that not only do cancer and orthognathic surgeries exhibit distinct postoperative pain distributions compared to other dentofacial surgeries, but also that specific procedural variations within the same surgical category can significantly influence pain trajectories.

This study represents the first application of such an analytical approach in the context of dental surgery postoperative pain. The insights gained emphasize that postoperative pain management strategies should be tailored not only to the broad category of surgery but also to the specific procedural details. Recognizing that even within the same type of surgery, such as cancer surgeries, the inclusion of procedures like neck dissection or flap reconstruction can lead to different pain experiences, clinicians can adopt more nuanced and personalized approaches to pain control. This approach enables better preoperative patient education regarding expected pain trajectories and facilitates more predictive and individualized analgesic planning.

Furthermore, the analytical framework introduced here serves as a robust foundation for future research to refine pain management protocols and deepen the understanding of how procedural variations influence postoperative pain. This study provides a novel perspective on postoperative pain patterns and establishes a practical framework for improving patient care in dental surgery.

The study design was approved by the Institutional Review Board of Seoul National University Dental Hospital (IRB number: CRI07005), and all research was performed in accordance with relevant guidelines and the Declaration of Helsinki. The requirement for informed consent was waived by the IRB of Seoul National University Dental Hospital owing to the retrospective nature of the study using electronic medical records (EMRs).

Patients aged 18 years and older who underwent elective surgeries in the oral and maxillofacial surgery department under general anesthesia and were hospitalized for > 36h postoperatively were included in the study. Patients with chronic obstructive pulmonary disease, head injury, adverse reactions to analgesics or narcotics, history of drug misuse or addiction, severe metabolic disorders or infections, or psychiatric disorders were excluded from the study.

The anesthesiologists surveyed VAS scores at 0, 6, 12, 24, and 36h postoperatively. We also collected information on surgery name, diagnosis, sex, age, weight, height, American Society of Anesthesiologists (ASA) classification, anesthesia type, surgery time, and number of patients using patient-controlled analgesia (PCA). Postoperative pain scores were measured at 0h after discharge from the post-anesthesia care unit (PACU).

The collected data on surgery names were further analyzed and categorized into three groups: ‘Cancer surgery’, ‘Orthognathic surgery’, and ‘Other dentofacial surgery’. This detailed classification allowed us to group procedures based on their clinical characteristics and surgical objectives. For frequency analysis, these major categories were further subdivided based on specific procedure names extracted from the surgical records, ensuring detailed insights into the distribution of surgical types.

Given that the study utilized precollected data, handling missing data was crucial. Few patients with missing data on ASA classification and anesthesia type were included in clusters based on the proportions of pre-existing patients within each cluster. In cases where VAS measurements were missing for specific periods, linear interpolation was performed based on the VAS values from other time points.

This study retrospectively analyzed postoperative VAS data using a modern cluster-based method, specifically k-means clustering. The method involves plotting VAS scores at 0, 6, 12, 24, and 36h postoperatively to identify patterns of pain trajectory and grouping patients with similar patterns into clusters. To determine the optimal number of clusters, the elbow method was employed, as illustrated in Fig.2a. The graph derived from the elbow method identifies the point at which the rate of change in slope decreases significantly, allowing for the selection of the corresponding value as the number of clusters, ‘k’. Once ‘k’ was determined, patients were grouped into clusters, which were subsequently visualized in two dimensions using principal component analysis (Fig.2b). Each oval in the figure represents a cluster, and the clinical significance of the pain trajectory patterns within each cluster was assessed.

In this study, the clustering approach revealed that setting ‘k’=3 provided the most clinically diverse and meaningful distinctions in postoperative pain trajectories. This enabled a robust analysis of the varying pain patterns and their potential clinical implications.

Statistical analyses were performed using R software for Windows, version 4.4.0 (R Foundation for Statistical Computing, Vienna, Austria). The elbow method was used to determine the optimal number of clusters, and k-means clustering was performed accordingly to analyze patients with similar trends in VAS scores over time. The clusters were visualized in two dimensions using principal component analysis. After examining the VAS trajectories of each cluster, the final number of clusters was determined based on clinically significant interpretations.

Demographic information, visual analog scale (VAS) changes, PCA usage, surgery classifications, and frequency analyses were performed for each cluster. Continuous variables were expressed as “mean ± standard deviation,” while categorical data were presented as “frequencies (percentages, %).” The normality of continuous variables was confirmed using the Shapiro–Wilk test, followed by repeated measures analysis of variance (ANOVA) to ascertain the statistical significance of the mean differences. If necessary, pairwise t-tests were conducted for each time point to evaluate the differences in pain scores, and p-values were adjusted using the Bonferroni method. Cross-analysis was performed using the chi-square test for categorical data.

Author notes

1. Qurani Alifitriah Tartar and Kyung Nam Park contributed equally to this project as co-first authors.

2. Kwang-Suk Seo and Myong-Hwan Karm contributed equally to this project as co-corresponding authors.

Authors and Affiliations

1. Department of Dental Anesthesiology, School of Dentistry and Dental Research Institute, Seoul National University, 101 Daehak-ro, Jongno-gu, Seoul, 03080, Republic of Korea

   Qurani Alifitriah Tartar,Kyung Nam Park,Kwang-Suk Seo&Myong-Hwan Karm

Contributions

T.Q.: Data curation, formal analysis, investigation, methodology, validation, visualization, writing – original draft.P.K.: Data curation, formal analysis, investigation, methodology, validation, visualization, writing – original draft.S.K.S: Conceptualization, formal analysis, investigation, methodology, project administration, resources, supervision, validation, writing – review & editing.K.M.H.: Conceptualization, project administration, supervision, validation, writing – review & editing.All authors reviewed the manuscript.

Corresponding authors

Correspondence to Kwang-Suk Seo or Myong-Hwan Karm.

Competing interests

The authors declare no competing interests.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions