Are Dental Procedures Associated With Pyogenic Vertebral Osteomyelitis?

Abstract

Background

Dental procedures can lead to bacteremia and have been considered a potential risk factor for pyogenic vertebral osteomyelitis (PVO). However, data on the association between dental procedures and PVO are limited.

Questions/purposes

(1) After controlling for relevant confounding variables, are dental procedures associated with an increased risk of PVO? (2) Does antibiotic prophylaxis before dental procedures effectively decrease the risk of PVO?

Methods

A case-crossover study was conducted to investigate the association between dental procedures and PVO using a Japanese claims database. The advantage of this study design is that confounding factors that do not vary over time are automatically adjusted for, because cases act as their own controls. From April 2014 to September 2021, the database included 8414 patients who were hospitalized for PVO. Of these, 50% (4182 of 8414) were excluded because they had not undergone any dental procedures before the index date, a further 0.1% (10 of 8414) were excluded because they were younger than 18 years at the index date, and a further 7% (623 of 8414) were excluded because they did not have at least 20 weeks of continuous enrollment before the index date, leaving 43% (3599 of 8414) eligible for analysis here. The mean age was 77 ± 11 years, and 55% (1985 of 3599) were men. Sixty-five percent (2356 of 3599) of patients had a diagnosis of diabetes mellitus, and 42% (1519 of 3599) of patients had a diagnosis of osteoporosis. We compared the frequency of dental procedures between a 4-week hazard period before the admission date for PVO and two control periods, 9 to 12 weeks and 17 to 20 weeks before the admission date for PVO, within individuals. We calculated odds ratios and 95% confidence intervals using conditional logistic regression analysis.

Results

Comparing the hazard and matched control periods within individuals demonstrated that dental procedures were not associated with an increased risk of PVO (OR 0.81 [95% CI 0.72 to 0.92]; p < 0.001). Additional analysis stratified by antibiotic prophylaxis use showed that antibiotic prophylaxis was not associated with a lower OR of developing PVO after dental procedures (with antibiotic prophylaxis: OR 1.11 [95% CI 0.93 to 1.32]; p < 0.26, without antibiotic prophylaxis: OR 0.72 [95% CI 0.63 to 0.83]; p < 0.001). Our sensitivity analyses, in which the exposure assessment interval was extended from 4 to 8 or 12 weeks and exposure was stratified by whether the dental procedure was invasive, demonstrated results that were consistent with our main analysis.

Conclusion

Dental procedures were not associated with an increased risk of subsequent PVO in this case-crossover study. The effectiveness of antibiotic prophylaxis was not demonstrated in the additional analysis that categorized exposure according to the use of antibiotic prophylaxis. Our results suggest that the association between dental procedures and PVO may have been overestimated. Maintaining good oral hygiene may be important in preventing the development of PVO. The indications for antibiotic prophylaxis before dental procedures should be reconsidered in view of the potential risk of adverse drug reactions to antibiotic prophylaxis and the emergence of drug-resistant pathogens. Larger randomized controlled trials are needed to confirm these findings and assess the role of antibiotic prophylaxis.

Level of Evidence

Level III, therapeutic study.

Introduction

Pyogenic vertebral osteomyelitis (PVO) is a devastating disease; it requires prolonged antibiotic treatment, carries a substantial risk of recurrence, and 40% of patients in whom it develops have poor physical function despite treatment [2, 17, 25, 35]. The incidence of PVO is estimated to affect 5 to 11 per 100,000 individuals per year in developed countries, and has increased in recent decades [1, 5]. Possible reasons for this increase include the aging of society, higher prevalence of comorbidities, and improved diagnosis [1, 3, 5, 20].

PVO is commonly caused by the hematogenous transfer of bacteria from different sites in the body [22]. Although the main source of infection is speculated to be urinary tract infection, skin wounds, intra-abdominal organ infection, or oral cavity, little evidence regarding this question has appeared because all previous studies were case series [7, 11, 14, 21, 25, 34]. Dental procedures are reported to cause bacteremia [12], and a few case series and case reports of PVO after dental procedures have been reported [7, 21, 34]. However, these studies were limited by the absence of control patients, and recent studies have cast doubt on the infection risk after dental procedures and the necessity of routine antibiotic prophylaxis in unselected patients [26, 30, 32]. A large, well-designed study to confirm the association between dental procedures and the occurrence of PVO would be important. To this end, we considered that a large claims database represented a good setting in which to evaluate the relationship between dental procedures and PVO and the possible benefits of antibiotic prophylaxis because routinely collected claims data provide a substantial amount of data for rare events and reduce recall bias of dental procedures. For the present study, we opted to use a case-crossover design to control for time-invariant (value of the variable does not change across time) within-person confounding variables.

We therefore asked: (1) After controlling for relevant confounding variables, are dental procedures associated with an increased risk of PVO? (2) Does antibiotic prophylaxis before dental procedures effectively decrease the risk of PVO?

Patients and Methods

Data Source

To conduct this study, we used a commercially available administrative claims database provided by DeSC Healthcare Inc. As of 2022, the total number of individuals in the database was approximately 12 million. The database consists of individuals with a broad range of ages and social backgrounds (Supplemental Table 1; http://links.lww.com/CORR/B239), and includes information on an enrollee’s characteristics, diagnoses, procedures, and drugs. Individuals can be followed through a unique encrypted identifier, even if they visit or are hospitalized at multiple medical institutions. The need for individual informed consent is waived because all information is anonymized. This database has been used in clinical studies [23, 28]. We used it for the present study because routinely collected claims data provide a substantial amount of data to evaluate rare events and reduces recall bias of exposure.

Study Design

To control for time-invariant (value of the variable does not change across time) within-person confounding, we applied a case-crossover design, in which patients with a record of dental procedures and hospitalization for PVO served as their own controls [9, 15, 16]. This design compares exposure status in the period immediately before the outcome (hazard period) with exposure status in preceding periods (control periods) within the same individual (Fig. 1). The advantage of this design is that time-invariant confounding factors can be automatically adjusted, even if they are unknown or unmeasured, such as genetic predisposition [15]. The case-crossover design requires three assumptions: the exposure should be transient, the exposure trend is constant, and the outcome onset should be abrupt. Given that dental procedures are transient exposures with a stable trend over time and that PVO usually occurs suddenly, our study meets the assumptions required for a case-crossover design. The case-crossover design has been widely used to investigate the association of dental procedures with other infections, such as infective endocarditis and periprosthetic joint infection [4, 30-32].

Fig. 1.

This figure represents the hazard and control periods in the case-crossover study.

This study follows the Strengthening the Reporting of Observational Studies in Epidemiology guidelines [33].

Study Population and Outcome Definition

We identified all patients hospitalized with a diagnosis of PVO (Supplemental Table 2; http://links.lww.com/CORR/B240) between April 2014 and September 2021 who had a record of dental procedures during the same period. The index date was defined as the first admission date for PVO. Patients younger than 18 years were excluded from this study. All patients were required to have at least 20 weeks of continuous enrollment before the index date.

Between April 2014 and September 2021, the database followed 8414 patients hospitalized for PVO. Of these, 50% (4182 of 8414) were excluded because they had not received dental procedures before the index date, another 0.1% (10 of 8414) were excluded because they were younger than 18 years at the index date, and another 7% (623 of 8414) were excluded because they did not have at least 20 weeks of continuous enrollment before the index date, leaving 43% (3599 of 8414) eligible for the case-crossover study (Fig. 2).

Fig. 2.

This flowchart describes the selection of the study population.

Descriptive Data

The mean age was 77 ± 11 years, and 55% (1985 of 3599) were men (Table 1). Sixty-five percent (2356 of 3599) of patients had a diagnosis of diabetes mellitus, and 42% (1519 of 3599) had a diagnosis of osteoporosis. The most frequently used antibiotics after admission were first-generation cephalosporins (32% [1144 of 3599]), third-generation cephalosporins (25% [889 of 3599]), and carbapenem (15.7% [565 of 3599]).

Table 1.

Characteristics of patients with dental procedures and pyogenic vertebral osteomyelitis

Characteristic	Total (n = 3599)
Male sex	55 (1985)
Age in years	77 ± 11
Osteoporosis	42 (1519)
Diabetes mellitus	65 (2356)
Rheumatoid arthritis	12 (447)
Malignancy	48 (1739)
Hemodialysis	6 (226)
Year of onset
2014	0 (9)
2015	2 (63)
2016	7 (241)
2017	12 (425)
2018	16 (587)
2019	26 (948)
2020	28 (997)
2021	9 (329)

Exposure Definition

Exposure was defined as any dental procedure (Supplemental Table 3; http://links.lww.com/CORR/B241) [27, 32]. This exposure definition was selected after a discussion among the authors, one of whom was a dentist (KA). Dental claims data in Japan include information about general dental practice and oral and maxillofacial surgeries performed by dentists and hygienists. The sensitivity and specificity of these dental procedure data are 67% to 100% and 99% to 100%, respectively [24]. Exposure to dental procedures was measured during a hazard period of 1 to 4 weeks and two control periods, 9 to 12 weeks and 17 to 20 weeks, before the index date (Fig. 1). The hazard period reflects the finding that PVO is more likely to occur within 4 weeks after a dental procedure [7, 21, 34]. We set 4-week washout periods between the hazard and control periods to avoid carryover effects and potential autocorrelation [18]. Additionally, to investigate the effectiveness of antibiotic prophylaxis, we categorized exposure depending on the use of antibiotic prophylaxis [32]. We identified antibiotic prophylaxis prescribed to patients from 21 days before to the day of the dental procedure [32].

Ethical Approval

The ethical committee of Kyoto University Graduate School and Faculty of Medicine (number R3523) approved this study.

Statistical Analysis

We used descriptive statistics to show the characteristics of patients with PVO in this case-crossover study. We performed conditional logistic regression analysis to estimate the odds ratios and determined 95% confidence intervals for the risk of PVO associated with dental procedures in the hazard period compared with the two matched control periods within the same individuals. In our analysis, only discordant sets (where exposure status differed between the hazard and at least one control period) were used in the analysis, while concordant sets (where exposure status is the same in both the hazard and all control periods) did not contribute to the analysis [15, 16]. Consequently, people with constant exposure or no exposure throughout the study period were excluded from the analysis.

A series of sensitivity analyses were conducted to test the robustness of this study. In the first, we extended the exposure assessment window from 4 to 8 or 12 weeks [8, 30]. When the exposure assessment window was 8 weeks, the hazard period was 1 to 8 weeks and the two control periods were 17 to 24 weeks and 33 to 40 weeks before the index date. When the exposure assessment window was 12 weeks, the hazard period was 1 to 12 weeks and the two control periods were 25 to 36 weeks and 49 to 60 weeks before the index date. In the second sensitivity analysis, we defined exposure according to whether the dental procedure was invasive (for example, tooth extraction) or noninvasive (for example, periodontal treatment) (Supplemental Table 3; http://links.lww.com/CORR/B241) [32].

All statistical analyses were conducted using SAS version 9.4 (SAS Institute) and R 4.2.1 (R Foundation).

Results

Dental Procedures and Odds of PVO

A comparison of the hazard and control periods within individuals showed that dental procedures were not associated with an increased risk of PVO (OR 0.81 [95% CI 0.72 to 0.92]; p < 0.001) (Table 2). In the main analysis, although 18% (663 of 3599) of patients underwent at least one dental procedure during the hazard period, 31% (1132 of 3599) of patients had dental procedures in the control periods (Table 2). The distribution of concordant and discordant matched pairs for the presence or absence of dental procedures during the case and control periods is shown (Supplemental Table 4; http://links.lww.com/CORR/B242).

Table 2.

Association of dental procedures with pyogenic vertebral osteomyelitis

	Hazard period	Control periodsa	ORb (95% CI)	p value
Main analysis
Exposure assessment period, 4 weeksc (n = 3599)	18 (663)	31 (1132)	0.81 (0.72-0.92)	< 0.001
Secondary analysisd
With antibiotic prophylaxis (n = 2737)	9 (250)	15 (410)	1.11 (0.93-1.32)	0.26
Without antibiotic prophylaxis (n = 3264)	14 (448)	27 (889)	0.72 (0.63-0.83)	< 0.001
Sensitivity analysis
Exposure assessment period, 8 weekse (n = 3222)	28 (909)	44 (1412)	0.81 (0.73-0.91)	< 0.001
Exposure assessment period, 12 weeksf (n = 2826)	34 (973)	53 (1492)	0.74 (0.66-0.84)	< 0.001
Invasive dental proceduresg (n = 1856)	4 (78)	8 (153)	0.94 (0.71-1.25)	0.67
Noninvasive dental proceduresg (n = 3407)	19 (634)	32 (1099)	0.80 (0.71-0.90)	< 0.001

Data are presented as % (n) unless noted otherwise.

^aPatients who underwent a dental procedure in at least one control period.

^bOR of exposure to dental procedures in the hazard period compared with the two control periods within the same individuals.

^cHazard period was 1 to 4 weeks and two control periods were 9 to 12 weeks and 17 to 20 weeks before the index date.

^dCategorizing the exposure depending on use of antibiotic prophylaxis.

^eHazard period was 1 to 8 weeks and two control periods were 17 to 24 weeks and 33 to 40 weeks before the index date.

^fHazard period was 1 to 12 weeks and two control periods were 25 to 36 weeks and 49 to 60 weeks before the index date.

^gClassifying the exposure according to whether the dental procedure was invasive or nonivasive.

Does Antibiotic Prophylaxis Before Dental Procedures Reduce Risk of PVO?

When exposure was categorized according to the use of antibiotic prophylaxis, the use of antibiotic prophylaxis was not associated with a lower OR of developing PVO after dental procedures (with antibiotic prophylaxis: OR 1.11 [95% CI 0.93 to 1.32]; p < 0.26, without antibiotic prophylaxis: OR 0.72 [95% CI 0.63 to 0.83]; p < 0.001).

The results of a series of sensitivity analyses, in which extending the exposure assessment window from 4 to 8 or 12 weeks and stratifying exposures according to whether the dental procedures were invasive, were consistent with our main analysis (Table 2).

Discussion

PVO can be devastating because it is associated with a high risk of recurrence, and approximately half of patients who experience it have impaired physical function despite treatment. To date, it remains unclear whether dental procedures can lead to the development of PVO. This case-crossover study revealed that dental procedures did not increase the risk of PVO. Our additional analysis, in which categorization of exposure depended on whether antibiotic prophylaxis was used, did not show that antibiotic prophylaxis use was associated with a lower odds of developing PVO after dental procedures.

Limitations

First, diagnoses recorded in claims databases are less accurate than those in planned prospective studies. Although it is unclear whether blood culturing and MRI were undertaken to confirm the diagnosis of PVO, we only included patients hospitalized for PVO, which could have minimized potential misclassification bias. Second, no information was available on microorganism type, diagnostic method, or onset of symptoms. In particular, measurement bias caused by diagnostic delay might lead to a reverse association of dental procedures and PVO. However, our sensitivity analyses, in which we extended the exposure assessment window from 4 to 8 or 12 weeks, showed consistent results. Thus, we do not consider that this bias much affected our results. Third, although we had access to prescription records, it is impossible to know whether patients took prescribed medicines. This could have led to an underestimation of the effect of antibiotic prophylaxis. Moreover, we speculated that selection bias arising from the greater likelihood that higher-risk patients such as older people would receive antibiotic prophylaxis could influence the effect of antibiotic prophylaxis negatively. Thus, larger randomized controlled trials are warranted to confirm these results and assess the role of antibiotic prophylaxis. Lastly, although the case-crossover design can ignore confounding factors that do not change over time, this study design could not estimate the magnitude of the effect of dental procedures on the risk of PVO.

Dental Procedures and Odds of PVO

Several studies have investigated the association between dental procedures and infectious disease [4, 29-31]. Although a case-crossover study from the United States showed an association between invasive dental procedures and infective endocarditis for those at high risk [31], a case-only study from Taiwan showed no association of dental procedures with infective endocarditis [4]. Another case-crossover study also found no association between dental procedures and periprosthetic joint infection [30]. Finally, a second case-control study found that dental procedures had a negative correlation with infection after spinal instrumentation surgery [29]. Our main results are consistent with those of these previous studies and provided meaningful insight into the etiology of PVO. Unexpectedly, we also found that dental procedures were associated with a decreased risk of PVO. One explanation is that patients who regularly receive periodontal treatment to maintain oral hygiene are protected from PVO. In general, daily oral activities such as tooth brushing have the same bacteremia risk as dental procedures in the presence of poor oral hygiene [13]. A previous report showed that invasive dental procedures such as dental extractions have a stronger risk for subsequent infective endocarditis than noninvasive procedures, such as periodontal treatment [31]. Our sensitivity analyses, in which exposure was categorized according to whether the dental procedures were invasive, were consistent with this previous study [31]. Accordingly, we speculate that dental procedures, particularly periodontal treatment, could improve patients’ oral hygiene and mitigate the risk of PVO, despite the bacteremia risk. A second explanation for this finding is that the diagnostic delay was related to an inverse association between dental procedures and hospitalization for PVO. The diagnosis of PVO is sometimes difficult and delayed [10]. This interval between the onset of symptoms and diagnosis and treatment could decrease the frequency of dental procedures before hospitalization for PVO. However, our sensitivity analysis, which extended the exposure assessment window from 4 to 8 or 12 weeks, showed consistent results and is likely long enough to have alleviated the influence of diagnostic delay, which is reported to be a median (interquartile range) of 19 days (9 to 45) [19]. Thus, we do not consider that diagnostic delay much affected our results. Based on our findings, surgeons might consider that there is no need to hesitate in performing dental procedures owing to the risk of PVO.

Does Antibiotic Prophylaxis Before Dental Procedures Reduce the Risk of PVO?

Our study did not demonstrate the effectiveness of antibiotic prophylaxis. Given our main result that dental procedures were not associated with an increased risk of PVO, it is reasonable to expect that the effectiveness of antibiotic prophylaxis in preventing PVO would be minimal. Consistent with our findings, recent studies found no effectiveness of antibiotic prophylaxis and raised questions about whether the use of antibiotic prophylaxis is worthwhile [4, 29, 30]. There is no doubt that unnecessary antibiotic medication carries risks of allergy, adverse effects, and the emergence of drug-resistant pathogens [6]. Based on our findings, surgeons should generally avoid prescribing antibiotic prophylaxis before dental procedures, albeit with due regard to clinical judgment. Our database was insufficiently detailed to allow us to explore whether specific risk factors such as prior infection after spinal instrumentation might justify the use of antibiotic prophylaxis.

Conclusion

This large claims database study using a case-crossover design suggests that dental procedures are not associated with an increased risk of subsequent PVO. An additional analysis that categorized exposure according to the use of antibiotic prophylaxis did not show the effectiveness of antibiotic prophylaxis. Our results suggest that the association between dental procedures and PVO may have been overestimated and that maintaining good oral hygiene may be important in preventing the development of PVO. The indications for antibiotic prophylaxis before dental procedures should be re-evaluated, considering the potential risk of adverse drug reactions to antibiotic prophylaxis and the emergence of drug-resistant pathogens. Further randomized controlled trials on a larger scale are needed to confirm these results and evaluate the role of antibiotic prophylaxis.

Supplementary Material

abjs-482-716-s001.docx ^{(21.6KB, docx)}

abjs-482-716-s002.docx ^{(22.3KB, docx)}

abjs-482-716-s003.docx ^{(25.5KB, docx)}

abjs-482-716-s004.docx ^{(22KB, docx)}

abjs-482-716-s005.docx ^{(51.9KB, docx)}