SEARCH

SEARCH BY CITATION

Keywords:

  • horse;
  • radiography;
  • dentistry;
  • cheek tooth;
  • apical infection

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgement
  9. Manufacturer's address
  10. References

Reasons for performing study: Radiography is commonly used for the diagnosis of equine cheek teeth (CT) infection but, to our knowledge, no study to date has evaluated the relative values of individual specific radiographic signs when making a diagnosis.

Objectives: To investigate the sensitivity and specificity of individual radiographic signs identified from the literature for the diagnosis of CT apical infection using a retrospective case-control study.

Methods: Cropped radiographs taken using computed radiography of 41 apically infected CT and 41 control CT were independently blindly evaluated by 3 clinicians for the presence of 12 predetermined radiographic signs associated with CT apical infection. A final diagnosis of either noninfected or infected was made. Sensitivity and specificity were calculated for the presence or absence of each radiographic sign for each clinician. Uni- and multivariable conditional logistic regression were used to determine strength of association of the 12 radiographic signs with apical infection.

Results: Median sensitivity and specificity for the diagnosis of CT apical infection were 76 and 90%, respectively. Periapical sclerosis, clubbing of one or 2 roots, degree of clubbing and periapical halo formation had the highest sensitivities (73–90%), with moderate specificity (61–63%). Multivariable conditional logistic regression revealed that severity of periapical sclerosis and extensive periapical halo were strongly associated with CT apical infection.

Conclusions: The presence of periapical sclerosis and formation of a periapical halo were strongly associated with CT apical infection. Computed radiography appears to have a higher sensitivity but similar specificity to previously published results using film radiography to detect CT apical infection.

Potential relevance: These findings may aid practitioners when interpreting radiographs of equine CT as to the relative significance of their findings.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgement
  9. Manufacturer's address
  10. References

Radiography is currently the most commonly employed diagnostic imaging technique for investigating apical infection of cheek teeth (CT) in horses. The complex 3-dimensional structure of the head means that interpretation of radiographs of this region can be difficult in some cases and diagnosis of equine apical infection is reported to be made with confidence in only 50–57% of cases, especially in the more caudally positioned maxillary CT where secondary dental sinusitis is common (Wyn-Jones 1985; Gibbs and Lane 1987; Tremaine and Dixon 2001). Two studies that have investigated the accuracy of conventional film radiography for diagnosis of equine dental disorders found sensitivities of 52–69% and specificities of 70–95% (Weller et al. 2001; Barakzai 2005). To date, a single study has investigated the role of computed radiography in the diagnosis of equine CT apical infection (Casey et al. 2009) but, to the best of the author's knowledge, no study to date has investigated the sensitivity and specificity of specific radiographic signs for diagnosis of equine CT apical infection.

Reported signs in the early stages of apical infection include widening of the periodontal space and thinning of the lamina dura (LD) (Baker 1971; Gibbs and Lane 1987; Fig 1). More advanced signs of CT apical infection may occur if infection has been present for many weeks and the apex may develop lytic changes, which manifest as periapical radiolucent halos (Fig 2) or a rounded or clubbed appearance of the tooth roots (Baker 1971; Wyn-Jones 1985; Gibbs and Lane 1987; Fig 3). Increased radiopacity or sclerosis of the bone supporting the CT is often also noted in more advanced cases (Baker 1971; Gibbs and Lane 1987; Fig 2).

image

Figure 1. A control (noninfected) tooth with clearly visible lamina dura around the rostral root (white arrows), though the LD cannot be fully discerned around the caudal root (black arrows).

Download figure to PowerPoint

image

Figure 2. An apically infected cheek tooth with a moderate periapical halo (black arrows) with widening of the periodontal ligament, mild root clubbing and moderate periapical sclerosis. The lamina dura is not distinguishable in any regions of this tooth, but it can be noted in the apical region of the more caudal tooth (white arrow).

Download figure to PowerPoint

image

Figure 3. An apically infected cheek tooth with moderate clubbing of both rostral and caudal roots, widening of the periodontal ligament, mild periapical sclerosis and fragmentation of the rostral root.

Download figure to PowerPoint

Other signs of CT apical infection include abnormal apical deposition of cementum, which appears as diffuse areas of mineral density deposition around the reserve crown (Baker 1971; Gibbs and Lane 1987), and dystrophic calcification of the nasal conchae in chronic maxillary CT infections (Gibbs and Lane 1987; Tremaine and Dixon 2001). This reserve crown cementosis in the apical region of teeth may also be associated with increasing age but in aged horses is usually more regular in appearance, affects multiple (if not all) cheek teeth in the row and is associated with a significant proportion of the reserve crown, not just the apical region (P. Dixon, personal communication; Fig 4). Cementomaformation has also been described and appears as discrete rounded mineral density radiopacities adjacent to affected roots (Wyn-Jones 1985; Dixon and Dacre 2005; Fig 5). Dental dysplasia (Fig 6) or displacement may also lead to CT apical infection usually due to deep periodontal disease leading to spread of infection to the apex (Dacre et al. 2008a,b).

image

Figure 4. An apically infected cheek tooth with generalised lucency of the reserve crown and cementoma formation (black arrows). The lamina dura can be visualised at the rostral border of this tooth (white arrows) but not elsewhere and is also not visible around the adjacent (normal) teeth due to superimposition of other radio-opacities.

Download figure to PowerPoint

image

Figure 5. A noninfected cheek tooth, which is dysplastic.

Download figure to PowerPoint

image

Figure 6. Median sensitivity and specificity (with 95% confidence intervals) based on the presence or absence of each radiographic sign. The numbers of radiographic lesions (1–12) correspond to those in Table 1.

Download figure to PowerPoint

The aim of this study was to investigate the sensitivity and specificity of specific radiographic signs for detection of CT apical infection in the horse. An additional aim was to compare our results using computed radiography to those of previously published studies that have utilised conventional film radiography for diagnosis of CT infections (Weller et al. 2001; Barakzai 2005). The results of this study may give veterinary practitioners guidelines as to the reliability of their specific radiographic findings in individual cases of suspected CT infection.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgement
  9. Manufacturer's address
  10. References

Case details

Case records for horses admitted to the Dick Veterinary Equine Hospital, Easter Bush Veterinary Centre for investigation and treatment of CT apical infection between November 2005 and July 2007 were analysed retrospectively. A case was defined as an infected CT. Inclusion criteria for cases were: 1) a full clinical history and documentation of examination findings, including detailed oral examination using a dental mirror or oral endoscope; 2) a latero30°dorsal-lateroventral or latero 45° ventro-laterodorsal oblique digital radiograph of the affected CT row; 3) gross or histological confirmation of apical infection of the extracted tooth. The dental age (age in years relative to time of intraoral eruption) of each case tooth CT was calculated according to the horse's age and standard eruption times (Muyelle et al. 1997).

Control CT were selected from lateral oblique radiographs of horses presented for reasons other than dental disease (such as skull trauma remote to the CT, headshaking or epistaxis) or the unaffected contralateral CT row of a horse with apical infection that had: 1) no clinical signs of CT apical infection (unilateral nasal discharge, cuteanous draining tract or facial swelling); and 2) a normal occlusal surface (free of dentinal defects) on oral examination. Apically infected CT were matched to control CT by dental age and by position within the cheek tooth row. If a control CT of the exact same age was not available, a CT with a dental age one year older or younger than the case tooth was selected instead.

Radiography

Radiographs of case and control CT were cropped from the original latero30°dorsal–lateroventral oblique/latero45°ventro-laterodorsal oblique DICOM image, to show only the affected tooth with extended borders i.e. minimal amounts of the neighbouring teeth and either the ventral border of the mandible, the apical portion of the maxillary bone or a small amount of the paranasal sinuses (Figs 1–6).

Each image underwent blinded evaluation by 3 equine surgeons from the same institution (one RCVS and ECVS diploma holder and 2 RCVS certificate holders) experienced in reading dental radiographs. Data sheets evaluating 12 specific radiographic signs of CT apical infection, some with sub-categories (Table 1) were compiled for each case and control tooth.

Table 1. Table showing the 12 specific radiographic signs and their sub-categories assessed for each cheek tooth
Radiographic signPresence of radiographic sign and sub-categories
 1. Periapical halo formationNoYes
MildModerateExtensive
 2. Periapical sclerosisNoYes
MildModerateExtensive
 3. Clubbing of root/apicesNoOne rootTwo roots
 4. Severity of clubbingNoMildModerateExtensive
 5. Loss of lamina dura (LD)NoYes
 6. Widening of periodontal ligamentNoYes
 7. Displaced toothNoYes
 8. Reserve crown fragmentationNoYes
 9. Generalised radiolucency of the reserve crownNoYes
10. Cementum deposition on reserve crown/apicesNoYes
Age relatedPathological
11. Cementoma formationNoYes
MildModerateExtensive
12. Dysplastic toothNoYes

Statistical analyses

All statistical analyses were performed using STATA 10.1 statistical software (Stata Statistical Software: Release 10.1, 2005)1.

Median sensitivity and specificity for the diagnosis of CT apical infection and the presence or absence of the 12 radiographic signs were calculated. Fisher's exact 95% confidence intervals (95% CI) were calculated for the median value of sensitivity and specificity (Abramson 2004).

Cohen's kappa analysis was used to detect interclinician agreement. A weighted kappa analysis was used for radiographic signs measured on an ordinal scale (Landis and Koch 1977). Differences in detection of periapical infection between rostral (Triadan 06–08) and caudal (Triadan 09–11) maxillary CT, and maxillary and mandibular CT were determined using a Chi-squared test.

Diagnostic odds ratios (DOR; Glas et al. 2003) were calculated to determine the association between periapical infection and each of the 12 radiographic signs using uni- and multivariable conditional logistic regression. A maximum model containing all explanatory variables (i.e. all radiographic signs) was created. Explanatory variables were subsequently removed one at a time and were only retained if the likelihood ratio test P value was <0.05. Variables with P≥0.05 were added back into the final model one at a time to determine their effect on the odds ratios in the final model. If addition of the variable altered the odds ratio by >20% (Dohoo et al. 2003), this variable was considered to be a confounder. In addition, biologically plausible interactions between variables in the final model were identified and interaction terms were created and assessed. There were no interaction terms included in the final model. A multivariable model was generated using the entire dataset and also for each of the 3 clinicians individually.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgement
  9. Manufacturer's address
  10. References

There were 38 horses that contributed 41 infected CT to the study (3 horses contributed 2 CT each); 30 horses contributed 41 control CT (one horse contributed 3 CT and 4 horses contributed 2 CT). The distribution of CT position and dental age are shown in Table 2.

Table 2. Distribution of dental age and Triadan tooth number in case and control groups
 CasesControls
Total teeth4141
Total horses3836
Median dental age (years)7.47.5
Range (years)1–221–23
Triadan tooth numberPositionMaxillaryMandibularPositionMaxillaryMandibular
06210621
07630763
08720872
0914109141
10311031
11101110

Kappa values indicating the degree of agreement between each clinician are shown in Table 3. A kappa value <0.2 implies slight agreement, 0.2–0.4 fair agreement, 0.4–0.6 moderate agreement, 0.6–0.8 substantial agreement and >0.8 almost perfect agreement (Dohoo et al. 2003).

Table 3. Kappa values (s.e.) for overall diagnosis of CT apical infection and between clinicians for overall diagnosis of CT apical infection and for each individual radiographic sign
Radiographic signInter-clinician comparison
Clinician 1+2Clinician 1+3Clinician 2+3
Kappa (s.e.)Kappa (s.e.)Kappa (s.e.)
  • *

    Indicates a weighted kappa was performed.

Diagnosis of CT apical infection0.56 (0.09)0.34 (0.09)0.53 (0.09)
 1. *Periapical halo formation0.59 (0.08)0.26 (0.07)0.17 (0.07)
 2. *Periapical sclerosis0.50 (0.08)0.28 (0.07)0.29 (0.08)
 3. *Clubbing (1 root/2 roots)0.53 (0.08)0.03 (0.08)0.18 (0.08)
 4. *Clubbing (severity)0.60 (0.08)0.04 (0.07)0.25 (0.08)
 5. *Loss of lamina dura0.38 (0.08)0.09 (0.08)0.25 (0.08)
 6. *Widening of the periodontal ligament0.33 (0.08)0.04 (0.08)0.22 (0.08)
 7. Displaced tooth0.00 (0.00)0.26 (0.23)0.00 (0.00)
 8. Reserve crown fragmentation0.52 (0.13)0.22 (0.14)0.09 (0.12)
 9. Reserve crown lucency0.46 (0.18)0.10 (0.08)0.08 (0.07)
10. *Reserve crown cementum0.51 (0.09)0.10 (0.08)0.28 (0.09)
11. *Cementoma formation0.58 (0.07)0.07 (0.07)0.16 (0.08)
12. Dysplastic tooth0.79 (0.15)0.23 (0.16)0.23 (0.16)

The sensitivity and specificity of radiography for the overall diagnosis of CT apical infection were 73 and 93% for Clinician 1, 76 and 90% for Clinician 2 and 83 and 68% for Clinician 3. Median sensitivity and specificity for the diagnosis of CT apical infection as well as for each of the 12 radiographic signs for the 3 clinicians are shown in Table 4. Median sensitivity and specificity for each of the 12 radiographic signs are shown in Figure 6.

Table 4. Median sensitivity (SENS) and specificity (SPEC for the 3 clinicians in the overall diagnosis of apical CT infection and for each radiographic sign (presence or absence)
Radiographic signClinician median
SENS (95% CI)SPEC (95% CI)
CT apical infection - overall diagnosis76 (60–88)90 (77–97)
 1. Periapical halo formation78 (52–82)61 (45–76)
 2. Periapical sclerosis90 (67–97)61 (45–86)
 3. Clubbing (1 root/2 roots)73 (57–86)63 (47–78)
 4. Clubbing (degree)73 (57–86)61 (45–76)
 5. Loss of lamina dura90 (77–97)34 (20–51)
 6. Periodontal ligament widening68 (52–82)73 (57–86)
 7. Displaced tooth2 (1–13)100 (92–100)
 8. Reserve crown fragmentation22 (11–38)90 (77–97)
 9. Reserve crown lucency12 (4–26)100 (91–100)
10. Reserve crown cementum39 (24–56)80 (65–91)
11. Cementoma formation34 (20–51)88 (74–96)
12. Dysplastic tooth7 (2–20)95 (84–99)

Results of the univariable conditional logistic regression for each radiographic sign and its association with apical infection are shown in Table 5. Due to small numbers in their extensive categories, moderate and extensive periapical sclerosis and periapical halo formation were re-categorised as moderate/extensive. These results show all variables to be significantly associated with CT apical infection with the exception of CT displacement (P = 0.17) and CT dysplasia (P = 0.80). Cementosis of the reserve crown was significantly associated with CT apical infection (P<0.001), although when subcategorised this was found only to be pathological cementosis (P<0.001) rather than age-related cementosis (P = 0.49).

Table 5. Results of the univariable conditional logistic regression analysis
VariableUnivariable conditional logistic regression
VariableDORCIPOverall P
  1. DOR = diagnostic odds ratio; 95% CI = 95% confidence interval; P values in bold are based on likelihood ratio tests. All reported P values that correspond to individual levels of categorical variables were based on Wald tests.

ClinicianClinician 11.0 1.00
Clinician 21.00.6–1.71.0
Clinician 31.00.6–1.71.0
Periapical halo formationNone1.0 <0.001
Mild3.61.8–7.40.001
Moderate/Extensive11.75.3–25.8<0.001
Periapical sclerosisNone1.0 <0.001
Mild5.32.5–11.3<0.001
Moderate/Extensive90.024.1–335.8<0.001
Clubbing 1/2 rootsNone1.0 <0.001
1root2.61.3–5.00.005
2 roots8.23.9–17.5<0.001
Clubbing (degree)None1.0 <0.001
Mild1.50.8–2.80.263
Moderate8.73.5–21.2<0.001
Extensive24.26.2–94.1<0.001
Loss of lamina duraNone1.0 <0.001
Yes/Unsure4.62.3–9.3<0.0001
Periodontal ligament wideningNone1.0 <0.001
Yes/Unsure3.92.3–6.9<0.0001
Displaced toothNo1.0 0.17
Yes4.00.4–35.80.17
Reserve crown fragmentationNo1.0 0.009
Yes2.91.3–6.80.01
Radiolucency of reserve crownNo1.0 0.003
Yes3.11.4–6.60.005
Reserve crown cementosisNo1.0 <0.001
Age-related0.70.3–1.80.49
Pathological5.62.6–12.1<0.001
Cementoma formationNone1.0 <0.001
Mild4.92.0–12.20.001
Moderate3.50.8–16.30.01
Extensive6.91.3–36.10.02
Dysplastic toothNo1.0 0.80
Yes1.10.4–3.00.8

Results of the final multivariable model are presented in Table 6. In the final model, mild (DOR 4.3, 95%CI 1.8–9.6), moderate/extensive periapical sclerosis (DOR 72.0, 95%CI 17.6–295.6) and mild (DOR 2.6, 95%CI 1.0–6.3) and moderate/extensive periapical halo formation (DOR 4.4, 95%CI 1.5–12.9) were significantly associated with a diagnosis of periapical infection. Clinician was included in the model to account for repeated observations and differences in radiographic interpretations between clinicians. In all multivariable models generated for each of the 3 clinicians, the presence of periapical sclerosis was a consistent risk factor for periapical infection. Extensive periapical halo formation was retained in the model for Clinician 3.

Table 6. Results of the multivariable conditional logistic regression analysis demonstrating the association between periapical infection and radiographic signs
VariableMultivariable conditional logistic regression
VariableDORCIPOverall P
  1. DOR = diagnostic odds ratio; 95% CI = 95% confidence interval. P values in bold are based on likelihood ratio tests. All reported P values that correspond to individual levels of categorical variables were based on Wald tests.

ClinicianClinician 11 0.01
Clinician 20.60.3–1.40.24
Clinician 30.40.2–0.90.03
SclerosisNone1 <0.001
Mild4.11.8–9.60.001
Moderate/extensive72.017.6–295.6<0.0001
Periapical halo formationNone1 0.02
Mild2.61.0–6.30.039
Moderate/extensive4.41.5–12.90.008

There was no significant difference in the detection of apical infection between rostral (06–08) and caudal (09–11) maxillary CT for any clinician (Clinician 1 P = 0.36, Clinician 2 P = 0.8, Clinician 3 P = 0.29). The acuracy of specific radiographic signs for detection of infection in rostral compared to caudal maxillary CT could not be analysed statistically due to relatively small numbers. Descriptively, moderate to extensive sclerosis was present in 9/15 (60%) of rostral maxillary CT compared to 8/15 (53%) of caudal maxillary CT; and moderate to extensive periapical halo formation was present in 7/15 (47%) of rostral maxillary CT compared to 6/15 (40%) of caudal maxillary CT. There was no significant difference in the detection of apical infection between maxillary and mandibular CT for any clinician (Clinician 1 P = 0.13, Clinician 2 P = 0.44, Clinician 3 P = 0.47).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgement
  9. Manufacturer's address
  10. References

Based on the optimal balance between high sensitivity and specificity as well as consistent interclinician agreement, the presence of 4 specific radiographic signs (periapical sclerosis, periapical halo formation, clubbing of one or 2 roots and severity of clubbing) were identified from our population of horses as being most consistent for the diagnosis of CT apical infection. However, when matching of cases and controls was taken into account and after adjusting for the confounding effects of other radiographic signs in the multivariable model, only the presence of periapical sclerosis and periapical halo formation were strongly associated with CT apical infection.

The major limitation of this study was the lack of a definitive diagnosis made in control teeth. We felt that given the low prevalence of periapical infection in the general equine population, using our inclusion criteria for controls i.e. normal occlusal surface of the tooth (Casey and Tremaine 2010) and no clinical signs of apical infection would produce a reliable control population for use in this study. However, without extraction and subsequent sectioning/histopathological examination of these teeth we have no definitive evidence that these teeth were not apically infected. This is accepted as a limitation of this study, although if all control CT were to be extracted and sectioned, our control population would probably be restricted to available post mortem material, for which other parameters such as age, breed and presence of clinical signs are often limited.

For the overall diagnosis of CT apical infection, moderate interobserver agreement was noted between Clinicians 1 and 2 (kappa = 0.56) and Clinicians 2 and 3 (0.53) with fair agreement noted between Clinicians 1 and 3 (0.34). These results are comparable to the 0.41–0.54 recorded by a previous study using 3 clinicians investigating the sensitivity of radiography for the diagnosis of CT apical infection (Weller et al. 2001) and better than the 0.38 recorded for 2 clinicians in a second study (Barakzai 2005), although it should be noted that there were major differences in the study design of these 2 previously published studies compared to the current study. Clinician 3 also appeared more sensitive (83%) compared to Clinicians 1 and 2 (73% and 76%), although this was offset by Clinician 3 having a reduced specificity (68%) compared to Clinicians 1 and 2 (93% and 90%).

However, for evaluation of the 12 individual radiographic signs with the exception of displaced tooth, Clinicians 1 and 2 had fair to moderate agreement (kappa 0.33–0.79) but quite poor agreement was noted between Clinician 3 and the other 2 clinicians for all variables (kappa 0.00–0.29). These findings are interesting as they may reflect the variability in methods of interpreting radiographs by the veterinary profession as a whole. It should be taken into account that the 3 clinicians that read radiographs in this study were working at a referral institution and specialising in equine soft tissue surgery. It is possible that sensitivity, specificity and inter-observer agreement may be lower for vets working in first opinion practice.

Diagnosis of apical infection

Median sensitivity (76%) and specificity (90%) for the overall diagnosis of apical infection generated in this study means that the probability of misclassifying disease was relatively low (probability of a false positive 10% or false negative 24%). These diagnostic values would be improved in a ‘real life’ clinical situation when other information such as clinical signs, results of oral examination and other diagnostic tests are available, as has been shown by previous authors (Gibbs and Lane 1987; Weller et al. 2001). Clinically, we feel it is better from both a welfare and economic point of view to make a false negative diagnosis of CT apical infection rather than a false positive one. A false positive diagnosis would involve removal of a healthy tooth which is invasive and can be associated with complications (Pritchard et al. 1992; Dixon et al. 2005, 2008), is expensive and associated with CT overgrowths for the remainder of the horse's life (Townsend et al. 2008). A false negative diagnosis on initial presentation often means that surgery (extraction) may be delayed by some weeks to months; however, in the vast majority of cases there are no increased risks of surgical complications occurring associated with such a delay though it is our experience that cementosis/cementoma formation may make extraction more difficult.

Although our study design was different, we found a similar specificity (90%) but higher sensitivity (76%) of radiography for diagnosis of CT apical infection compared to previous studies (specificity 70–95%, sensitivity 52–69%, Weller et al. 2001; Barakzai 2005). The improved sensitivity in this study may be due to the use of computed radiography, as compared to conventional film radiography as utilised in previous studies. The diagnostic sensitivity (76%) and specificity (90%) values generated in this study are similar to the 80% sensitivity and 86% specificity reported in the only other published abstract investigating the role of computed radiography in the diagnosis of apical infection (Casey et al. 2009). Although computed radiography has a lower spatial resolution than conventional film radiography (Cowen et al. 2007) this is thought to be clinically insignificant and not great enough to hinder the detection of pathological lesions (Swee et al. 1997). The greater dynamic range and improved contrast resolution, accentuates differences in opacity between different tissue types (Armbrust 2007; Marlof et al. 2008). Improved detection rates of pulmonary nodules (Kelcz et al. 1994) in the human field and increased sensitivity of computed radiography in detection of small volume pneumoperitoneum in canine cadavers (Marlof et al. 2008) has been shown. Generally this has not been the case in human dental imaging with poorer detection rates of simulated internal resorption cavities (Kamburoglu et al. 2007), periapical bone lesions (Kullendorf et al. 1996) and artificial bone lesions (Hadley et al. 2008) using direct radiography compared to conventional film radiography although the radiographic techniques used for digital and film radiography in humans are vastly different.

Individual radiographic signs

Based on the unadjusted values of sensitivity, specificity and interclinician agreement, the most consistent radiographic signs for the diagnosis of apical infection were periapical sclerosis, periapical halo formation, clubbing of one or 2 roots and degree of clubbing (median sensitivity 73–90%, median specificity 61–63%). These 4 signs are probably indicative of advanced CT infection, where lysis of the tooth root and supporting bone has occurred and this may explain their high sensitivity values (Baker 1971; Wyn-Jones 1985; Gibbs and Lane 1987). Analysis at the univariable level also demonstrated a dose-response effect for these radiographic signs: increasing severity of the radiographic sign was more likely to be associated with infection.

Conversely, radiographic signs detectable in the early stages of apical infection, in the authors' experience (e.g. widening of the periodontal space and loss of the LD) were found to be less reliable for detecting CT apical infection, in particular due to the very wide variability observed between clinicians. This may indicate differences in the abilities of individual clinicians to detect, interpret and weight these changes based on their personal experiences. Although the observation of a loss of LD was found to be one of the most sensitive radiographic signs (90%), it was also the least specific (34%) (i.e. low probability of false negatives [10%] but very high probability of false positives [66%]). A test with such a low specificity is undesirable in a clinical situation as it may result in the incorrect removal of a CT. Loss of LD was also associated with very low interclinician agreement (kappa 0.03–0.38) making it potentially unreliable. It is unfortunate that these early signs of infection are unreliable for making an early diagnosis, as apicoectomy or endodontic therapies could then be performed which aim to preserve the tooth, but are only successful in teeth without advanced changes (Simhofer et al. 2008).

The remaining 7 radiographic signs: cementum deposition around the reserve crown, cementoma formation, reserve crown fragmentation, reserve crown lucency, dental dysplasia and dental displacement had low sensitivity (range 2–39%) but good specificity (range 88–100%). Given the results from this study, it appears that these signs have little discriminatory ability in detecting periapical infection, generating a high proportion of false negatives but few false positives. As mentioned previously, a false negative diagnosis with respect to CT apical infection, although not desirable, has fewer welfare implications for the horse. Whilst dental displacement was included as an assessment category in this study, it is likely that only severe displacements where there is obvious overlap with the CT either side would be visible on lateral oblique radiographs, and hence we had very low numbers of displaced teeth identified in this study (n = 1 each for case and controls).

Although the independent evaluation of sensitivity and specificity of radiographic lesions is useful for comparison with other studies, a conditional multivariable approach was essential to account for the matched study design and the confounding effect of radiographic signs that are likely to be correlated. In the final model, mild sclerosis (DOR 4.1, 95%CI 1.8–9.6), moderate/extensive periapical sclerosis (DOR 72.0, 95%CI 17.6–295.6), mild periapical halo formation (DOR 2.6, 95%CI 1.0–6.3) and moderate/extensive periapical halo formation (DOR 4.4, 95%CI 1.5–12.9) were significantly associated with a diagnosis of apical infection. The model also took into account the variability of diagnoses by different clinicians and confirmed the results from the kappa analysis: Clinicians 1 and 2 were more similar than Clinician 3 with respect to the final diagnosis of periapical infection.

No significant difference was noted in the detection of CT apical infection between the rostral (Triadan 06–08) and caudal (Triadan 09–11) maxillary CT or between maxillary and mandibular CT. This in contrast to previous studies where radiographic signs of apical infection are more readily identified in the mandibular or rostral maxillary CT, whose apical regions lie within maxillary or mandibular bone rather than the air or fluid-filled maxillary sinuses (Baker 1971; Wyn-Jones 1985; Gibbs and Lane 1987; Dixon et al. 2000). Again this difference may be explained by the use of computed radiography in this study, although the only previous study using computed radiography in the diagnosis of equine apical infection also found a difference in detection rate between rostral and caudal maxillary CT (Casey et al. 2009). Alternatively, the cropped images may have influenced the results obtained. The images were cropped in such a manner that only a small amount of the surrounding teeth and apical region was visible so that other peripheral radiographic findings such as fluid or masses in the sinuses and abnormalities of adjacent teeth were not taken into account. Due to a lack of statistical power, individual radiographic signs could not be evaluated for each of the rostral and caudal maxillary CT groups, although when the radiographic signs present in the final model were evaluated descriptively there appeared to be similar numbers of both rostral and caudal maxillary CT displaying moderate to extensive periapical sclerosis and periapical halo formation. This may suggest that the findings of the final model can be applied equally to all the maxillary CT. Despite these signs being reportedly more difficult to see in the caudally positioned teeth due to the thinner bone surrounding their apices (Wyn-Jones 1985; Gibbs and Lane 1987). The cropping of the images does not replicate the ‘real life’ scenario or the method used for other studies to date, when the entire radiograph was available for evaluation, but, given our results, it is possible that this method of close evaluation of individual CT apices may offer some advantages.

Conclusion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgement
  9. Manufacturer's address
  10. References

This study has shown that blinded evaluation of digitally cropped radiographs taken using computed radiography appear to be more sensitive than previous methods for detecting CT apical infection. There were makred differences in the evaluation of individual radiographic signs by 1 clinician compared to the other 2, and this may reflect variation in methods of interpreting radiographs within the veterinary profession as a whole. When clinician variability and the confounding effects of other radiographic signs are accounted for multivariable conditional logistic regression identified periapical sclerosis and periapical halo formation as being significantly associated with CT apical infection. There was no significant difference in detection rate of CT apical infection between rostral and caudal maxillary CT, or maxillary and mandibular CT.

Acknowledgement

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgement
  9. Manufacturer's address
  10. References

Neil Townsend was a Horse Trust Clinical Training Scholar for the duration of this study.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgement
  9. Manufacturer's address
  10. References
  • Abramson, J.H. (2004) WINPEPI (PEPI-for-Windows): Computer programs for epidemiologists. Epidemiol. Perspect Innov. 1, 6.
  • Armbrust, L.J. (2007) Digital images and digital capture. In: Veterinary Diagnostic Radiology, 5th edn., Ed: D.E.Thrall, Saunders Elsevier, Philadelphia. pp 22-37.
  • Baker, G.J. (1971) Some aspects of dental radiology. Equine vet. J. 3, 46-51.
  • Barakzai, S.Z. (2005) Use of Scintigraphy for the Diagnosis of Suspected Equine Dental and Paranasal Sinus Disorders. MSc Thesis, University of Edinburgh.
  • Casey, M.B., Browne, W.J. and Tremaine, W.H. (2009) The diagnostic value of computed radiography in the investigation of pulpitis of equine cheek teeth. In: Proceedings of the British Equine Veterinary Association Congress, Equine Veterinary Journal, Newmarket. p 102.
  • Casey, M.B. and Tremaine, W.H. (2010) The prevalence of secondary dentinal lesions in cheek teeth from horses with clinical signs of pulpitis compared to controls. Equine vet. J. 42, 30-36.
  • Cowen, A.R., Davies, A.G. and Kengyelics, S.M. (2007) Advances in computed radiography systems and their physical imaging characteristics. Clin. Radiol. 62, 1132-1141.
  • Dacre, I.T., Kempson, S. and Dixon, P.M. (2008a) Pathological studies of cheek teeth apical infections in the horse: 4. Aetiopathological findings in 41 apically infected mandibular cheek teeth. Vet. J. 178, 241-351.
  • Dacre, I.T., Kempson, S. and Dixon, P.M. (2008b) Pathological studies of cheek teeth apical infections in the horse: 5. Aetiopathological findings in 57 apically infected maxillary cheek teeth. Vet. J. 178, 352-363.
  • Dixon, P.M. and Dacre, I.T. (2005) A review of equine dental disorders. Vet. J. 169, 165-187.
  • Dixon, P.M., Dacre, I., Dacre, K., Tremaine, W.H., McCann, J. and Barakzai, S. (2005) Standing oral extraction of cheek teeth in 100 horses (1998-2003). Equine vet. J. 37, 105-112.
  • Dixon, P.M., Hawkes, C.S. and Townsend, N.B. (2008) Complications of oral surgery. Vet. Clin. N. Am.: Equine Pract. 24, 499-514.
  • Dixon, P.M., Tremaine, W.B., Pickles, K., Kuhns, L., Hawe, C., McCann, J., McGorum, B.C., Railton, D.I. and Brammer, S. (2000) Equine dental disease part 4: A long-term study of 400 cases: Apical infections of cheek teeth. Equine vet. J. 32, 182-194.
  • Dohoo, I., Martin, W. and Stryhm, H. (2003) Screening and diagnostic tests. In: Veterinary Epidemiological Research, Eds: I.Dohoo, W.Martin and H.Stryhm, AVC Inc, Charlottetown. p 92.
  • Gibbs, C. and Lane, G.C. (1987) Radiographic examination of the facial, nasal and paranasal sinus regions of the horse. II. Radiological findings. Equine vet. J. 19, 474-482.
  • Glas, A.F., Lijmer, J.G., Prins, M.H., Bonsel, G.J. and Bossuyt, P.M.M. (2003) The diagnostic odds ratio: A single indicator of test performance. J. clin. Epidemiol. 56, 1129-1135.
  • Hadley, D.L., Replogle, K.J., Kirkam, J.C. and Best, A.M. (2008) A comparison of five radiographic systems to d-speed film in the detection of artificial one lesions. J. Endod. 34, 1111-1114.
  • Kamburoglu, K., Barenboim, S.F. and Kaffe, I. (2007) Comparison of conventional film with different digital and digitally filtered images in the detection of simulated internal resorption cavities – an ex vivo study in human cadaver jaws. Oral Surg. oral Med. oral Pathol. oral Radiol. Endod. 105, 790-797.
  • Kelcz, F., Zink, F.E., Peppler, W.W., Kruger, D.G., Ergun, D.L. and Mistretta, C.A. (1994) Conventional chest radiography vs dual energy computed radiography in the detection and characterisation of pulmonary nodules. Am. J. Roentgenol. 162, 271-278.
  • Kullendorf, B., Nilsson, M. and Rohlin, M. (1996) Diagnostic accuracy of digital direct dental radiography for the detection of periapical bone lesions: Overall comparison between conventional and direct digital radiography. Oral Surg. oral Med. oral Pathol. oral Radiol. Endod. 82, 344-350.
  • Landis, J.R. and Koch, G.G. (1977) The measurement of observer agreement for categorical data. Biometrics 33, 159-174.
  • Marlof, A., Blaik, M., Ackerman, N., Watson, E., Gibson, N. and Thompson, M. (2008) Comparison of computed radiography and conventional radiography in the detection of small volume pneumoperitoneum. Vet. Radiol. Ultrasound 49, 227-232.
  • Muyelle, S., Simeons, P., Lauwers, H. and van Loon, G. (1997) Ageing draft and trotter horses by their dentition. Vet. Rec. 141, 17-20.
  • Pritchard, M.A., Hackett, R.P. and Erb, H.N. (1992) Long-term outcome of tooth repulsion in horses: A retrospective study of 61 cases. Vet. Surg. 2, 145-149.
  • Simhofer, H., Stoian, C. and Zetner, K. (2008) A long term study of apicoectomy and endodontic treatment of apically infected teeth in 12 horses. Vet. J. 178, 411-418.
  • Swee, R.B., Gray, J.E. and Beabout, J.W. (1997) Screen-film vs. computed radiography imaging of the hand: A direct comparison. Am. J. Roentgenol. 168, 539-542.
  • Townsend, N.B., Dixon, P.M. and Barakzai, S.Z. (2008) Evaluation of the long-term oral consequences of equine exodontia in 50 horses. Vet. J. 178, 419-424.
  • Tremaine, W.H. and Dixon, P.M. (2001) A long-term study of 277 cases of equine sinonasal disease. Part 1: Details of horses, historical, clinical and ancillary diagnostic findings. Equine vet. J. 33, 274-282.
  • Weller, R., Livesy, L., Maierl, J., Nuss, K., Bowen, I.M., Cauvin, E.R.J., Weaver, M., Schumacher, J. and May, S.A. (2001) Comparison of radiography and scintigraphy in the diagnosis of dental disorders in the horse. Equine vet. J. 33, 49-58.
  • Wyn-Jones, G. (1985) Interpreting radiographs 6: Radiology of the equine head (Part 2). Equine vet. J. 17, 417-425.

Author contributions The initiation and conception of this study was by S.B., the execution by N.T., C.H., R.R. and S.B., the statistics by N.T. and L.B. and the writing by N.T., L.B. and S.B.