Tooth-level predictors of tooth loss and exposed bone after radiation therapy for head and neck cancer

Abstract

Background.

The objective of this study was to identify tooth-level risk factors for use during preradiation dental care management to predict risk of tooth failure (tooth lost or declared hopeless) and exposed bone after radiation therapy (RT) for head and neck cancer (HNC).

Methods.

The authors conducted a prospective observational multicenter cohort study of 572 patients receiving RT for HNC. Participants were examined by calibrated examiners before RT and then every 6 months until 2 years after RT. Analyses considered time to tooth failure and chance of exposed bone at a tooth location.

Results.

The following pre-RT characteristics predicted tooth failure within 2 years after RT: hopeless teeth not extracted pre-RT (hazard ratio [HR], 17.1; P < .0001), untreated caries (HR, 5.0; P < .0001), periodontal pocket 6 mm or greater (HR, 3.4; P = .001) or equaling 5 mm (HR, 2.2; P = .006), recession over 2 mm (HR, 2.8; P = .002), furcation score of 2 (HR, 3.3; P = .003), and any mobility (HR, 2.2; P = .008). The following pre-RT characteristics predicted occurrence of exposed bone at a tooth location: hopeless teeth not extracted before RT (risk ratio [RR], 18.7; P = .0002) and pocket depth 6 mm or greater (RR, 5.4; P = .003) or equaling 5 mm (RR, 4.7; P = .016). Participants with exposed bone at the site of a pre-RT dental extraction averaged 19.6 days between extraction and start of RT compared with 26.2 days for participants without exposed bone (P = .21).

Conclusions.

Individual teeth with the risk factors identified in this study should be considered for extraction before RT for HNC, with adequate healing time before start of RT.

Practical Implications.

The findings of this trial will facilitate evidence-based dental management of the care of patients receiving RT for HNC. This clinical trial was registered at Clinicaltrials.gov. The registration number is NCT02057510.

Head and neck cancers (HNCs) are among the 10 most common cancers worldwide.¹ Most patients with HNC receive radiation therapy (RT), often with concomitant chemotherapy. RT usually is delivered in fractions of approximately 2 Gy per day, 5 days per week, for 5 through 7 weeks, for a total dose of 50 through 70 Gy (5,000–7,000 cGy). Short-term adverse effects may include oral mucositis, oral candidiasis, hyposalivation, and taste changes.² Longer-term adverse effects can include hyposalivation, gingival recession, caries, tooth loss, and osteoradionecrosis.^3–8 Osteoradionecrosis manifests as persistent exposure of intraoral bone and can result in significant morbidity, including pain, infection, fistula, paresthesia, and jaw fracture. It commonly occurs after dental extraction but also can occur owing to odontogenic or periodontal infection and denture trauma or without an identified cause.⁹

Because of increased risk of experiencing osteoradionecrosis and other oral complications, patients with HNC should be referred for dental evaluation and care management before RT. The primary goal is to restore or extract diseased teeth before RT, so as to avoid the need for dental extraction and potential resulting osteoradionecrosis after RT.¹⁰ Therefore, teeth with poor long-term prognosis often are extracted pre-RT. However, there is limited evidence to guide decision making on which teeth should be extracted or retained. As a result, pre-RT dental care management varies widely across dental practices, driven by expert opinion, access, and practitioner experience.¹¹

The objective of our study was to identify tooth-level risk factors before RT that predict tooth loss and exposed bone at that specific site after RT. Such information would guide evidence-based dental care management for this patient population.

METHODS

Study design

We conducted a prospective observational cohort study called Observational Study of Dental Outcomes in Head and Neck Cancer Patients (also known as OraRad). Study participants were assessed before beginning RT and then every 6 months until 2 years after RT. We enrolled 572 participants across 6 clinical sites. Institutional review board approval was obtained at each site. All participants provided written informed consent.

Inclusion and exclusion criteria

Inclusion criteria were 18 years or older; diagnosis of head and neck squamous cell carcinoma or salivary gland cancer and intending to receive external beam RT with curative intent, or a diagnosis of a nonsquamous cell carcinoma, nonsalivary gland cancer head and neck malignancy and expected to receive at least 4,500 cGy RT to the head and neck region; and having at least 1 natural tooth remaining after pre-RT dental extractions. Patients receiving palliative RT or with a history of curative RT for HNC were excluded.¹²

Study procedures

Study visits were conducted before RT and 6, 12, 18, and 24 months after RT initiation. The baseline (pre-RT) assessment was conducted after completing clinical pre-RT dental care management, including extractions.

Study assessments included missing teeth, hopeless teeth, exposed bone, caries, periodontal measures, mobility, and other assessments.¹²

A tooth was recorded as missing if no part of the tooth was clinically visible. A tooth was recorded as hopeless if it was present and met any of these criteria: nonrestorable due to fracture or extensive caries, amputated crown with root remaining, and persistent or uncontrolled odontogenic or periodontal infection. Tooth failure was defined as a tooth that was either newly missing or declared hopeless since the last study visit.⁵

Exposed bone was defined as visible exposed bone in the mouth, with loss of the overlying soft tissue. Location of exposed bone was recorded using tooth numbers of teeth present or normally present in that location.⁸

Caries assessment was done by recording decayed, missing, and filled surfaces (DMFS) at the baseline and 6-, 12-, 18-, and 24-month visits. A no. 23 or no. 2A explorer was used to detect caries. Four tooth surfaces on the anterior teeth and 5 tooth surfaces on the posterior teeth (except third molars) were scored for DMFS.⁴

Periodontal measures were collected on all teeth except third molars at the baseline and 12- and 24-month visits. Pocket depth and the distance from the cementoenamel junction (CEJ) to the gingival margin (CEJ-GM) was measured using a UNC 15 probe at 6 sites per tooth. Bleeding on probing was recorded at each tooth’s 6 periodontal probing sites. Furcation involvement was scored on multirooted teeth as Class 0, 1, 2, 3, or 4, using Glickman classification¹³ and a Naber probe.⁶

Mobility was measured on all teeth present except third molars, using the nonworking ends of the dental mirror and probe, pressed on the buccal and lingual surfaces of the tooth. Mobility was rated as Class 0, 1, 2, or 3, using the classification by Grace and Smales.¹⁴

All study personnel received detailed training on conducting clinical assessments, completing study forms, entering data, and other procedures. All clinical examiners underwent annual inperson calibration on healthy volunteers for DMFS and periodontal measurements.

Statistical methods

The statistical methods are described briefly here; the Appendix, available online at the end of this article, gives a detailed description.

Time to Tooth Failure

These were time-to-event (survival) analyses, in which the event was tooth failure and possible event times were 6, 12, 18, and 24 months. Third molars were excluded. The analyses had the form of Cox regression, so that relative hazards are reported for comparisons. The Appendix, available online at the end of this article, gives detailed definitions of at-risk teeth and events, analysis methods, and software. Adjusted analyses added person-level characteristics, listed in the Appendix.

Exposed Bone

The location of an exposed-bone lesion was defined using the tooth numbers to which it was closest. The outcome, an exposed-bone report, occurred for a tooth if that tooth was included in an exposed bone lesion’s location at a follow-up visit. Analyses used generalized estimating equations to estimate the probability of an exposed bone at a tooth location, so that relative risks are reported for comparisons. Adjusted analyses added person-level characteristics, listed in the Appendix, available online at the end of this article.

RESULTS

We enrolled 572 participants across 6 clinical sites from 2014 through 2018. Of these, 52 participants had no follow-up visits. Tooth failure data were thus available for 520 participants. Exposed bone data were available for 1 additional participant based on chart review. Table 1 provides pre-RT characteristics of the 521 participants with follow-up for exposed bone.

Table 1.

Characteristics of the 521 participants with follow-up for exposed bone.

CHARACTERISTIC	VALUE
Age, Y (SD)	58.0 (11.0)
Male Sex, No. (%)	407 (78)
Race, No. (%) *
White	431 (85)
Black	39 (8)
Other (including multiple)	37 (7)
Education Past High School, No. (%) †	376 (72)
Dental Insurance, No. (%)	334 (64)
Smoking, No. (%) ‡
Current	23 (4)
Former	269 (52)
Never	228 (44)
Radiation Type, No. (%) †
Any intensity-modulated RT	489 (94)
Proton only	30 (6)
Primary Site of RT, No. (%) §
Oropharynx	245 (51)
Oral cavity	68 (14)
Larynx or hypopharynx	36 (7)
Salivary gland	50 (10)
Other	86 (18)
Total RT Dose to Primary Site, cGy (SD) ‡	6,592 (624)
Mean RT Dose to the Mandible, cGy (SD) ‡	3,552 (1,328)
Mean RT Dose to the Parotid Glands, cGy (SD) ¶
Ipsilateral dose	3,499 (1,394)
Contralateral dose	1,928 (1,035)
Received Surgery Before RT, No. (%) #	290 (60)
Received Chemotherapy Concurrently With RT, No. (%) #	327 (67)
Received Routine Dental Care in the Prior 12 Months, No. (%)	381 (73)
Had ≥ 1 Dental Extractions as Part of Pre-RT Dental Care Management, No. (%)	156 (30%)
Teeth Extracted During Pre-RT Dental Care Management, Mean No. (SD)
All participants	1.4 (3.3)
Those who had extractions	4.8 (4.6)
Teeth Present at Baseline Visit (After Pre-RT Extractions), Mean No. (SD) ‡	23.0 (5.7)
Had ≥ 1 Hopeless Teeth Present at Baseline Visit, No. (%) ‡	30 (6)
Hopeless Teeth Present at Baseline Visit, Mean No. (SD) ‡
All participants	0.1 (0.7)
Those who had hopeless teeth	2.4 (1.6)
Had ≥ 1 Teeth With Untreated Caries Present at Baseline Visit, No. (%) ‡	160 (31)
Teeth With Untreated Caries at Baseline Visit, Mean No. (SD) ‡
All participants	0.9 (2.0)
Those who had untreated caries	2.9 (2.6)
Had ≥ 1 Teeth With a Periodontal Pocket of ≥ 5 mm at Baseline Visit, No. (%) **	298 (61)
Teeth With ≥ 1 Periodontal Pockets of ≥ 5 mm at Baseline Visit, Mean No. (SD) **
All participants	2.4 (3.5)
Those who had probing depth ≥ 5 mm	4.0 (3.7)
Maximum Furcation Score at Baseline Visit, No. (%) **
0	119 (24)
1	211 (43)
2	160 (33)
Teeth With Furcation Score ≥ 1 at Baseline Visit, Mean No. (SD) **
All participants	3.4 (2.9)
Those with max furcation score ≥ 1	4.5 (2.5)
Teeth With Furcation Score of 2 at Baseline Visit, Mean No. (SD) **
All participants	0.7 (1.3)
Those with maximum furcation score 2	2.2 (1.5)
Had ≥ 1 Teeth With CEJ-GM†† Distance < 0 at Baseline Visit, No. (%)**	471 (96)
Teeth With CEJ-GM Distance < 0 at Baseline Visit, Mean No. (SD) **
All participants	12.2 (7.2)
Those who had CEJ-GM < 0	12.7 (6.9)
Had ≥ 1 Teeth With Mobility Score ≥ 1 at Baseline Visit, No. (%) ‡‡	112 (22)
Teeth With Mobility Score ≥ 1 at Baseline Visit, Mean No. (SD) ‡‡
All participants	0.7 (1.8)
Those who had mobility score ≥ 1	3.1 (2.6)
Had ≥ 1 Teeth With Bleeding on Probing at Baseline Visit, No. (%) §§	453 (92)
Teeth With Bleeding on Probing at Baseline Visit, Mean No. (SD) §§
All participants	8.7 (6.5)
Those who had teeth with bleeding on probing	9.4 (6.2)

^*Five hundred and seven participants reported race.

^†Five hundred and nineteen participants reported education and radiation type.

^‡Five hundred and twenty participants had data for smoking status, total radiation therapy (RT) dose to the primary site, mean RT dose to the mandible, number of teeth present at the baseline visit, number of hopeless teeth at the baseline visit, and number of teeth with untreated caries at the baseline visit.

^§Four hundred and eighty-five participants had data for primary RT location.

^¶For ipsilateral, 476 participants had dose with side identifiable; for contralateral, 520 participants had dose with side identifiable.

^#Four hundred and eighty-seven participants had surgery and chemotherapy data.

^**Four hundred and ninety participants had periodontal data, including furcation involvement.

^††CEJ-GM: Cementoenamel junction to gingival margin.

^‡‡Five hundred and eighteen participants had mobility data.

^§§Four hundred and ninety-one participants had bleeding on probing data.

Tooth failure after RT

The 520 participants with follow-up for tooth failure after RT had a total of 11,993 teeth present when RT began (excluding third molars) that were not considered to be hopeless, as defined in the Methods section above. Of these 11,993 teeth, 214 (1.8%) failed (were exfoliated, extracted, declared hopeless) within 2 years of follow-up after RT. Eighty-two of the 520 participants had at least 1 tooth failure within 2 years.⁵

Several pre-RT characteristics of individual teeth were associated significantly with risk of failure of that particular tooth within 2 years after RT. Table 2 and eTable 1, available online at the end of this article, list unadjusted and adjusted hazard ratios (HRs), 95% CIs, and P values.

Table 2.

Predictors of tooth failure within 2 years after radiation therapy.

COMPARISON	UNADJUSTED*			ADJUSTED†
	Hazard Ratio	95% CI	P Value	Hazard Ratio	95% CI	P Value
Retained Hopeless Teeth vs Other Teeth Present ‡	47.8	21.6 to 105.9	< .0001	17.1	7.9 to 37.1	< .0001
Untreated Caries vs Teeth Without Caries	10.0	5.3 to 18.9	< .0001	5.0	2.7 to 9.3	< .0001
Deepest Periodontal Pocket, mm
≥ 6 vs < 4	6.0	2.5 to 14.3	< .0001	3.4	1.6 to 7.0	.001
5 vs < 4	3.5	1.9 to 6.3	< .0001	2.2	1.3 to 4.0	.006
4 vs < 4	2.1	1.3 to 3.2	.001	1.7	1.1 to 2.5	.018
Furcation Score
2 vs 0	6.7	3.1 to 14.3	< .0001	3.3	1.5 to 7.3	.003
1 vs 0	1.6	0.7 to 3.6	.31	1.4	0.6 to 3.1	.40
Gingival Recession, Cementoenamel Junction to Gingival Margin Distance, mm
< −2 vs ≥ 0	8.3	4.5 to 15.5	< .0001	2.8	1.5 to 5.2	.002
−1 or −2 vs ≥ 0	2.8	1.6 to 5.0	.0006	2.8	0.99 to 3.2	.053
Mobility, Any Grade vs None	5.9	3.1 to 11.3	< .0001	2.2	1.2 to 4.0	.008
Bleeding on Probing, Yes vs No	1.5	0.7 to 3.1	.28	1.0	0.5 to 1.9	.98

^*These analyses include the same participants and teeth included in the adjusted analyses. Unadjusted analyses including all participants and teeth (that is, that have the data needed for the unadjusted analysis) are in eTable 1.

^†Adjusted analyses omitted teeth with missing data for adjusters. With exceptions as noted, analyses were adjusted for these person-level characteristics: baseline number of teeth (5 categories: ≤ 14, 15–24, 25–26, 27, 28); age (continuous); whether the person had untreated caries at baseline; alcohol use (< 12 drinks per year vs ≥ 12 drinks per year); radiation type (any intensity-modulated radiation therapy [RT] vs proton); smoking (never, former, or current); clinic; compliance change baseline to follow-up (4 categories, defined in the Methods section); radiation dose; nadir of saliva after RT as a percentage of baseline. Exceptions were as follows. For the tooth-level predictor untreated caries, the person-level adjuster “whether the person had untreated caries at baseline” was not included. For furcation score, only teeth with furcations were included, and a further adjuster was added, premolar vs molar.

^‡For this comparison only, teeth were at risk at baseline if they were present in the mouth, and only newly missing teeth were counted as events.

A tooth that was declared hopeless at the pre-RT examination but retained going into RT had the largest risk of extraction within 2 years, with adjusted HR of 17.1 compared with teeth not declared hopeless (P < .0001). The next strongest predictor of tooth failure was untreated caries pre-RT, with adjusted HR of 5.0 compared with teeth without untreated caries (P < .0001).

Compared with teeth with a maximum probing depth (PD) less than 4 mm, a maximum PD of 6 mm or more predicted tooth failure, with an adjusted HR of 3.4 (P = .001), whereas a maximum PD of 5 mm and 4 mm predicted tooth failure with adjusted HRs of 2.2 (P = .006) and 1.7 (P = .018), respectively.

A furcation score of 2 was predictive of tooth failure with an adjusted HR of 3.3 compared with a furcation score of 0 (P = .003). Gingival recession with the gingival margin 2 mm or more below the CEJ predicted tooth failure with an adjusted HR of 2.8 compared with teeth with gingival margin at or above the CEJ (P = .002). Finally, the presence of any degree of tooth mobility before RT predicted tooth failure with an adjusted HR of 2.2 compared with teeth with no mobility (P = .008).

Exposed intraoral bone after RT

Thirty-five participants received diagnoses of a total of 37 exposed bone lesions during 2 years after RT.⁸ Average (SD) maximum RT dose to the exposed bone area was 5,456 (1,768) cGy. Of these 37 lesions, 13 occurred at the site of dental extraction performed for pre-RT management. Participants with exposed bone at the site of pre-RT dental extraction had a mean (SE) of 19.6 (5.0) days between extraction and start of RT compared with 26.2 (1.4) days for participants who did not experience exposed bone (P = .21).

Several pre-RT characteristics of individual teeth predicted risk for exposed bone at that particular site within 2 years after RT. Table 3 and eTable 2, available online at the end of this article, list unadjusted and adjusted relative risks (RRs), 95% CIs, and P values.

Table 3.

Predictors of exposed bone within 2 years after radiation therapy.

COMPARISON	UNADJUSTED*			ADJUSTED†
	Relative Risk	95% CI	P Value	Relative Risk	95% CI	P Value
Retained Hopeless Teeth vs Missing Pre-RT	11.8	2.9 to 47.0	.0005	18.7	4.0 to 88.1	.0002
Retained Hopeless Teeth vs Extracted Pre-RT	2.9	0.6 to 13.9	.17	6.4	1.2 to 35.2	.034
Extracted Pre-RT vs Missing Pre-RT	4.0	1.1 to 14.2	.031	2.9	0.83 to 10.5	.096
Untreated Caries vs Teeth Without Caries	1.2	0.3 to 4.8	.78	1.2	0.4 to 3.8	.81
Deepest Periodontal Pocket, mm
> 6 vs < 4	3.5	0.6 to 19.6	.15	5.4	1.8 to 16.0	.003
5 vs < 4	2.8	0.8 to 10.2	.13	4.7	1.3 to 16.4	.016
4 vs < 4	1.4	0.5 to 4.2	.51	2.2	0.7 to 6.5	.17
Furcation Score
2 vs 0	1.4	0.4 to 5.3	.60	1.4	0.3 to 5.6	.66
1 vs 0	0.3	0.1 to 1.0	.056	0.3	0.1 to 1.1	.063
Gingival Recession, Cementoenamel Junction to Gingival Margin Distance, mm
< −2 vs ≥ 0	0.9	0.3 to 2.5	.87	1.1	0.4 to 2.7	.91
−1 or −2 vs ≥ 0	1.6	0.5 to 5.7	.46	2.4	0.6 to 9.5	.20
Mobility, Any Grade vs None	4.5	1.1 to 18.2	.032	3.7	0.9 to 15.3	.071
Bleeding on Probing, Yes vs No	0.7	0.3 to 1.6	.42	0.7	0.3 to 1.6	.39

^*These analyses include the same participants and teeth included in the adjusted analyses. The adjusters had little missing data; unadjusted analyses including all participants and teeth (that is, that had the data needed for the unadjusted analysis) give nearly identical results.

^†Adjusted analyses omitted teeth with missing data for adjusters. With exceptions as noted, analyses were adjusted for these person-level characteristics: baseline [BL] status (had BL extractions vs had no BL extractions but at least 1 tooth declared hopeless at BL vs had no BL extractions or teeth declared hopeless), number of teeth extracted at BL, smoking status (current, former, never), enrollment site, primary radiation therapy (RT) site (5 categories listed in Table 1), and total dose to the primary site. Exceptions were for the tooth-level predictor BL tooth status (missing before pre-RT care, extracted during pre-RT care, hopeless, present and not hopeless), these adjusters were omitted: person-level BL status and number of teeth extracted at baseline. For furcation involvement, only molar teeth were included and only these adjusters were used: BL status (had BL extractions vs had no BL extractions but at least 1 tooth declared hopeless at BL vs had no BL extractions or teeth declared hopeless), number of teeth extracted at BL, and smoking status (current, former, never).

A tooth declared hopeless at the pre-RT examination but retained going into RT had the largest risk of developing exposed bone at that site, with an adjusted RR of 18.7 compared with teeth already missing (P = .0002). Leaving a hopeless tooth in place was associated with a 6.4-fold higher adjusted risk of developing exposed bone compared with extracting a tooth shortly before RT (P = .034). Extraction of a tooth as part of pre-RT dental care management was associated with a 2.9-fold greater risk of developing exposed bone at that site compared with a missing tooth; however, this was not statistically significant (P = .096).

Compared with teeth with a maximum PD of 4 mm, a maximum PD of 6 mm or greater predicted exposed bone with an adjusted RR of 5.4 (P = .003), whereas a maximum PD of 5 mm predicted exposed bone with an adjusted RR of 4.7 (P = .003).

DISCUSSION

Patients who are to undergo RT for HNC should be referred for dental evaluation and care management before starting RT.¹⁵ This includes identifying teeth with poor long-term prognoses, considering the expected hyposalivation and increased risk of developing caries and gingival recession after RT. Such teeth often are extracted pre-RT if they will be in the field of high-dose RT (> 5,000 cGy). The rationale is to avoid extractions after RT owing to increased risk of developing osteoradionecrosis. In some cases (for example, fractured teeth), the determination of poor long-term prognosis is clear. However, in many situations with varying degrees of severity (for example, periodontal disease), this determination is not as clear. This decision making is complicated by the fact that a tooth with a good long-term prognosis in an otherwise healthy patient may have a poor long-term prognosis in a patient with significant hyposalivation after RT for HNC. Owing to limited evidence to guide such decision making, pre-RT dental care management, when available, often varies widely between centers, which can lead to suboptimal outcomes.¹¹

Our study provides evidence at the tooth level to help guide this complex decision-making process. We examined pre-RT tooth-level risk factors that increased risk of experiencing tooth failure, defined as a tooth being extracted or declared hopeless after RT. We used this definition because extractions are often avoided in areas that have received high-dose RT owing to risk of developing osteoradionecrosis. We found that the greatest risk of experiencing tooth failure after RT was for teeth that were already identified as hopeless but not extracted before RT. Such teeth had a 17-fold higher risk of experiencing tooth failure after RT than other teeth present in the mouth. The next highest risk of experiencing tooth failure was for teeth with untreated caries going into RT, which had a 5-fold higher risk of experiencing tooth failure than teeth without active caries at the pre-RT assessment. Because there is urgency to start RT to treat cancer, limited time is available for pre-RT dental care management. Priority is given to completing needed extractions, and caries management is sometimes deferred until after RT. However, our data show that if carious teeth are not restored before or soon after RT, there is a high risk of experiencing tooth failure.

Periodontal disease presents a particular challenge during the pre-RT dental assessment because of its varying degrees of severity. A finding reported in 2022 was that RT for HNC led to a significant increase in gingival recession.⁶ Our analyses showed that teeth with the gingival margin already at 2 mm or more apical to the CEJ before RT had a 2.8-fold increased risk of experiencing tooth failure after RT. We also found that teeth with at least 1 periodontal pocket 5 mm or greater had significantly increased risk of experiencing tooth failure than teeth with maximum pocket depth less than 4 mm. A furcation score of 2 or above and any degree of mobility pre-RT also predict increased risk of tooth failure after RT.

We also examined characteristics of teeth at the pre-RT assessment that predicted exposed bone at those sites after RT. Like tooth failure, the highest risk of developing exposed bone was for teeth already identified as hopeless but not extracted before RT. Leaving a hopeless tooth in place was associated with a 6.4-fold higher risk of developing exposed bone at that site than extracting such teeth before RT. We also found that a tooth with at least 1 periodontal pocket of 5 mm or greater was at increased risk of developing exposed bone at that site. Pocket depth of 6 mm or greater was associated with still greater risk. This finding is relevant as not all cases of exposed bone follow an extraction. Thus, our data confirm previous reports that periodontal disease is also a risk factor for exposed bone after RT.¹⁶ Although not the focus of our report, person-level factors such as tobacco and alcohol use can also increase risk of developing exposed bone.⁸

One of the challenges during pre-RT dental care management is that a dental extraction performed shortly before RT actually can cause osteoradionecrosis if the extraction site is not allowed to heal adequately before RT begins.¹⁷ Of the 37 exposed bone lesions recorded in our study, 13 lesions (35%) occurred at the site of a dental extraction completed for pre-RT management. A factor governing this risk is the duration of the healing period between the extraction and first day of RT. This period was an average of 19.6 days for participants who experienced exposed bone at the extraction site vs 26.2 days for those who did not. Although this difference was not statistically significant (both groups had few exposed bone events, so statistical power was low for this comparison), these findings are clinically significant. This difference of 6.6 days suggests that from a purely dental perspective, a healing period of at least 3 weeks is preferable compared with the recommendation of 2 weeks being used now.¹⁵ Our results are supported by a 2022 systematic review, which reported a significantly greater risk of experiencing osteoradionecrosis in patients with less than 2 weeks between extractions and start of RT compared with a period of 2 weeks through 1 month.¹⁸ In some cases, however, an optimal healing period may not be achievable owing to an urgent need to start RT for tumor control. The desirable healing period is also likely to be influenced by the degree of trauma to bone due to extraction. As most cases of osteoradionecrosis occur in the posterior mandible, pre-RT extractions of mandibular molars likely require a longer healing period than extractions in other areas. Risk of developing osteoradionecrosis posed by pre-RT dental extractions also raises the question of whether such extractions should be performed at all shortly before RT, especially given the urgency to start RT. It is worth repeating, however, that leaving a hopeless tooth in place is associated with a 6.4-fold higher risk of developing exposed bone at that site than extracting a tooth shortly before RT. Thus, our findings support the rationale for the practice of pre-RT dental extractions of teeth with poor prognosis but also indicate that if the healing period is not adequate, this practice can lead to osteoradionecrosis, the problem it is intended to prevent. On the basis of our findings, the Box lists characteristics of teeth that should be considered for extraction before RT, if they are in an area expected to receive more than 5,000 cGy of RT. It should be noted that the final decision about which teeth to extract should take into account all relevant factors, including oral hygiene status and practices as well as the presence of single vs multiple risk factors for tooth loss and osteoradionecrosis.

Box. Risk factors for tooth loss and exposed bone in patients who will receive radiation therapy for head and neck cancer.

• Only roots remaining

• Not restorable due to any reason, including fractured crown

• Active abscess that persists or recurs after treatment

• Clinically evident tooth mobility

• Periodontal pocket depth of 5 mm or greater

• Furcation score of 2 or higher

• Gingival margin apical to the cementoenamel junction by 2 mm or greater

Two important systematic reviews on this topic were published in 2022. Urquhart and colleagues¹⁹ reported that patients with HNC who needed pre-RT dental intervention may have had an increased risk of experiencing osteoradionecrosis compared with those who did not. As the studies reviewed included both pre- and post-RT extractions, the authors could not exclude the possibility that post-RT extractions contributed to the increased risk of developing osteoradionecrosis. These findings are consistent with published evidence that patients with poor dental status at the start of RT, who are more likely to need pre-RT dental extractions, are also more likely to lose additional teeth after RT.⁵ Another systematic review and meta-analysis by Normando and colleagues²⁰ reported a significant association between osteoradionecrosis and post-RT dental extractions but not pre-RT dental extractions. They concluded that performing dental extractions before RT reduces the risk of developing osteoradionecrosis compared with post-RT extractions. This is consistent with our study, which reports a 6.4-fold higher risk of developing exposed bone for hopeless teeth that were not extracted before RT compared with teeth extracted before RT.

The strengths of our study include the prospective multicenter design with a large sample size, rigorous annual calibration and training of study personnel, and external monitoring by multiple groups. A limitation of this analysis is that there were relatively few tooth failures and exposed bone events during the 2-year study period. However, this is not surprising because the study participants all received comprehensive pre-RT dental care management. Furthermore, the number of events was adequate to identify multiple risk factors for each of the outcomes. We were not able to separately assess risk only for teeth receiving high doses of RT, as such tooth-level dose data were not available for most teeth. Another limitation is that we assessed and recorded the presence of exposed intraoral bone, whereas a diagnosis of osteoradionecrosis was only made by the treating clinician. As the commonly used definitions of osteoradionecrosis require persistence of exposed bone for at least several weeks, we were unable to determine definitively such a diagnosis for each case of exposed bone.²¹ Additional studies with longer follow-up times would be helpful in identifying risk factors for long-term risk of developing osteoradionecrosis.

CONCLUSIONS

This study provides quantitative estimates of risk of experiencing tooth failure and exposed bone after RT, based on pre-RT characteristics of individual teeth. These risk ratios will help guide the evidence-based dental care management of patients undergoing RT for HNC.

ABBREVIATION KEY

BL

Baseline

CEJ

Cementoenamel junction

CEJ

Cementoenamel

GM

junction to gingival margin

DMFS

Decayed, missing, and filled surfaces

GM

Gingival margin

HNC

Head and neck cancer

IMRT

Intensity-modulated radiation therapy

PD

Probing depth

RT

Radiation therapy

APPENDIX: DETAILED STATISTICAL METHODS

Number of participants with follow-up data

Of the 572 participants who enrolled in the Observational Study of Dental Outcomes in Head and Neck Cancer Patients (OraRad) study and were eligible for follow-up, 520 had 1 or more follow-up visits and provided data about tooth failure. One more person provided data about exposed bone based on chart review, giving a total of 521 participants. Table 1 describes the 521 with some follow-up for exposed bone.

Time to tooth failure

This study was a time-to-event analysis. Possible event times were 6, 12, 18, and 24 months, as discussed in the article by Brennan and colleagues.^e1 Third molars were excluded from all analyses because the study did not record whether third molars were present at baseline. Otherwise, the definitions of teeth at risk and of the event are as follows:

1. For comparisons other than those in items B and C below: Teeth were at risk if they were present at baseline and had not been deemed hopeless. An at-risk tooth had an event during follow-up if it was deemed hopeless or exited the mouth via extraction or exfoliation.

2. For the comparison of retained hopeless teeth at baseline vs teeth extracted pre-radiation therapy (RT) vs teeth missing pre-RT vs teeth present and not hopeless: The group of teeth at risk was increased to include the 77 teeth that were declared hopeless at baseline but retained in the mouth. An at-risk tooth had an event if the tooth exited the mouth via extraction or exfoliation.

3. For the comparison of furcation scores: The group of teeth at risk was as under item A except that it included only teeth with furcations, that is, molars and premolars 5 and 12. The event was defined as under item A.

Analyses were clustered Cox regressions computed in the R system (R Project) using the survival package (https://CRAN.R-project.org/package=survival), specifically using the cluster option and the estimating equation method with robust SEs (Huber sandwich estimator) and Wald P values, with ties handled using the Efron approximation. Sample code is

$$\begin{gathered} \text{tf2un}<-\mathrm{c}\mathrm{o}\mathrm{x}\mathrm{p}\mathrm{h}(\text{Surv(time}=\text{ptar2\$'Alt}\text{TL}\text{time}\text{to}\text{failure/censor',} \\ \text{event=ptar2\$'Alt}\text{TL}\text{Event}\left.=1,\text{Censor}=0^{\mathrm{\text{'}}}\right)\sim \\ \text{BLExtHpls,}\text{cluster=PID,}\text{data}=\text{ptar2,}\text{na.action="na.omit")} \end{gathered}$$

Adjusted analyses were computed by adding adjusters to the analyses described above. Adjusters included person-level characteristics that had P value below .05 in figure 2 of the article from Brennan and colleagues.^e1 With exceptions noted below, analyses were adjusted for the following person-level characteristics: baseline number of teeth (5 categories: ≤ 14, 15–24, 25–26, 27, 28; Brennan and colleagues^e1 give the rationale for these categories), age (continuous), whether the person had untreated caries at baseline, alcohol use (<12 drinks per year vs ≥ 12 drinks per year), radiation type (any intensity-modulated RT vs proton), smoking (never, former, current), clinic, compliance with hygiene (4 categories: compliant at baseline and follow-up, compliant at baseline but not follow-up, compliant at follow-up but not baseline, compliant at neither baseline nor follow-up), radiation dose to the parotid glands (the first principal component of 4 measures, the maximum and mean dose to the right and left parotid glands), and nadir of salivary flow after RT as a percentage of baseline. Exceptions were for the tooth-level predictor of untreated caries, the person-level adjuster of whether the person had untreated caries at baseline was not included. For furcation score, a further adjuster was added: premolar vs molar.

These analyses omitted teeth that had missing data for the comparison variable (for example, baseline maximum pocket depth) or for the adjusters. Because some adjusters had substantial amounts of missing data (especially periodontal measures and compliance with hygiene), the teeth included in the unadjusted analyses were a rather larger superset of the teeth included in the adjusted analyses. The Results section shows unadjusted analyses of the same subset of teeth that were included in the adjusted analyses. eTable 1 shows unadjusted analyses using all teeth that had data for a given comparison; the results are similar to those shown in Table 2.

Exposed bone

In OraRad, the location of an exposed-bone lesion was defined using the tooth numbers to which it was closest, so a lesion’s location could extend across more than 1 tooth number; for example, 1 lesion extended across 4 teeth. For this analysis, the outcome, an exposed bone report at a tooth, was defined as occurring if that tooth was included in an exposed bone lesion’s location at a follow-up visit and not occurring otherwise. With this definition, 35 participants and 63 teeth had events.

Analyses used generalized estimating equations computed in the R system using the geepack package (https://cran.r-project.org/web/packages/geepack/index.html),^e2 specifically the geeglm function, with the log (not logit) link with offset natural log of years to the last follow-up visit (that is, 0.5, 1, 1.5, or 2 years), working correlation AR1 within person for teeth included in the analysis, fixed scale parameter, and robust standard errors. Pairwise comparisons between categories used the R system emmeans package (https://cran.r-project.org/web/packages/emmeans/index.html), specifically the emmeans function with Satterthwaite degrees of freedom, no adjustment for multiple comparisons, and default settings otherwise. Sample code is

$$\begin{gathered} \text{eb12un}\text{<-geeglm}\text{('Exposed}\text{bone'}\sim \text{offset}\text{('}\mathrm{l}\mathrm{n}(\text{MAXVISIT/12)}\text{for}\text{EB')} \\ +\text{BLExtHpls,}\mathrm{i}\mathrm{d}=\text{PID,}\text{family=binomial}(\text{link="log"),}\text{data=ptnm,} \\ \text{corstr=‘arl’, scale.}\text{fix}=\text{TRUE, na.action}=\text{na.omit}) \\ \text{emmeans}\text{(eb12un,}\text{pairwise}\sim \text{BLExtHpls, lmer.df}\text{=}\text{"satterthwaite",} \\ \text{adjust="none")} \end{gathered}$$

For comparisons involving pocket depth, bleeding on probing, tooth mobility, furcation involvement, and recession, many teeth (locations) had missing data; for example, if a tooth was missing, none of these measurements was defined for that tooth. The analysis thus included a category for teeth that otherwise would have been included but for which the predictor was missing; for example, for bleeding on probing, a tooth was in 1 of these categories: yes (that is, a site on the tooth bled on probing), no (no site on the tooth bled on probing), or missing (the tooth was missing but would have been included had it not been missing). The missing category did not include third molars, for which periodontal measurements were not made, nor did it include participants who did not have periodontal measurements because they would need antibiotic prophylaxis.

Except as described in the previous paragraph, these analyses omitted teeth that had missing data for the comparison variable or for the adjusters. The adjusters (described in the following paragraph) had little missing data, so the Results section presents unadjusted analyses including all participants and teeth that have the data needed for the unadjusted analysis.

With exceptions noted below, analyses were adjusted for person-level characteristics that had P values below .05 in Table 3 of the article by Treister and colleauges^e3 baseline status (had baseline extractions vs had no baseline extractions but at least 1 tooth declared hopeless at baseline vs had no baseline extractions or teeth declared hopeless), number of teeth extracted at baseline, smoking (current, former, never), enrollment site, primary RT site (5 categories: oropharynx, oral cavity, larynx or hypopharynx, salivary gland, other), and total dose to the primary site. Exceptions were as follows: for the tooth-level predictor baseline tooth status (missing before pre-RT care, extracted during pre-RT care, hopeless, present and not hopeless), these adjusters were omitted: person-level baseline status and number of teeth extracted at baseline. For furcation involvement, only molars were included and only these person-level adjusters were used: baseline status (as above), number of teeth extracted at baseline, and smoking (current, former, never).

eTable 1.

Predictors of tooth failure within 2 years after radiation therapy.^*

COMPARISON	HAZARD RATIO	95% CI	P VALUE
Retained Hopeless Teeth vs Other Teeth Present †	35.4	15.9 to 78.8	< .0001
Untreated Caries vs Teeth Without Caries	9.6	5.5 to 16.9	< .0001
Deepest Periodontal Pocket, mm
≥ 6 vs < 4	6.5	2.8 to 15.0	< .0001
5 vs < 4	3.5	2.0 to 6.3	< .0001
4 vs < 4	2.0	1.4 to 3.0	.0003
Furcation Score
2 vs 0	9.8	4.5 to 21.3	< .0001
1 vs 0	2.4	1.1 to 5.4	.036
Gingival Recession, Cementoenamel Junction to Gingival Margin Distance, mm
< −2 vs ≥ 0	8.7	5.0 to 15.2	< .0001
−1 or −2 vs ≥ 0	3.5	2.1 to 6.0	< .0001
Mobility, Any Grade vs None	5.6	3.2 to 9.8	< .0001
Bleeding on Probing, Yes vs No	1.5	0.9 to 2.7	.16

^*Unadjusted analyses including all participants with the necessary data.

^†For this comparison only, teeth were at risk at baseline if they were present in the mouth, and only newly missing teeth were counted as events.

eTable 2.

Predictors of exposed bone within 2 years after radiation therapy, including the category of missing.^*

COMPARISON	UNADJUSTED†			ADJUSTED‡
	Relative Risk	95% CI	P Value	Relative Risk	95% CI	P Value
Retained Hopeless Teeth vs Missing Pre-RT §	11.8	2.9 to 47.0	.0005	18.7	4.0 to 88.1	.0002
Extracted Pre-RT vs Missing Pre-RT	4.0	1.1 to 14.2	.031	2.9	0.83 to 10.5	.096
Retained Hopeless Teeth vs Extracted Pre-RT	2.9	0.6 to 13.9	.17	6.4	1.2 to 35.2	.034
Untreated Caries vs Teeth Without Caries	1.2	0.3 to 4.8	.78	1.2	0.4 to 3.8	.81
Deepest Periodontal Pocket, mm
≥ 6 vs < 4	3.5	0.6 to 19.6	.15	5.4	1.8 to 16.0	.003
5 vs < 4	2.8	0.8 to 10.2	.13	4.7	1.3 to 16.4	.016
4 vs 4	1.4	0.5 to 4.2	.51	2.2	0.7 to 6.5	.17
Missing¶	3.7	1.5 to 9.1	.005	4.1	1.6 to 10.3	.003
Furcation Score
2 vs 0	1.4	0.4 to 5.3	.60	1.4	0.3 to 5.6	.66
1 vs 0	0.3	0.1 to 1.0	.056	0.3	0.1 to 1.1	.063
Missing¶	1.2	0.5 to 3.0	.73	0.8	0.3 to 2.1	.71
Gingival Recession, Cementoenamel Junction to Gingival Margin Distance, mm
< −2 vs ≥ 0	0.9	0.3 to 2.5	.87	1.1	0.4 to 2.7	.91
−1 or −2 vs ≥ 0	1.6	0.5 to 5.7	.46	2.4	0.6 to 9.5	.20
Missing¶	2.7	1.1 to 6.6	.034	2.6	1.01 to 6.9	.047
Mobility
Any grade vs none	4.5	1.1 to 18.2	.032	3.7	0.9 to 15.3	.071
Missing¶	2.5	1.1 to 5.9	.033	2.1	0.9 to 5.1	.091
Bleeding on Probing
Yes vs no	0.7	0.3 to 1.6	.42	0.7	0.3 to 1.6	.39
Missing¶	2.4	1.01 to 5.7	.048	2.1	0.9 to 5.1	.094

^*This version of Table 3 includes results for the category of missing, when one was included for a comparison. In this table, the relative risks for a given predictor are relative to the predictor’s reference group, which is different for each predictor, so the missing category is not expected to have the same relative risk for the different predictors.

^†These analyses include the same participants and teeth included in the adjusted analyses. The adjusters had little missing data; unadjusted analyses including all participants and teeth (that is, that have the data needed for the unadjusted analysis) give nearly identical results.

^‡Adjustments are the same as in the version of Table 3 included in the main article.

^§RT: Radiation therapy.

^¶The analysis included a category capturing teeth for which the predictor was missing. The missing category does not include third molars, for which Observational Study of Dental Outcomes in Head and Neck Cancer Patients did not measure periodontal data, and it does not include study participants who did not have periodontal measures obtained because antibiotic prophylaxis would have been necessary. Teeth labeled missing were almost entirely missing non–third-molar teeth in people who otherwise had periodontal measurements obtained, with the few exceptions being teeth that were present but for which pocket depth or free gingival margin to cementoenamael junction distance could not be measured for any of the 6 sites on the tooth.