Vitamin D Status and Five Year Changes in Periodontal Disease Measures among Postmenopausal Women: The Buffalo OsteoPerio Study

Abstract

Background

Vitamin D is hypothesized to prevent periodontal disease progression through its immune modulating properties and its role in maintaining systemic calcium concentrations. We investigated associations between plasma 25(OH)D (collected 1997–2000) and the five-year change in periodontal disease measures from baseline (1997–2000) to follow-up (2002–2005) among 655 postmenopausal women in a Women’s Health Initiative Observational Study ancillary study. Exploratory analyses were conducted in 628 women who also had 25(OH)D measures at follow-up.

Methods

Four continuous measures of five-year change in periodontal disease were assessed using alveolar crestal height (ACH), clinical attachment level (CAL), probing pocket depth (PD), and percent of gingival sites that bled upon assessment. Linear regression was used to estimate beta-coefficients (β), standard errors (SE), and p-values corresponding to change in periodontal disease (either a 1 mm change in ACH, CAL or PD, or 1 unit change in the percent of gingival sites that bled) for a 10 nmol/L difference in 25(OH)D. Models were adjusted for age, education, dental visit frequency, smoking, diabetes status, current medications affecting bone health, baseline measures of periodontal disease, body mass index, and recreational physical activity.

Results

No statistically significant associations were observed between baseline 25(OH)D and change in periodontal disease measures, overall or in a subset (n=442) of women with stable 25(OH)D concentrations (women whose 25(OH)D changed less than ± 20 nmol/L from baseline to follow-up). Results also did not vary significantly in analyses that were stratified by baseline periodontal disease status.

Conclusion

No association between baseline 25(OH)D and the subsequent five-year change in periodontal disease measures was observed. Vitamin D status may not influence periodontal disease progression. More studies are needed to confirm these results.

Study summary

In our prospective study in postmenopausal women, baseline vitamin D status, assessed using 25-hydroxyvitamin D concentrations, was not associated with the five-year progression of periodontal disease defined as changes in alveolar bone, clinical attachment, probing pocket depth, or gingival bleeding.

INTRODUCTION

Periodontal disease is a common, chronic, inflammatory disease of aging which, if not controlled, can lead to tooth loss. It is estimated that 8.7%, 30.0% and 8.5% of the US population over age 30¹ have mild, moderate and severe disease, respectively, based on a full-mouth periodontal examination and the current Centers for Disease Control and Prevention and the American Academy of Periodontology (CDC/AAP) definition². Among persons 50 to 64 years and ≥ 65 years, prevalence of any periodontal disease is estimated to be even higher at approximately 57% and 70% of the population, respectively¹. Modifiable factors that reduce development and progression of periodontal disease are of interest to the general public and to dental professionals who want to reduce the burden of tooth loss.

Vitamin D status has been hypothesized to prevent and reduce the progression of periodontal disease³. In the last decade, research has focused on vitamin D as a potential anti-inflammatory⁴ and anti-microbial agent⁵. Vitamin D is also essential in maintaining bone health and mineralization⁶, presumably inclusive of alveolar bone. In a cross-sectional analysis using data from postmenopausal women enrolled in the Buffalo Osteoporosis and Periodontal Disease (OsteoPerio) Study, an ancillary study of the Women’s Health Initiative Observational Study (WHIOS), we previously showed that vitamin D status, assessed with plasma concentrations of 25-hydroxyvitamin D (25(OH)D), was associated with clinical measures of oral health⁷. Women with 25(OH)D concentrations ≥ 50 compared to < 50 nmol/L had reduced odds of gingival bleeding, (a measure of gingival inflammation) and reduced odds of moderate-to-severe periodontitis assessed using the CDC/AAP definition. However, vitamin D status was not significantly associated with radiographic measures of alveolar crestal height (ACH), which tend to reflect the chronic phase of destructive periodontitis.

Most^7–12, although not all¹³, previous cross-sectional and case-control studies have supported vitamin D status as a potential modifiable risk factor for periodontal disease. Few studies^14–18 have examined associations between vitamin D status and the periodontal disease measures taken over time. Garcia et al.¹⁴ conducted a one-year study of 51 patients with moderate-to-severe chronic periodontal disease attending a periodontal disease maintenance program. Patients who reported baseline use of calcium and vitamin D supplements compared to non-users had less periodontal disease (considering collectively a number of clinical measures) at baseline, six months and 12 months, although results were not statistically significant at 12 months. In a larger epidemiologic study of 550 men, Krall et al.¹⁵ found no association between self-reported baseline intake of vitamin D from foods and supplements and the seven-year progression in alveolar bone loss. In another study of 562 men, Alshouibi et al.¹⁷ used a repeated-measures cross-sectional design to examine associations between vitamin D intake and periodontal disease collected one to four times from 1986–1998. Vitamin D intake was associated with lower odds of moderate-to-severe periodontal disease. However, these studies did not consider sunlight exposure as a source of vitamin D when assessing vitamin D status. Vitamin D can be synthesized in the skin upon exposure to ultraviolet B radiation¹⁹. A recently published study by Jimenez et al.¹⁸ showed that higher predicted 25(OH)D concentrations were associated with a lower incidence of self-reported tooth loss in a 20-year prospective cohort of men. This study was limited by its use of a predictor score, as opposed to a direct measures of the biomarker 25(OH)D, to assess vitamin D status. The authors also relied on self-reported incident periodontal disease outcomes instead of clinical measures of disease progression. One randomized clinical trial of 145 men and women, nested within a larger clinical trial of bone loss of the hip, found that supplementation of vitamin D (700 IU/day) and calcium (500 mg/day) was associated with decreased odds of tooth loss over three years (27% of the placebo group vs. 13% of the supplemented group lost ≥ 1 tooth)¹⁶. However, this trial examined supplementation with both calcium and vitamin D together, limiting the ability to differentiate if both or just one nutrient was most influential.

The purpose of this manuscript is to build upon the currently published cross-sectional analyses in the OsteoPerio Study⁷ . We previously observed that adequate compared to inadequate or deficient vitamin D status was associated with reduced odds of periodontal disease ⁷, however we could not determine temporality from this cross-sectional study. We had the ability to examine prospective associations between baseline plasma 25(OH)D and the five-year change in periodontal disease measures inclusive of clinical attachment level (CAL), probing pocket depth (PD), ACH, and percent of gingival sites that bled upon assessment in the OsteoPerio Study, a well-defined cohort of postmenopausal women. The biomarker 25(OH)D reflects contribution from all three sources of vitamin D: diet, supplements, and sunlight. We hypothesized that baseline plasma 25(OH)D concentrations would be inversely associated with changes in periodontal disease measures, reflecting progression of disease over five years.

MATERIALS AND METHODS

Study Sample

There were 1,362 women who participated in the baseline OsteoPerio Study conducted from 1997–2000 as previously described⁷. Of these, 5 women were missing data on the study baseline questionnaires. An additional 39 women were missing data on one or more baseline periodontal disease measures (ACH [n=16], CAL [n=20], PD [n=19], gingival bleeding measures [n=13]). Of the remaining 1,318 women, baseline plasma sample collection was implemented after the study began and were available for 25(OH)D assays in 921 women. One additional woman was excluded because her baseline 25(OH)D was determined to be an extreme value (530 nmol/L). Of the remaining 920 women 675 (73%) attended the follow-up exam (2002–2005) and had follow-up periodontal disease measures needed to compute change in disease over time. Additionally, participants in this longitudinal sample were excluded if they were missing data on pertinent risk factors measured at baseline (education [n=10] and physical activity [n=10]). This left a final analytic sample of n=655. All participants signed informed consent. The study protocol was approved by the University at Buffalo’s Health Sciences Institutional Review Board and the study was conducted in accordance with the Helsinki Declaration of 1975, as revised in 2000.

Study Visit

These 655 study participants attended a clinic visit at baseline (1997–2000) and follow-up five years later (2002–2005). Prior to the study visits, participants completed questionnaires to assess their demographic information, family and medical history, lifestyle habits including physical activity, osteoporosis risk factors, oral health history, and current medication and supplement use. At the clinic visit, questionnaires were reviewed for completeness, physical measurements were taken to assess weight and height, and participants underwent dual-energy x-ray absorptiometry (DXA)^§ to determine systemic bone density.

Oral Health Exam Outcome Measures at Baseline and Follow-up

Full-mouth oral clinical examinations were conducted by trained and calibrated dental examiners using standardized protocols, the details of which were described previously^{7, 20, 21}. Measures of ACH, PD, CAL, and gingival bleeding were assessed at baseline and follow-up and described in detail elsewhere^{7, 20–22}. ACH, in mm, was assessed using standardized intraoral radiographs^‖ and measured mesially and distally for each tooth present (excluding third molars and canines) for a maximum of 48 sites. The average ACH of all sites measured is referred to as the whole-mouth mean ACH. Using whole-mouth mean ACH levels and self-reported tooth loss to periodontitis, participants were grouped into three categories of periodontal disease: none, mild/moderate, or severe²⁰. A constant-force electronic periodontal probing system^¶ was used to measure PD, in mm, on six surfaces of each tooth and the corresponding CAL, in mm, was determined for each surface site assessed with the use of a manual periodontal probe^**. Measures of whole-mouth mean PD and CAL level were computed. Women were also categorized as having none/mild, moderate or severe periodontal disease using the CDC/AAP Working Group definition²³. Gingival bleeding was assessed by inserting a periodontal probe^δ approximately two mm into the gingival sulcus/pocket at three gingival sites per tooth, except third molars. Each site was scored either a 0 (absence) or 1 (presence) for bleeding. The mean of all gingival bleeding scores was computed representing the proportion of all sites assessed that bled in the mouth (when multiplied by 100 this represents the percent of bleeding sites in a mouth).

Measure of Five-Year Change in Periodontal Disease

Using the periodontal disease measures taken at baseline and follow-up, we defined continuous measures of five-year change in periodontal disease (follow-up value - baseline value), with positive values representing loss in alveolar bone and clinical attachment, worsening of PD over time, or a greater percentage of gingival sites that bled over time (Table 1). In every case, change was computed from all available five-year measurements and their corresponding measurements at baseline. For ACH, the change at each site was determined using overlayed oral radiographs taken at the two time periods. Change measures from all sites were averaged and this is referred to as the ACH side-by-side mean change. This method was developed by Hausmann et al.²⁴ and described for use in this study by LaMonte et al.²¹ The whole-mouth mean change for CAL and PD and the change in the percent of all gingival sites in the mouth that bled upon assessment were determined by subtracting the whole-mouth mean CAL/PD or the percentage of bleeding sites measures at baseline from follow-up.

Table 1.

Change in periodontal disease measures from baseline (1997–2000) to follow-up (2002–2005) among 655 participants in the Buffalo OsteoPerio Study who had baseline plasma 25(OH)D concentrations and attended the follow-up exam

Periodontal Disease Change Measure	Mean	SD	Median	Min	Max
ACH side-by-side mean change (mm)*	0.18	0.22	0.16	− 0.45	2.90
CAL whole-mouth mean change (mm)*	− 0.17	0.50	− 0.19	− 1.59	3.23
PD whole-mouth mean change (mm)*	− 0.15	0.39	− 0.15	− 1.88	2.14
Change in the percent of gingival sites that bled upon assessment*	− 21.8	24.0	− 19.6	− 89.7	77.3

^*Change measures were calculated by subtracting baseline periodontal disease measures from follow-up periodontal disease measures with positive values indicating a greater loss in alveolar bone, clinical attachment, or worsening of PD over time or a greater percent of gingival sites that bled over time.

Assessment of Plasma 25-hydroxyvitamin D

Blood draws were conducted on the same day as the clinical periodontal examinations. Assessment of plasma 25(OH)D at baseline and follow-up has been previously reported²⁵. Assays were conducted over a consistent four-month period using a competitive chemiluminescence immunoassay^††. Using the investigators’ blinded, duplicate quality control samples nested in each batch, the within pair coefficient of variation was 4.9%. Only a subset of our sample (n=628) had 25(OH)D measures at follow-up. For both baseline and follow-up 25(OH)D measures, we adjusted values for season of blood draw using the residual method. Residuals were computed from regression of 25(OH)D on day of the year of blood draw and added back to sample mean as previously described⁷.

Statistical Analysis

Data were analyzed with statistical software^‡‡. In order to better understand how loss to follow-up may have influenced our results, we compared characteristics of study participants included in the current prospective analysis (n=655) to study participants who had baseline plasma 25(OH)D concentrations but were not included in our current analyses because they did not attend follow-up, did not have periodontal disease measures needed to compute change measures, or were missing data on pertinent covariates (n=265). Next, in our current analytic sample of 655 women, summary measures of periodontal change variables were described (Table 1) and mean baseline plasma 25(OH)D was described according to participant characteristics and periodontal disease risk factors (Table 2). Differences in means were examined using analysis of variance (ANOVA) and differences between categorical variables were examined using χ² tests. Tests were considered statistically significant with p-values of <0.05 (two-sided).

Table 2.

Mean and standard deviation (SD) of baseline plasma 25(OH)D concentrations by baseline characteristics: the Buffalo OsteoPerio Study (n=655)

Baseline characteristic	Baseline plasma 25(OH)D
	N	Mean (SD)	p-value*
Overall sample	655	60.64 (21.9)	---
Categories of vitamin D status in 25(OH)D (nmol/L)
Deficient (<30)	47	22.94 (1.3)	<0.0001
Inadequate (≥30 to <50)	159	41.36 (0.7)
Adequate (≥50 to <75)	300	61.74 (0.5)
Adequate (≥75)	149	90.88 (0.7)
ACH definition of periodontal disease20
None	175	61.93 (22.4)	0.367
Mild/moderate	335	59.45 (20.7)
Severe	145	61.82 (24.1)
CDC/AAP definition of periodontal disease23
None/mild	145	62.98 (23.6)	0.297
Moderate	404	59.70 (21.3)
Severe	106	61.01 (21.9)
Percent of gingival sites that bled at baseline (%) in tertiles
Tertile 1: (0–22.2)	223	61.51 (22.3)	0.004
Tertile 2: (22.6–45.0)	214	63.64 (22.0)
Tertile 3: (45.1–100)	218	56.80 (20.9)
Number of teeth at baseline
6 to 14	51	53.46 (24.8)	0.038
15 to 24	227	62.13 (22.7)
24 to 28	377	60.71 (20.9)
Age (years)
< 60	143	64.30 (22.3)	0.018
60 – < 70	322	60.88 (22.3)
≥ 70	190	57.47 (20.6)
Race
White	646	60.66 (21.8)	0.806
Other	9	58.85 (28.5)
Education
≤ High school diploma or GED†certificate	134	58.92 (22.8)	0.310
School after high school	521	61.08 (21.7)
Cigarette Smoking Status
Never	364	60.25 (22.2)	0.598
Former	276	61.38 (21.1)
Current	15	56.25 (30.0)
Waist circumference (cm) in tertiles‡
Tertile 1 ( 61 – 78)	223	68.47 (22.3)	<0.0001
Tertile 2 (79 – 87)	214	60.33 (20.4)
Tertile 3 (88 – 135)	209	52.92 (20.1)
Waist to hip ratio in tertiles‡
Tertile 1 (0.65 – 0.77)	214	66.66 (22.5)	<0.0001
Tertile 2 (0.77 – 0.82)	216	58.43 (18.9)
Tertile 3 (0.82 – 1.20)	215	57.29 (22.9)
Body Mass Index (kg/m2)
Underweight or Normal (< 25)	289	67.34 (22.3)	<0.0001
Overweight (25 – < 30)	233	58.17 (20.9)
Obese (≥ 30)	133	50.39 (17.7)
Recreational physical activity (METδ-hrs/week)
None	97	53.71 (21.9)	<0.0001
< 12.5	275	58.17 (21.1)
≥ 12.5	283	65.40 (21.7)
Hormone therapy use‡
Never	194	58.25 (23.1)	0.035
Past only	116	58.17 (19.9)
Current	338	62.64 (21.4)
Osteoporosis related medication use or bone therapies‖
No	288	57.85 (21.6)	0.004
Yes	367	62.82 (22.0)
Self-reported history of osteoporosis
No	573	60.17 (21.8)	0.146
Yes	82	63.93 (22.7)
Worst site T score
Normal	118	61.26 (21.4)	0.019
Low bone density	300	62.85 (21.2)
Osteoporosis	237	57.52 (22.8)
Self-reported history of diabetes
No	632	61.06 (22.0)	0.009
Yes	23	48.90 (15.8)
Days since the last dental cleaning‡
≤ 90	254	61.27 (21.4)	0.567
>90	386	60.26 (22.4)
Frequency of dental visits
Never or only with a problem	50	52.47 (20.0)	0.017
Once a year	86	59.66 (20.5)
More than once a year	519	61.58 (22.2)
Frequency of flossing teeth‡
Not every week	116	61.51 (23.5)	0.083
Once a week	62	62.80 (17.6)
More than once a week	189	57.12 (23.0)
Everyday	287	61.90 (21.0)

^*P-values from analysis of variance of mean baseline plasma 25(OH)D concentrations across level of baseline characteristics.

^†GED, Generalized educational development.

^‡Sample size does not add up to 655 due to missing data on this variable.

^δMET, Metabolic equivalent task.

^‖Current use of osteoporosis related medications or bone therapies include women taking current hormone therapy drugs, bone drugs (miacalcin, alendronate), or selective estrogen receptor modulator (SERM) drugs (raloxifene).

Scatterplots of periodontal disease change measures on baseline 25(OH)D concentrations were examined and assumptions for statistical methods were checked and met. In the full analytic sample (n=655), linear regression was used to regress periodontal disease change measures on baseline 25(OH)D (Table 3). Regression coefficients, standard errors (SE), and affiliated p-values for the baseline 25(OH)D per each 10 nmol/L increment are presented. Regression coefficients represent either a 1 mm change in ACH, CAL or PD, or 1 unit change in the percent of gingival sites that bled for a 10 nmol/L difference in 25(OH)D. We examined whether adjustment for common periodontal disease risk factors, as noted in the literature (see Table 2), confounded our regression models. We used a forward selection process where confounder selection was based on confounders that changed the beta-coefficient at least 10%. Multivariable regression models were adjusted for the following covariates assessed at baseline age, education, frequency of dental visits, smoking status, self-reported history of diabetes, current use of osteoporosis related medications or bone therapies (e.g., current use of hormone therapy drugs, bone drugs [miacalcin, alendronate], or selective estrogen receptor modulator [SERM] drugs [raloxifene]), and baseline measures of periodontal disease. Models examining ACH side-by-side change were also adjusted for baseline whole-mouth mean ACH. Models examining CAL or PD whole-mouth mean change were also adjusted for baseline whole-mouth mean CAL or PD, respectively; and models examining change in the percent of all gingival sites in the mouth that bled were also adjusted for the baseline measure of percent of gingival sites that bled. Adjustment for baseline measures of periodontal disease helps to differentiate between cross-sectional and longitudinal relationships. Tests were considered statistically significant with p-values of <0.05 (two-sided).

Table 3.

β-coefficients ± standard errors (SE) for changes in periodontal disease measures corresponding to a 10 nmol/L difference in baseline 25(OH) D concentrations (nmol/L) among all women with baseline 25(OH)D measures (n=655) and limited to women with stable 25(OH)D concentrations over time (n=442): the Buffalo OsteoPerio Study

	Baseline 25(OH)D		Baseline 25(OH)D Among articipants who changed less than ± 20 nmol/L in five-years
	n=655		n=442
	β-coefficient (SE)	P-value*	β-coefficient (SE)	P-value*
ACH Side-by-side mean change (mm)
Age-adjusted Model	− 0.001 (0.004)	0.724	−0.003 (0.005)	0.579
Multivariable Model 1†δ	−0.003 (0.004)	0.474	−0.005 (0.006)	0.397
Multivariable Model 2‡δ	−0.002 (0.004)	0.593	−0.004 (0.006)	0.530
CAL Whole-mouth mean change (mm)
Age-adjusted Model	− 0.001 (0.009)	0.889	0.003 (0.012)	0.788
Multivariable Model 1†	0.001 (0.008)	0.905	0.005 (0.012)	0.686
Multivariable Model 2‡	0.001 (0.009)	0.882	0.008 (0.012)	0.501
PD Whole-mouth mean change (mm)
Age-adjusted Model	0.009 (0.007)	0.214	0.013 (0.009)	0.157
Multivariable Model 1†	0.0009 (0.006)	0.883	0.006 (0.008)	0.432
Multivariable Model 2‡	0.0005 (0.006)	0.930	0.007 (0.008)	0.429
Change in the percent of gingival sites that bled upon assessment (%)
Age-adjusted Model	0.864 (0.433)	0.046	0.389 (0.572)	0.497
Multivariable Model 1†	0.034 (0.281)	0.903	−0.209 (0.366)	0.568
Multivariable Model 2‡	0.112 (0.295)	0.704	−0.155 (0.382)	0.685

^*P-value for associated beta-coefficient.

^†Model 1 is adjusted for age, education, frequency of dental visits, smoking status, diabetes status, and current use of osteoporosis related medications or bone therapies, and baseline periodontal disease measure.

^‡Model 2 is equivalent to Model 1, but further adjusted for BMI and recreational physical activity.

^δAnalyses for baseline 25(OH)D include only 651 women and analyses for stable 25(OH)D include only 439 women because some women’s cement-enamel junction were not visible on their baseline radiograph to determine baseline whole-mouth mean ACH.

The multivariable model is also shown with and without further adjustment for body mass index (BMI) and self-reported recreational physical activity, both of which are strong predictors of vitamin D status^{26, 27}. Inclusion of these variables in the model will explain variation in 25(OH)D and could result in over-adjustment, therefore we presented the results with and without their inclusion in the model. We also examined these analyses stratified by baseline periodontal disease status for ACH, CAL and PD change measures and stratified by tertiles of baseline measures of percent of gingival sites that bled upon assessment (Table 4). P-values for interaction between baseline 25(OH)D and baseline periodontal disease status were estimated by addition of an interaction term to the logistic regression model. P-values <0.20 (two sided) for interactions were considered statistically significant.

Table 4.

Adjusted^* β-coefficient ± standard errors (SE) for changes in periodontal disease measures corresponding to a 10 nmol/L difference in baseline 25(OH)D concentrations (nmol/L) stratified by baseline periodontal disease status^† (n=655): the Buffalo OsteoPerio Study

	n	β-coefficient (SE)	P-value‡
ACH Side-by-side mean change (mm)δ
None	175	0.003 (0.004)	0.473
Moderate	332	− 0.0009 (0.005)	0.857
Severe	144	− 0.012 (0.014)	0.385
P for interaction			0.319
CAL Whole-mouth mean change (mm)
None/Mild	145	− 0.014 (0.013)	0.286
Moderate	404	0.010 (0.011)	0.354
Severe	106	− 0.037 (0.030)	0.219
P for interaction			0.317
PD Whole-mouth mean change (mm)
None/Mild	145	− 0.007 (0.010)	0.531
Moderate	404	0.005 (0.007)	0.517
Severe	106	− 0.026 (0.024)	0.291
P for interaction			0.586
Change in the percent of gingival sites that bled upon assessment (%)
Tertile 1: Percent of gingival sites that bled at baseline (0–22.2)	223	−0.468 (0.431)	0.279
Tertile 2: Percent of gingival sites that bled at baseline (22.6–45.0)	214	0.516 (0.509)	0.312
Tertile 3: Percent of gingival sites that bled at baseline (45.1–100)	218	0.417 (0.617)	0.500
P for interaction			0.542

^*Models are adjusted for age, education, frequency of dental visits, smoking status, diabetes status, current use of osteoporosis related medications or bone therapies, baseline periodontal disease measures, BMI and recreational physical activity.

^†Periodontal disease status is defined using the whole-mouth mean ACH and self-reported tooth loss due to periodontitis20 when analyzing the ACH side-by-side mean change measures. Periodontal disease status is defined using CDC/AAP defined categories23 when analyzing the CAL and PD whole-mouth mean change measures. Change in the percent of gingival sites that bled upon assessment is stratified by tertiles of percent of gingival sites that bled at baseline.

^‡P-value for associated beta-coefficient.

^δAnalyses include only 651 women because four women’s cement-enamel junctions were not visible on their baseline radiograph to determine baseline whole-mouth mean ACH.

Further, sensitivity analyses were conducted to explore the effect of teeth lost between baseline and follow-up. Lost teeth resulted in missing measurements that could lead to biased estimates of periodontal disease change. To address this concern, we conducted a sensitivity analysis by imputing a wide range of plausible values (from 2 to 10 mm of change) to estimate the extent tooth loss may have influenced results for whole-mouth mean change values.

Further exploratory analyses using linear regression were also conducted to examine the association between periodontal disease change measures and baseline 25(OH)D among participants with stable vitamin D status over time (those whose 25(OH)D changed less than ± 20 nmol/L [n=422]). An increase in 25(OH)D concentrations of approximately 20 nmol/L, in the absence of sunlight, requires a substantial increase in vitamin D intake per day (~1,000 IU/day in supplementation) ^{28, 29}. As two-thirds of our sample did not increase or decrease their 25(OH)D concentrations beyond 20 nmol/L, we considered women outside this range (± 20 nmol/L) to have greatly changed their vitamin D status over time.

RESULTS

We examined characteristics of women included (n=655) and excluded (n=265) from these analyses. Women excluded had slightly lower mean 25(OH)D concentrations (mean [SD] =58.22 [24.4] nmol/L) than women included (60.64 [21.9] nmol/L), although this difference was not statistically significant (p-value=0.14). A greater percentage of excluded women had severe periodontal disease based on measures of ACH and tooth loss due to periodontal disease (30.2% vs. 22.1%, p-value=0.03) and the CDC/AAP definition of periodontal disease (20.0% vs. 16.2%, p-value=0.25), although these differences were only statistically significant for disease based on ACH and tooth loss. Women excluded versus included had fewer teeth at baseline (mean [SD]=22.48 [5.5) vs. 23.72 [5.0], p-value=0.001), were older (68.91 [7.5] vs. 65.60 [6.6], p-value<0.0001), more likely to be a race other than Caucasian (4.9% vs. 1.4%, p-value=0.002), more likely to be current smokers at baseline (5.3% vs. 2.3%, p-value=0.01), more likely to have osteoporosis at baseline (42.6% vs. 36.2%, p-value=0.03) and self-reported diabetes at baseline (6.8% vs. 3.5%, p-value=0.03). Statistically significant differences with respect to baseline measures of percent of gingival sites that bled, education, BMI, waist circumference, recreational physical activity, hormone therapy use, current use of osteoporosis related medications or bone therapies, and measures of dental hygiene were not observed (data not shown).

Summaries of the changes in periodontal disease measures over 5 years are shown in Table 1. The mean ACH side-by-side change indicates loss, on average, in alveolar bone over time. For the other periodontal disease change measures based on CAL, PD and gingival bleeding, mean changes were negative suggesting that, on average, there was a slight improvement in these measures over time.

At baseline, 25(OH)D concentrations ranged from 5.91 to 146.01 nmol/L with 7% (n=47) of the sample having deficient (25(OH)D <30 nmol/L) and 24% (n=159) having inadequate (25(OH)D <50 nmol/L) vitamin D status³⁰ (Table 2). Mean baseline 25(OH)D concentrations did not vary significantly by baseline periodontal disease status as defined using the ACH or CDC/AAP categorical definitions. Twenty-two percent of women were defined as having none/mild and 16% with severe periodontal disease at baseline based on the CDC/AAP definition. The majority of our sample (62%) had moderate disease. Mean baseline 25(OH)D concentrations were lower among women with a greater percentage of gingival sites that bled at baseline (tertile 3) compared to women with fewer sites that bled (tertile 1) and among women with fewer compared to more teeth at baseline. Mean 25(OH)D concentrations were lower for older women (≥ 70 years) and women with greater waist circumferences, waist-to-hip ratios, and BMIs and women who self-reported not engaging in recreational physical activity. Mean concentrations were also lower for women who reported never using hormone therapy or using it in the past and concentrations were lower in women who did not take osteoporosis related medications or bone therapies. Women who had DXA-defined osteoporosis or self-reported a history of diabetes also had lower 25(OH)D concentrations as did women who never frequented the dentist or only went with a problem.

Table 3 shows regression coefficients and standard errors (SE) for change in periodontal disease measures regressed on baseline 25(OH)D. In the age-adjusted model, only change in the percent of gingival sites that bled upon assessment was associated with 25(OH)D status. A 10 nmol/L greater baseline 25(OH)D concentration was associated with ~0.9 percentage points more gingival sites that bled at follow-up than at baseline (p-value=0.046). Adjustment for additional covariates attenuated this association (Model 1 β-coefficient (SE) = 0.034 (0.281), p-value=0.903). There were no other statistically significant associations observed between baseline 25(OH)D and change in any other periodontal disease measure.

Among the 655 women in these analyses, 6.7% (n=24) self-reported loss of at least one tooth due to periodontal disease. Imputing 2 mm, 5 mm or 10 mm for ACH, CAL and PD change measures for teeth lost due to periodontal disease over follow-up did not greatly influence the results. The 25(OH)D β-coefficients (SE) and p-values for Model 2 with a 10 mm imputation were −0.007 (0.008), p-value=0.381 for ACH, 0.004 (0.011), p-value=0.722 for CAL and 0.002 (0.009), p-value=0.803 for PD. Table 3 also shows exploratory analyses where we restricted the analyses of baseline 25(OH)D to include only participants whose vitamin D status was stable over time (defined as changed less than ± 20 nmol/L from baseline to follow-up). This slightly strengthened the beta-coefficients when examining change measures for ACH, CAL, and PD. The beta-coefficient in the model examining change in the percent of gingival sites that bled became negative. Even so, vitamin D status remained unrelated to periodontal disease progression.

Table 4 shows associations between baseline 25(OH)D and periodontal disease change measures stratified by baseline periodontal disease status. Among women with severe periodontal disease at baseline, greater concentrations of baseline 25(OH)D were consistently associated with less periodontal disease progression defined using measures of ACH, CAL and PD, but these associations were not statistically significant.

DISCUSSION

This is the largest prospective study to date on vitamin D status and progression of periodontal disease in postmenopausal women. We observed no associations between vitamin D status, assessed with baseline 25(OH)D concentrations, and the subsequent five-year change in periodontal disease measures inclusive of changes in ACH, CAL, PD and gingival bleeding. These results did not vary greatly by baseline periodontal disease status. There was some suggestion that among women with severe periodontal disease at baseline, higher baseline 25(OH)D concentrations may protect against periodontal disease progression defined using ACH, CAL and PD, but these results were not statistically significant. We cannot say for certain that the results differed by baseline disease severity. Additionally, these results do not support our previous cross-sectional findings in the same cohort of women⁷ where we observed that adequate compared to inadequate or deficient vitamin D status was associated with a decreased odds of periodontal disease.

Among our 655 participants, 16.2% and 61.7% of our sample had severe and moderate periodontal disease, respectively, at baseline (1997–2000). This is comparable to recent estimates of 11.7% (50–64 years) and 11.2% (65+ years) of U.S. adults with severe disease and 37.7% (50–64 years) and 53.0% (65+ years) of U.S. adults with moderate disease as reported using nationally representative data¹. Although our baseline prevalence estimates of periodontal disease are comparable to national prevalence estimates, previous data from our cohort showed small changes in periodontal disease measures over five years²¹. We speculated this was attributable to a low prevalence of periodontal disease risk factors (e.g., minimal current smokers, low prevalence of diabetes, etc.) in this cohort of postmenopausal women²¹. It is possible our null results are explained by the minimal progression of disease over five years (e.g., 0.18 mm loss in alveolar bone on average). Perhaps a longer follow-up period is needed to observe meaningful changes in disease status. A previous report of disease progression over 10 years cites 0.7 to 1.4 mm of alveolar bone loss among individuals 25–65 years at baseline and 2.8 mm among those 60–70 years.³¹

Our null results are not likely explained by a lack of variation in vitamin D status. In our sample seven percent of women demonstrated deficient baseline vitamin D status (i.e., 25(OH)D <30 nmol/L). This is similar to the three to five percent of older women reported to have deficient vitamin D status (defined as 25(OH)D ς5 nmol/L) in a nationally representative dataset³² in which associations between vitamin D status and clinical measures of periodontal disease have been observed^{10, 11}.

Two previously conducted studies^{14, 17} examined associations between vitamin D intake and repeat measures of periodontal disease over time. Garcia et al.¹⁴ conducted a 12 month study among 51 patients in a periodontal disease maintenance program consisting of postmenopausal women and men 50 to 80 years. They showed that supplement users, defined as users of both calcium (≥ 1,000 mg/day) and vitamin D (≥ 400 IU/day) supplements for > 18 months at baseline, had borderline statistically significant (p=0.058) better periodontal disease measures (assessed collectively as attachment loss, bleeding on probing, gingival index, pocket depth and furcation involvement) at 12 months than non-supplement users. Non-users did not use vitamin D or calcium supplements and had low dietary intake of these nutrients. Supplement users, although not different with respect to ACH from non-users, had denser oral bone at six and 12 months. Follow-up in this study design was short and supplement users of calcium were not examined differently from users of vitamin D. In the Department of Veterans Affairs Dental Longitudinal Study of men (mean age 62.9 years), vitamin D intake of ≥ 800 compared to < 400 IU/day was associated with a 33% and 46% decreased odds of severe periodontal disease defined with measures of CAL and PD and moderate-to-severe disease defined using measures of alveolar bone loss, respectively¹⁷. Data collected on diet and periodontal disease one to four times from 1986–1998 was used in a repeated-measures cross-sectional design. Neither of these studies examined progression of disease over time in relation to vitamin D intake.

In a different study conducted by Krall et al.¹⁵, in men aged ≥ 65 years, high compared to low dietary plus supplemental calcium, but not vitamin D intake, was shown to protect against the seven-year progression of periodontal disease as defined by alveolar bone loss. It is possible that the null results with vitamin D intake were explained in part by misclassification of vitamin D status when using the measure of oral vitamin D intake²⁶, or these results support what we observed, that vitamin D status is not associated with changes in alveolar bone loss. It is also possible that vitamin D intake is influential, but only among persons with low sun exposure, which was not explored in this study. Differently, Jimenez et al.¹⁸ observed a lower incidence of self-reported tooth loss and periodontitis in a 20-year prospective cohort of 42,730 of male health professionals with high compared to low predicted 25(OH)D concentrations. The multivariable model was adjusted for age, smoking, pipe use, chewing tobacco use, multivitamin use, vitamin E, vitamin C, dental profession, alcohol consumption, routine physical examination and diabetes status. After further adjustment for BMI, race and physical activity (factors used to develop the 25(OH)D predictor score) the periodontitis, but not the tooth loss, association was no longer statistically significant. This model may have been over-adjusted, or as the authors suggest, self-reported periodontitis, as compared to self-reported tooth loss, may be more susceptible to misclassification and led to the attenuation of the observed association. The predictor score may also reflect a healthy lifestyle score rather than a true indicator of vitamin D status. There has been one, three-year randomized clinical trial of combined vitamin D and calcium supplementation on tooth loss that showed a protective effect¹⁶. Unfortunately this study cannot differentiate the effect of one nutrient from another. It is also possible that the difference between Krall et al.’s and Jimenez et al.’s study findings compared to others could be explained by the outcome of tooth loss, which is representative of end stage periodontal disease as compared to incident periodontitis or changes in PD, CAL, ACH, and gingival bleeding.

Whether our results are generalizable to other populations remains unknown. Our study consists of well-educated, primarily Caucasian women. It is possible that loss of participants due to follow-up may have contributed to our null results as women who were excluded from the current analyses had slightly lower 25(OH)D concentrations and more severe periodontal disease. It is also questionable whether our follow-up period was long enough; however, this study is still the largest and longest prospective analysis of vitamin D status and periodontal disease progression in postmenopausal women. Our participants all had standardized full mouth oral exams and oral radiographs at baseline and follow-up. We were able to assess change in a varied set of periodontal disease measures using matched sites at both time points.

In conclusion, our study results suggest that baseline plasma vitamin D status does not influence the subsequent five-year change in chronic periodontal disease measures in postmenopausal women. Thus, supplementation of vitamin D for prevention of periodontal disease progression is not warranted at this time. Further replication of these findings is needed. We recommend that these associations be examined in studies with a longer follow-up period of periodontal disease progression.