Abstract
Background
Associations between vitamin D biomarkers and persistent high-risk human papillomavirus (hrHPV) detection have not been evaluated.
Methods
2011–2012 stored sera from 72 women aged 30–50 years with prevalent hrHPV (n = 116 type-specific infections) were tested for 5 vitamin D biomarkers: 25(OH)D and 4 emerging biomarkers, 1,25(OH)2D; 24,25(OH)2D; free vitamin D; and vitamin D binding protein (DBP). The hrHPV detection patterns (persistent vs transient/sporadic) were determined using cervicovaginal swabs collected monthly for 6 months. Associations between vitamin D and short-term type-specific hrHPV persistence were estimated using logistic regression. Our primary exposure was continuous 25(OH)D, with additional biomarkers evaluated as secondary exposures. Primary models were adjusted for age, race, body mass index, education, contraceptives, smoking, season, and calcium/phosphate levels. Sensitivity analyses were restricted from 19 hrHPV types to 14 used in cervical cancer screening.
Results
In primary analyses, nonsignificant positive associations with hrHPV persistence were observed for measures of 25(OH)D and 24,25(OH)2D. Associations were stronger and significant when restricting to 14 hrHPV types (25(OH)D per 10 ng/mL increase: adjusted odds ratio [aOR], 1.82 [95% confidence interval {CI}, 1.15–2.88] and aOR, 4.19 [95% CI, 1.18–14.88] DBP-adjusted; 25(OH)D ≥30 vs <30 ng/mL: aOR, 8.85 [95% CI, 2.69–29.06]; 24,25(OH)2D: aOR, 1.85 [95% CI, 1.18–2.88]). 1,25(OH)2D was unassociated with persistence.
Conclusions
Serum vitamin D measured by multiple biomarkers showed positive associations with short-term hrHPV persistence that were significant only when restricting to 14 clinically relevant hrHPV types.
Keywords: vitamin D, persistence, mid-adult, 25-dihydroxyvitamin-D, HPV
Serum vitamin D measured by multiple biomarkers was positively associated with short-term high-risk human papillomavirus (hrHPV) persistence, but associations were inconsistent and significant only when restricting to hrHPV types included in cervical cancer screening assays.
High-risk human papillomavirus (hrHPV) infections are a major cause of infection-related cancer globally [1]. Most hrHPV infections are transient, and approximately 90% of new infections become undetectable within 2 years [2]. Infections that persist, however, are more likely to progress to cervical cancer [3]. Up to 10% of women aged ≥30 years are estimated to carry persistent HPV infections, and these women are considered at highest risk for the development of cervical cancer [4]. While HPV infection characteristics such as genotype [5, 6] and viral load [7] are established risk factors for persistence, other cofactors are less well defined. Better understanding of modifiable risk factors that contribute to persistent detection of hrHPV may inform efforts to prevent carcinogenic progression.
Vitamin D sufficiency is associated with improved health outcomes, including lower incidence of certain cancers [8] and reduced risk and severity of respiratory and influenza-like infections [9], yet 40% of adults in the United States are vitamin D deficient [10]. Vitamin D can modulate immune responses, with lower serum 25(OH)D concentrations contributing to reduced innate immune function and increased susceptibility to infection [11, 12]. Vitamin D deficiency may limit production of antimicrobial peptides (AMPs) [13], which, when produced in the endothelial cells in the vagina and cervix, defend against sexually transmitted infections, including hrHPV [14, 15]. The virus’s established ability to evade the adaptive immune response [16] makes HPV-prevention strategies that address innate immunity (bolstered by optimal AMP production and activity) particularly appealing.
Most epidemiologic studies have assessed vitamin D status as serum concentrations of 25-hydroxyvitamin D, or 25(OH)D, but this biomarker is only 1 of several generated throughout the metabolism of vitamin D [17], and clinical cut-points of sufficiency remain contested [18]. Additional biomarkers may describe vitamin D bioavailability and activity more fully and thus more appropriately characterize an individual’s vitamin D status [19], particularly because 25(OH)D does not appear to be consistently associated with health outcomes [20]. These other biomarkers include calcitriol, the hormonally active metabolite of vitamin D otherwise known as 1,25(OH)2D; 24,25(OH)2D, the most concentrated product of 25(OH)D catabolism; vitamin D binding protein (DBP); free 25(OH)D (circulating unbound to DBP or albumin); and bioavailable 25(OH)D (unbound to DBP) [17, 19]. However, these newer biomarkers have not been factored into vitamin D recommendations [21], in part because of a lack of established cut-points indicating sufficiency.
To our knowledge, only cross-sectional assessments of the relationship between serum 25(OH)D and detection of hrHPV have been conducted [22–24]. Investigators evaluating 2003–2006 National Health and Nutrition Examination Survey data determined that 25(OH)D serum concentrations were inversely associated with detection of prevalent cervicovaginal hrHPV [22]. Analyses of 2009–2014 data found higher likelihood of detection of cervical and oral HPV among women with vitamin D concentrations <20 ng/mL vs ≥20 ng/mL, though there was no significant association between vitamin D and oral or penile HPV infection among men [23]. Our cross-sectional analysis of mid-adult women in Washington State yielded inconclusive results [24]. Using longitudinal data from the same study, we aimed to establish whether lower serum concentrations of 5 different vitamin D biomarkers are associated with higher likelihood of short-term persistent (vs transient or sporadic) detection of hrHPV. Prior research suggests that short-term persistence may correlate with long-term persistence and cervical cancer risk [25].
METHODS
Study Population
This secondary analysis utilized stored specimens and data from a 6-month longitudinal cohort study of HPV infections in healthy women 30–50 years of age in Seattle, Washington. Details of the parent study have been described previously [26]; in brief, mid-adult women affiliated with the University of Washington (faculty, staff, or student) were enrolled between March 2011 and January 2012. Women who were pregnant, had a history of hysterectomy, or had a serious medical condition that would preclude participation were excluded. All subjects provided written informed consent and the protocol was reviewed and approved by the University of Washington Institutional Review Board.
Data Collection
Upon enrollment, subjects received verbal and written instruction [27] from the study coordinator for self-collection of cervicovaginal swabs for HPV DNA testing into 1.5 mL of specimen transport medium [26]. Subjects provided specimens at enrollment and monthly for up to 6 months. Venous blood specimens were collected at enrollment and exit visits, and residual serum specimens were stored at –80°C for future testing. Subjects self-reported demographic (age, race/ethnicity, highest level of education, marital status), health (alcohol intake, smoking status, HPV vaccination history, oral contraceptive use), and sexual behavior characteristics (history of pregnancy, history of genital warts and/or other sexually transmitted infections, and number and recency of male sex partners) at enrollment. Participants’ weight and height were collected and body mass index (BMI [kg/m2]) was calculated. Information regarding vitamin D supplementation was not determined.
Laboratory Methods
Vitamin D Biomarker Testing
Serum concentrations of albumin, calcium, and phosphate were determined using an automated clinical chemistry analyzer (Beckman AU5812, Beckman Coulter, Brea, California). Concentrations of 1,25(OH)2D3, 1,25(OH)2D2, 24,25(OH)2D3, 25(OH)D3, and 25(OH)D2 were determined using immunoaffinity enrichment liquid chromatography–tandem mass spectrometry (LC-MS/MS), as previously described [28, 29]. Haplotypes and concentrations of vitamin D binding were determined using trypsin digestion LC-MS/MS, as previously described [30, 31].
HPV Genotyping
Self-collected vaginal swabs were directly genotyped as part of the parent study using the polymerase chain reaction–based Roche Linear Assay for 37 α-genus HPV types (Roche Diagnostics, Pleasanton, California) [26].
Exposure Definitions
Our primary exposure was serum concentration of 25(OH)D at baseline. As secondary exposures, we measured serum concentrations at baseline of 1,25(OH)2D, 24,25(OH)2D, free/unbound vitamin D, and total vitamin D adjusted for DBP and stratified by DBP haplotype. 25(OH)D and 1,25(OH)2D are composite measures and are calculated as the sums of 25(OH)D2 and 25(OH)D3, and 1,25(OH)2D2 and 1,25(OH)2D3, respectively. Free vitamin D was calculated using an established equation that accounts for affinity constants for albumin and DBP and, due to its presence in low concentrations in our population, was measured in (1000) × nmol/L.
We modeled our primary exposure as a continuous 10 ng/mL increase of serum 25(OH)D to aid in both interpretability and comparison to previous literature [22]. To assess clinically meaningful cut-points of total vitamin D sufficiency in secondary analyses, we generated 2 binary exposure variables, evaluating sufficiency as ≥20 ng/mL per recommendations from the Institute of Medicine [32], and then again, as ≥30 ng/mL, as suggested by the Endocrine Society [33]. Finally, the 4 novel biomarkers under examination were analyzed as continuous variables.
Given the potential clinical relevance of evaluating patterns of hrHPV persistence among women who supplement with vitamin D, we generated, post hoc, a cut-point of 50 ng/mL of 25(OH)D at baseline to compare women who likely supplement to those who likely do not [34].
Outcome Definitions
Analyses were restricted to 19 hrHPV types (16/18/26/31/33/35/39/45/51/52/53/56/58/59/66/68/73/82/IS39) [35, 36]. At enrollment, each hrHPV type within a woman (hereafter referred to as a “woman-type”) was classified as negative or positive; woman-types that were DNA negative at baseline were excluded from analyses.
Four consecutive type-specific positive monthly vaginal samples or 6 type-specific positives out of 7 total samples were classified as persistent type-specific hrHPV detection, our primary outcome of interest. Alternatively, a single type-specific positive sample at baseline (with subsequent negative samples for the same type) was classified as transient detection. Finally, detection patterns with >1 type-specific positive that did not meet the definition for persistent detection were classified as sporadic detection. To maximize our power to detect differences between infections that persisted and those that did not persist, we generated a dichotomous outcome variable by combining transient and sporadic detection.
Statistical Analysis
Analyses were restricted to women with a positive result for at least 1 of 19 hrHPV types at baseline and who completed 6 or more visits. Women could contribute up to 19 type-specific infections to analyses. We used logistic regression to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for associations between persistent vs transient/sporadic detection and women’s demographic and behavioral characteristics, and for associations between persistent vs transient/sporadic detection and vitamin D serum concentrations. We used generalized estimating equations with an independence working correlation structure to account for correlation between multiple type-specific infections within an individual woman.
A priori, we considered additional covariates for inclusion in our vitamin D and hrHPV analyses. For descriptive purposes, we produced a minimally adjusted model adjusting only for age (continuous) and self-reported race (dichotomized as white or either African American, Asian, American Indian or Alaska Native, Native Hawaiian or other Pacific Islander, other races, or multiple races). Our moderately adjusted model incorporated additional behavioral and biologic risk factors: current oral contraceptive use (yes, no), highest level of education (less than college, bachelor’s degree, master’s/doctoral degree), smoking (never, former, current smoker), BMI (<18.5, 18.5– <25, 25–29.9, ≥30 kg/m2), and serum concentrations of calcium (ng/mL) and phosphate (mg/dL). Our primary model includes all variables in our moderately adjusted model and also includes a B-spline with 5 degrees of freedom to account for season of enrollment, as serum concentration of vitamin D fluctuates significantly with time of year. Finally, we ran a fully adjusted model that also included sexual behavior risk factors (lifetime number of male sex partners [quintiles] and number of sex partners in the 6 months prior to enrollment [none, non-new only, only new, ≥2 new]) evaluated as precision variables.
Using the mgcv package in R [37] and adjusting for all covariates included in our primary adjustment model, we fit generalized additive models (GAMs) for each exposure to assess evidence of nonlinearity on the logistic scale of the relationship between our predictors and persistent detection of hrHPV.
Sensitivity Analyses
Restricting to a Subset of 14 hrHPV Types
We examined relationships between vitamin D concentrations and persistent hrHPV detection restricting to the 14 types of hrHPV included in most US Food and Drug Administration–approved assays for cervical cancer screening (16/18/31/33/35/39/45/51/52/56/58/59/66/68) [38]. We fit post hoc GAMs to assess evidence of nonlinearity on the logistic scale of the relationship between our predictors and persistent detection.
Vitamin D Concentrations at Baseline or Averaged Between Baseline and Exit
To evaluate whether fluctuations in vitamin D serum concentrations over time affect associations with persistent hrHPV detection, we considered the average of serum concentrations at baseline and exit as the exposure. In separate models, we also considered serum concentrations at baseline restricting to women who completed an exit visit.
Apart from GAMs evaluated in R software [37], analyses were conducted in Stata version 15 software [39].
RESULTS
Study Population Characteristics
One woman whose specimen was insufficient for HPV testing was excluded from analyses; another 4 women were excluded following laboratory result quality assurance due to uncertainty in patient-to-specimen matching. Of the remaining 404 women, 89 had prevalent hrHPV at baseline. Seventeen women who completed <6 visits were excluded. The mean age of the remaining 72 women was 36.6 years (standard deviation [SD], 5.9 years) (Table 1). The subset of women was predominantly white (89%), and 39% had a BMI ≥25 kg/m2. Nearly half of the study sample reported current hormonal contraceptive use, and 39% were current or former smokers. Persistent hrHPV detection was positively associated with smoking and higher BMI, though these associations were not significant (Supplementary Table 1).
Table 1.
Characteristics of Study Population of Mid-Adult Women With Prevalent High-Risk Human Papillomavirus Detected in Seattle, Washington, 2011–2012 (N = 72)
| Demographic, Health, and Behavioral Characteristics | No. (%) |
|---|---|
| Age, y, mean (SD) | 36.6 (5.9) |
| Lifetime No. of male sex partners, median (IQR) (n = 71)a | 12 (8–20) |
| White race | 64 (88.9) |
| Highest level of education | |
| Less than college | 11 (15.3) |
| Bachelor’s degree | 26 (36.1) |
| Master’s or doctoral degree | 35 (48.6) |
| Smoking status | |
| Never smoked | 44 (61.1) |
| Former smoker | 22 (30.6) |
| Current smoker | 6 (8.3) |
| Current hormonal contraceptive user | 33 (45.8) |
| Body mass index, kg/m2 | |
| Normal (18.5–24.9) | 44 (61.1) |
| Overweight (25–29.9) | 25 (34.7) |
| Obese (≥30) | 3 (4.2) |
| Male sex partner(s) within 6 mo prior to enrollmenta | |
| None | 8 (11.3) |
| Non-new partners only | 43 (60.6) |
| 1 new partner | 11 (15.5) |
| ≥2 new partners | 9 (12.5) |
Data are presented as No. (%) unless otherwise indicated.
Abbreviations: IQR, interquartile range; SD, standard deviation.
aOne woman excluded due to missing sexual history data.
Vitamin D Summary Statistics
Medians and interquartile ranges (IQRs) for all vitamin D biomarkers were within normal ranges, as reported previously [24]. Descriptive statistics for vitamin D biomarkers by persistent vs transient/sporadic detection pattern are presented in Table 2. Baseline median 25(OH)D was 31.63 ng/mL (IQR, 23.17–43.44 ng/mL) and 28.09 ng/mL (IQR, 23.79–34.05 ng/mL) for persistent vs transient/sporadic detection patterns, respectively. Within-woman variations in biomarkers between baseline and exit samples are presented in Supplementary Table 2.
Table 2.
Vitamin D Biomarkers: Descriptive Statistics Among Woman-Types, Stratified by Pattern of High-Risk Human Papillomavirus Detection (n = 116)
| Baseline Vitamin D Measure | Pattern of High-Risk HPV Detection | |||||||
|---|---|---|---|---|---|---|---|---|
| Persistent (n = 63) | Sporadic or Transient (n = 53) | |||||||
| Mean | (SD) | Median | (IQR) | Mean | (SD) | Median | (IQR) | |
| 25(OH)D, ng/mL | 34.09 | (12.83) | 31.63 | (23.17–43.44) | 30.59 | (9.73) | 28.09 | (23.79–34.05) |
| 1,25(OH)2D, pg/mL | 53.66 | (20.99) | 51.04 | (39.06–66.53) | 54.27 | (21.82) | 51.04 | (39.08–62.80) |
| 24,25(OH)2D, ng/mL | 2.65 | (1.56) | 2.32 | (1.60–3.35) | 2.16 | (1.30) | 1.67 | (1.34–2.71) |
| Free vitamin D, nmol/L × 1000 | 19.84 | (6.45) | 19.89 | (15.53–23.04) | 17.95 | (4.60) | 17.23 | (15.44–20.80) |
| DBP, μg/mL | 274.54 | (66.72) | 257.50 | (229.50–298.40) | 269.23 | (50.68) | 261.00 | (235.30–286.40) |
| Albumin, g/d | 4.25 | (0.27) | 4.30 | (4.00–4.40) | 4.35 | (0.26) | 4.40 | (4.30–4.50) |
| Calcium, mg/dL | 9.46 | (0.43) | 9.50 | (9.20–9.60) | 9.48 | (0.37) | 9.50 | (9.40–9.70) |
| Phosphate, mg/dL | 3.65 | (0.62) | 3.70 | (3.20–4.00) | 3.48 | (0.56) | 3.50 | (3.10–3.80) |
Abbreviations: DBP, vitamin D binding protein; HPV, human papillomavirus; IQR, interquartile range; SD, standard deviation.
hrHPV Summary Statistics
Among 72 women with prevalent hrHPV, we detected 116 woman-types. Sixty-three woman-types (54.3%) were characterized with a persistent detection pattern, 41 (35.3%) sporadic, and 12 (10.3%) positive only at baseline (transient). The most commonly detected hrHPV type at baseline was type 53 (21.6%), type 26 was not detected, and 75% of infections were positive for 1 of the 14 hrHPV types included in sensitivity analyses.
Associations Between Serum Concentrations of Vitamin D and Persistent Detection of hrHPV
Short-term persistent detection of hrHPV was not statistically significantly associated with serum concentrations of 25(OH)D (Table 3). Each 10 ng/mL increase in serum 25(OH)D at baseline was associated with a non–statistically significant increased likelihood of persistent detection of hrHPV, after adjustment for age, race, season of enrollment, BMI, smoking status, education level, and oral contraceptive use (adjusted OR [aOR], 1.26 [95% CI, .84–1.90]). Models including both fewer and more covariates produced estimates that were also not significantly different from 1.0.
Table 3.
Associations Between Concentrations of Serum Vitamin D Biomarkers at Study Baseline and High-Risk Human Papillomavirus Persistence Among Mid-Adult Women in Seattle, Washington (2011–2012) (n = 116 Type-Specific Infections)
| Vitamin D Measure | Minimally Adjusteda | Moderately Adjustedb | Moderately Adjusted, Including Adjustment for Season of Enrollmentc | Fully Adjustedd | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No. | OR | (95% CI) | P Value | No. | OR | (95% CI) | P Value | No. | OR | (95% CI) | P Value | No. | OR | (95% CI) | P Value | |
| 25(OH)D, continuous serum concentration, per 10 ng/mL increase | 116 | 1.34 | (.95–1.88) | .09 | 116 | 1.28 | (.87–1.87) | .21 | 116 | 1.26 | (.84–1.90) | .26 | 114 | 1.29 | (.78–2.14) | .33 |
| 25(OH)D, categorical (IOM) | ||||||||||||||||
| <20 ng/mL | 11 | 1.00 | … | 11 | 1.00 | … | … | 11 | 1.00 | … | … | 11 | 1.00 | … | ||
| ≥20 ng/mL | 105 | 0.73 | (.23–2.26) | .58 | 105 | 0.63 | (.16–2.52) | .51 | 105 | 0.56 | (.11–2.76) | .47 | 103 | 0.53 | (.08–3.58) | .51 |
| 25(OH)D, categorical (ES) | ||||||||||||||||
| <30 ng/mL | 63 | 1.00 | … | 63 | 1.00 | … | 63 | 1.00 | … | 63 | 1.00 | … | ||||
| ≥30 ng/mL | 53 | 2.38 | (1.02–5.56) | .05 | 53 | 2.57 | (.98–6.72) | .06 | 53 | 2.75 | (.97–7.80) | .06 | 51 | 2.26 | (.75–6.82) | .15 |
| 25(OH)D, categorical (post hoc analysis) | ||||||||||||||||
| <50 ng/mL | 103 | 1.00 | … | 103 | 1.00 | … | 103 | 1.00 | … | 101 | 1.00 | … | ||||
| ≥50 ng/mL | 13 | 2.20 | (.58–8.35) | .25 | 13 | 2.90 | (.73–11.4) | .13 | 13 | 2.73 | (.68–10.93) | .16 | 13 | 6.26 | (.72–54.0) | .10 |
| 1,25(OH)2D | 116 | 1.00 | (.98–1.02) | .97 | 116 | 1.01 | (.99–1.03) | .41 | 116 | 1.01 | (.99–1.04) | .30 | 114 | 1.02 | (.99–1.04) | .17 |
| 24,25(OH)2D | 116 | 1.28 | (.92–1.78) | .14 | 116 | 1.34 | (.94–1.92) | .11 | 116 | 1.38 | (.96–1.98) | .09 | 114 | 1.45 | (.98–2.13) | .06 |
| Free vitamin D | 116 | 1.07 | (.99–1.16) | .08 | 116 | 1.07 | (.98–1.16) | .15 | 116 | 1.06 | (.96–1.17) | .26 | 114 | 1.06 | (.96–1.18) | .22 |
| 25(OH)D adjusted for DBPe | 116 | 1.70 | (1.03–2.81) | .04 | 116 | 1.98 | (1.08–3.63) | .03 | 116 | 1.94 | (.97–3.87) | .06 | 114 | 1.74 | (.95–3.17) | .07 |
Abbreviations: CI, confidence interval; DBP, vitamin D binding protein; ES, Endocrine Society; IOM, Institute of Medicine Food and Nutrition Board; OR, odds ratio.
aAdjusted for binary race (white, nonwhite) and age.
bAdjusted for binary race, age, body mass index (BMI), highest education level, oral contraceptive use, smoking status, and calcium (ng/mL) and phosphate (mg/dL) levels.
cAdjusted for race, age, BMI, highest education level, oral contraceptive use, smoking status, calcium (ng/mL) and phosphate (mg/dL) levels, and season of enrollment.
dAdjusted for binary race, age, season of enrollment, BMI, highest education level, oral contraceptive use, smoking status, calcium (ng/mL) and phosphate (mg/dL) levels, number of lifetime sexual partners, and number of sexual partners in the last 6 months.
ePrimary predictor is 25(OH)D, adjusted for binary race, age, and DBP concentrations (μg/mL) and stratified by DBP haplotype.
Secondary analyses that classified 25(OH)D as sufficient or deficient based on clinically relevant cut-points yielded somewhat contradictory results. ORs assessing the likelihood of persistent detection of hrHPV given 25(OH)D ≥20 ng/mL were consistently below 1.0 but not statistically significant. However, women with higher concentrations of 25(OH)D at baseline (≥30 ng/mL) had a borderline statistically significant higher odds of persistent hrHPV detection (aOR under primary model, 2.75 [95% CI, .97–7.80]). Comparisons of persistent hrHPV detection between women who likely supplemented (≥50 ng/mL) and those who likely did not suggested a similar (though not statistically significant) relationship (aOR, 2.73 [95% CI, .68–10.93]).
Free vitamin D, 25(OH)D adjusted for DBP and stratified by haplotype, and 24,25(OH)2D were positively associated with hrHPV persistence, but results were nonsignificant or borderline significant under the primary models. No significant associations were observed for 1,25(OH)2D.
GAMs did not provide clear evidence of nonlinear trends between vitamin D serum concentrations and persistent detection of hrHPV (Supplementary Figures).
Sensitivity Analyses
Restricting to a Subset of 14 hrHPV Types
Compared to the main analyses, associations were stronger and statistically significant when restricting to 14 hrHPV types (n = 87 woman-types) (Table 4). Women with higher 25(OH)D had significantly higher odds of persistent detection of hrHPV compared to women with lower serum concentrations; this relationship was observed when 25(OH)D was modeled as both a continuous (aOR, 1.82 [95% CI, 1.15–2.88]) and categorical measure (25(OH)D ≥30 ng/mL: aOR, 8.85 [95% CI, 2.69–29.06]). Additional 25(OH)D measurements (aOR, 4.19 [95% CI, 1.18–14.88] for 25(OH)D adjusted for DBP and stratified by haplotype; aOR, 1.14 [95% CI, 1.01–1.30] for free vitamin D) and 24,25(OH)2D (aOR, 1.85 [95% CI, 1.18–2.88]) were also significantly positively associated with persistence. GAMs for 25(OH)D and free vitamin D suggested possible nonlinearity between vitamin D concentrations and persistence of 1 of 14 hrHPV types (Supplementary Figures).
Table 4.
Sensitivity Analysis for Associations Between Concentrations of Serum Vitamin D Biomarkers at Study Baseline and High-Risk Human Papillomavirus (HPV) Persistence Among Mid-Adult Women in Seattle, Washington (2011–2012), Restricting to 14 Clinically Relevant HPV Typesa (n = 87 Type-Specific Infections)
| Vitamin D Measure | Minimally Adjustedb | Moderately Adjustedc | Moderately Adjusted, Including Adjustment for Season of Enrollmentd | Fully Adjustede | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No. | OR | (95% CI) | P Value | No. | OR | (95% CI) | P Value | No. | OR | (95% CI) | P Value | No. | OR | 95% CI | P Value | |
| 25(OH)D, continuous serum concentration, per 10 ng/mL increase | 87 | 1.59 | (1.05–2.40) | .03 | 87 | 1.57 | (.96–2.55) | .07 | 87 | 1.82 | (1.15–2.88) | .01 | 86 | 2.06 | (.89–4.75) | .09 |
| 25(OH)D, categorical (IOM) | ||||||||||||||||
| <20 ng/mL | 7 | 1.00 | … | 7 | 1.00 | … | 7 | 1.00 | … | 7 | 1.00 | … | ||||
| ≥20 ng/mL | 80 | 1.00 | (.27–3.76) | >.99 | 80 | 0.85 | (.18–4.00) | .84 | 80 | 0.76 | (.10–5.65) | .78 | 79 | 0.44 | (.04–4.97) | .51 |
| 25(OH)D, categorical (ES) | ||||||||||||||||
| <30 ng/mL | 49 | 1.00 | … | 49 | 1.00 | … | 49 | 1.00 | … | 49 | 1.00 | … | ||||
| ≥30 ng/mL | 38 | 4.44 | (1.70–11.56) | .002 | 38 | 5.23 | (1.61–16.98) | .006 | 38 | 8.85 | (2.69–29.06) | <.001 | 37 | 12.17 | (.65–226.12) | .09 |
| 25(OH)D, categorical (post hoc analysis) | ||||||||||||||||
| <50 ng/mL | 79 | 1.00 | … | 79 | 1.00 | …- | 79 | 1.00 | … | 78 | 1.00 | -- | ||||
| ≥50 ng/mL | 8 | 2.82 | (.63–12.59) | .17 | 8 | 3.89 | (.65–23.34) | .14 | 8 | 4.92 | (.90–26.99) | .07 | 8 | 25.69 | (.65–1008.36) | .08 |
| 1,25(OH)2D | 87 | 1.00 | (.98–1.03) | .66 | 87 | 1.02 | (.99–1.05) | .20 | 87 | 1.02 | (1.00–1.05) | .10 | 86 | 1.04 | (1.00–1.07) | .06 |
| 24,25(OH)2D | 87 | 1.45 | (.93–2.25) | .10 | 87 | 1.59 | (.96–2.65) | .07 | 87 | 1.85 | (1.18–2.88) | .01 | 86 | 1.90 | (1.08–3.34) | .03 |
| Free vitamin D | 87 | 1.11 | (1.00–1.23) | .05 | 87 | 1.10 | (.98–1.24) | .10 | 87 | 1.14 | (1.01–1.30) | .04 | 86 | 1.16 | (1.00–1.34) | .05 |
| 25(OH)D adjusted for DBPf | 87 | 2.34 | (1.15–4.77) | .02 | 87 | 2.75 | (1.03–7.37) | .04 | 87 | 4.19 | (1.18–14.88) | .03 | 86 | 3.52 | (1.24–9.94) | .02 |
Abbreviations: CI, confidence interval; DBP, vitamin D binding protein; ES, Endocrine Society; IOM, Institute of Medicine Food and Nutrition Board; OR, odds ratio.
aRestricted to 14 high-risk human papillomavirus types included in most US Food and Drug Administration–approved assays for cervical cancer screening: 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68 [38].
bAdjusted for binary race (white, nonwhite) and age.
cAdjusted for binary race, age, body mass index (BMI), highest education level, oral contraceptive use, smoking status, and calcium (ng/mL) and phosphate (mg/dL) levels.
dAdjusted for race, age, BMI, highest education level, oral contraceptive use, smoking status, calcium (ng/mL) and phosphate (mg/dL) levels, and season of enrollment.
eAdjusted for binary race, age, season of enrollment, BMI, highest education level, oral contraceptive use, smoking status, calcium (ng/mL) and phosphate (mg/dL) levels, number of lifetime sexual partners, and number of sexual partners in the last 6 months.
fPrimary predictor is 25(OH)D, adjusted for binary race, age, and DBP concentrations (μg/mL) and stratified by DBP haplotype.
Vitamin D Concentrations at Baseline or Averaged Between Baseline and Exit
Among prevalent woman-types at baseline, 91% (n = 106) had a matched exit visit sample yielding sufficient sera for biomarker testing. Compared with primary analyses using baseline serum as an exposure measure, results restricting to women with sufficient exit samples were similar (Supplementary Table 3). Results using the average of the baseline and exit samples were not appreciably different for biomarkers evaluated as continuous exposures; however, in categorical analyses using averaged baseline and exit 25(OH)D sera concentrations as predictors, ORs were not consistently above or below 1.0 for the cut-points 20 ng/mL, 30 ng/mL, and 50 ng/mL across the different adjustment models.
Discussion
While previous research [22–24] has assessed associations between serum vitamin D concentrations and hrHPV infection cross-sectionally, our analysis utilizes longitudinal data to address this question. Among mid-adult women evaluated monthly for 6 months, serum concentrations of 25(OH)D were positively associated with a short-term pattern of persistently detected hrHPV. While associations were inconsistent and significant only in sensitivity analyses, this may be attributable to limited sample size contributing to low study power rather than lack of true association, and these associations warrant closer examination. When 25(OH)D was assessed using cut-points to determine sufficiency, findings suggested a potential nonlinear relationship, with an inverse association when using a cut-point of 20 ng/mL, but a positive association when using a cut-point of 30 ng/mL. Similarly, post hoc analyses intended to describe the association between women based on whether or not they likely supplement with vitamin D (25(OH)D ≥50 ng/mL) yielded estimates consistently >1.0. Similar nonlinear relationships have been observed in studies of vitamin D and bone fracture [40] and total mortality [41, 42], suggesting a protective effect of 25(OH)D concentrations between 20 and 30 ng/mL but the opposite for lower and higher vitamin D concentrations. Interestingly, in cross-sectional analyses of prevalent hrHPV in this same population, we observed nonsignificant inverse associations with 25(OH)D at cut-points 20 ng/mL and 30 ng/mL, and a significant positive association only at ≥50 ng/mL (aOR vs <50 ng/mL, 2.82 [95% CI, 1.12–7.11]) [24]; differences in sample sizes and comparator groups may, in part, explain contradictory findings for 25(OH)D ≥30 ng/mL in cross-sectional vs longitudinal analyses.
While our a priori hypothesis was that associations between vitamin D biomarkers and hrHPV persistence would be similar for established and probable carcinogenic types, the sensitivity analysis restricting to 14 hrHPV types yielded stronger associations after removing types with more tenuous carcinogenic associations (types 53/73/82/IS39; HPV-53, the second most common type [after HPV-16] in cytologically normal women in North America [43], is prevalent in both the general population and among women in our study sample, but rarely occurs alone in cancers [44]). When restricting to 14 hrHPV types with more established carcinogenic associations, we observed a significant positive association between vitamin D and persistent infection for all biomarkers except 1,25(OH)2D (notably, the vitamin D biomarker with the shortest half-life and greatest variability within individuals [45]). The positive associations observed for other biomarkers, including 25(OH)D (both adjusted for DBP [a relative measure of bioavailable vitamin D] and unadjusted), free vitamin D, and 24,25(OH)2D (positively correlated with 25(OH)D concentrations [46]) suggest a significant positive association between higher systemic vitamin D stores and hrHPV persistence. As in the primary analysis, comparisons using clinically relevant 25(OH)D cut-points suggested a (statistically nonsignificant) inverse relationship between 25(OH)D ≥20 ng/mL and persistence; the positive relationship between 25(OH)D ≥30 ng/mL and persistent hrHPV was strong and statistically significant in sensitivity analyses.
Our decision to combine sporadic and transient detection patterns (as a comparator group to persistent infections) was both operationally and scientifically motivated. Operationally, by combining the small numbers of sporadic or transient infections, we maximized power to detect differences with persistently detected infections. Scientifically, over 6 months, sporadic and transient detection patterns are nearly indistinguishable; transient infections, when followed further, may meet criteria for sporadic infections if redetected. Studies with longer follow-up should consider the potentially meaningful differences in characteristics of infections that are detected transiently vs sporadically, the latter of which may be suggestive of a viral load hovering at or near the level of detection and thus an inability to achieve clearance.
To our knowledge, our study is the first to address the relationship between vitamin D status and longitudinal pattern of hrHPV infection in mid-adult women. We were able to characterize women by pattern of infection rather than relying on the (established) correlation between prevalent infection and hrHPV persistence [3]. Modeling our adjustment of confounders in a phased approach allowed us to present transparent adjustment and enables comparison of our findings to the results of cross-sectional analyses in the same study population. Finally, our use of novel vitamin D biomarkers along the metabolic pathways precludes reliance on 25(OH)D alone in establishing vitamin D status and increases the validity of our findings by showing consistent associations across correlated biomarkers.
Our study is not without limitations. Although longitudinal data collection enables a fuller picture of hrHPV status compared with cross-sectional analyses, future studies may benefit from a period of observation longer than 6 months. However, there is evidence that short-term persistence predicts longer-term persistence and subsequent diagnosis of cervical intraepithelial neoplasia grade 2 or worse (CIN2+) over the next few years [25], and our definition of persistence in this study is predicated on that research. Our sensitivity analysis utilizing average vitamin D measures as predictors leveraged both baseline and exit biomarker samples to estimate vitamin D over the 6-month infection observation period. However, most associations we observed in this analysis were attenuated, and we no longer saw positive associations at the 30 ng/mL and 50 ng/mL 25(OH)D cut-points. It is unclear whether these inconsistent findings highlight the challenges inherent in generating an exposure variable that incorporates data collected at exit (thereby compromising the temporal exposure-outcome sequence), or whether these results cast doubt on those from our primary analyses. Future studies may want to consider collecting vitamin D data prior to ascertainment of HPV status.
Future studies should also aim to understand the relationship between vitamin D and hrHPV in women with elevated risk for cervical cancer mortality; nonwhite women [47], cigarette smokers [48], and women with a BMI ≥30 kg/m2 (indicating obesity) [49] were not well-represented in our study population. Additionally, as these analyses rely on data collected as part of a parent study in which vitamin D was not the primary focus, we lack information on relevant covariates, including potential dietary sources of vitamin D, sun exposure, and vitamin D supplementation.
Unexpected findings among women with the highest vitamin D concentrations in our cross-sectional analyses of this data prompted the question of whether higher concentrations of vitamin D preceded infection with hrHPV or whether the opposite occurred. It is possible women who were classified as likely using supplements were previously diagnosed as deficient, and an association between hrHPV status and 25(OH)D level prior to supplementation might better reflect a relationship between vitamin D status and detection of hrHPV. Future studies evaluating incident detection of hrHPV should consider capturing data on vitamin D supplementation.
Evaluation of the association between vitamin D status, as measured by 5 biomarkers measured along the micronutrient’s metabolic pathway, and short-term persistence of hrHPV yielded mixed results in a population of healthy mid-adult women in Seattle, Washington. Findings from studies with longer-term follow-up in populations with both more variability in vitamin D status and data on supplementation may be useful both in better understanding cofactors linked to hrHPV persistence and in motivating clinical recommendations for subpopulations of women.
Supplementary Data
Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.
Notes
Financial support. This work was supported by the National Institutes of Health (grant numbers R21AI129935 and P30 DK035816).
Potential conflicts of interest. All authors: No reported conflicts of interest.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
References
- 1.Broker TR. Global prevention and management of human papillomavirus related diseases: the pressing challenges and the compelling opportunities. Foreword. Vaccine 2012; 30(Suppl 5):vii–x. [DOI] [PubMed] [Google Scholar]
- 2.Moscicki A, Schiffman M, Kjaer S, Villa LL. Chapter 5: updating the natural history of HPV and anogenital cancer. Vaccine 2006; 24:S42–51. [DOI] [PubMed] [Google Scholar]
- 3.Schiffman M, Castle PE, Jeronimo J, Rodriguez AC, Wacholder S. Human papillomavirus and cervical cancer. Lancet 2007; 370:890–907. [DOI] [PubMed] [Google Scholar]
- 4.Bosch FX, Broker TR, Forman D, et al. , authors of ICO Monograph Comprehensive Control of HPV Infections and Related Diseases Vaccine Volume 30, Supplement 5, 2012 . Comprehensive control of human papillomavirus infections and related diseases. Vaccine 2013; 31(Suppl 7):H1–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Nam K, Chung S, Kim J, Jeon S, Bae D. Factors associated with HPV persistence after conization in patients with negative margins. J Gynecol Oncol 2009; 20:91–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Schmeink CE, Melchers WJ, Siebers AG, Quint WG, Massuger LF, Bekkers RL. Human papillomavirus persistence in young unscreened women, a prospective cohort study. PLoS One 2011; 6:e27937. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.van Duin M, Snijders PJ, Schrijnemakers HF, et al. Human papillomavirus 16 load in normal and abnormal cervical scrapes: an indicator of CIN II/III and viral clearance. Int J Cancer 2002; 98:590–5. [DOI] [PubMed] [Google Scholar]
- 8.Krishnan AV, Feldman D. Mechanisms of the anti-cancer and anti-inflammatory actions of vitamin D. Annu Rev Pharmacol Toxicol 2011; 51:311–36. [DOI] [PubMed] [Google Scholar]
- 9.Sabetta JR, DePetrillo P, Cipriani RJ, Smardin J, Burns LA, Landry ML. Serum 25-hydroxyvitamin D and the incidence of acute viral respiratory tract infections in healthy adults. PLoS One 2010; 5:e11088. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Parva NR, Tadepalli S, Singh P, et al. Prevalence of vitamin D deficiency and associated risk factors in the US population (2011–2012). Cureus 2018; 10:e2741. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.John EM, Schwartz GG, Koo J, Wang W, Ingles SA. Sun exposure, vitamin D receptor gene polymorphisms, and breast cancer risk in a multiethnic population. Am J Epidemiol 2007; 166:1409–19. [DOI] [PubMed] [Google Scholar]
- 12.Di Rosa M, Malaguarnera M, Nicoletti F, Malaguarnera L. Vitamin D3: a helpful immuno-modulator. Immunology 2011; 134:123–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wang TT, Nestel FP, Bourdeau V, et al. Cutting edge: 1,25-dihydroxyvitamin D3 is a direct inducer of antimicrobial peptide gene expression. J Immunol 2004; 173:2909–12. [DOI] [PubMed] [Google Scholar]
- 14.Yarbrough VL, Winkle S, Herbst-Kralovetz MM. Antimicrobial peptides in the female reproductive tract: a critical component of the mucosal immune barrier with physiological and clinical implications. Hum Reprod Update 2015; 21:353–77. [DOI] [PubMed] [Google Scholar]
- 15.Cole AM. Innate host defense of human vaginal and cervical mucosae. Curr Top Microbiol Immunol 2006; 306:199–230. [PubMed] [Google Scholar]
- 16.Stanley MA. Immune responses to human papilloma viruses. Indian J Med Res 2009; 130:266–76. [PubMed] [Google Scholar]
- 17.Lissner D, Mason RS, Posen S. Stability of vitamin D metabolites in human blood serum and plasma. Clin Chem 1981; 27:773–4. [PubMed] [Google Scholar]
- 18.Engelman CD. Vitamin D recommendations: the saga continues. J Clin Endocrinol Metab 2011; 96:3065–6. [DOI] [PubMed] [Google Scholar]
- 19.Jukic AMZ, Hoofnagle AN, Lutsey PL. Measurement of vitamin D for epidemiologic and clinical research: shining light on a complex decision. Am J Epidemiol 2018; 187:879–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Kim SM, Lutsey PL, Michos ED. Vitamin D and cardiovascular disease: can novel measures of vitamin D status improve risk prediction and address the vitamin D racial paradox? Curr Cardiovasc Risk Rep 2017; 11:3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Lutsey PL, Parrinello CM, Misialek JR, et al. Short-term variability of vitamin D-related biomarkers. Clin Chem 2016; 62:1647–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Shim J, Pérez A, Symanski E, Nyitray AG. Association between serum 25-hydroxyvitamin D level and human papillomavirus cervicovaginal infection in women in the United States. J Infect Dis 2016; 213:1886–92. [DOI] [PubMed] [Google Scholar]
- 23.Gupta A, Villa A, Feldman S, et al. Site and sex-specific differences in the effect of vitamin D on human papillomavirus infections: analyses of NHANES 2009–2014 [manuscript published online ahead of print 31 March 2020]. Sex Transm Infect 2020. doi:10.1136/sextrans-2020-054466. [DOI] [PubMed] [Google Scholar]
- 24.Troja C, Hoofnagle AN, Szpiro A, Stern JE, Lin J, Winer RL. Serum concentrations of emerging vitamin D biomarkers and detection of prevalent high-risk HPV infection in mid-adult women. Cancer Epidemiol Biomarkers Prev 2020; 29:1468–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Castle PE, Rodríguez AC, Burk RD, et al. , Proyecto Epidemiológico Guanacaste Group . Short term persistence of human papillomavirus and risk of cervical precancer and cancer: population based cohort study. BMJ 2009; 339:b2569. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Fu TJ, Fu XL, Hulbert A, et al. Short-term natural history of high-risk human papillomavirus infection in mid-adult women sampled monthly. Int J Cancer 2015; 137:2432–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Winer RL, Feng Q, Hughes JP, et al. Concordance of self-collected and clinician-collected swab samples for detecting human papillomavirus DNA in women 18 to 32 years of age. Sex Transm Dis 2007; 34:371–7. [DOI] [PubMed] [Google Scholar]
- 28.Laha TJ, Strathmann FG, Wang Z, de Boer IH, Thummel KE, Hoofnagle AN. Characterizing antibody cross-reactivity for immunoaffinity purification of analytes prior to multiplexed liquid chromatography-tandem mass spectrometry. Clin Chem 2012; 58:1711–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Strathmann FG, Laha TJ, Hoofnagle AN. Quantification of 1α,25-dihydroxy vitamin D by immunoextraction and liquid chromatography-tandem mass spectrometry. Clin Chem 2011; 57:1279–85. Erratum in: Clin Chem 2012; 58:1373. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Henderson CM, Lutsey PL, Misialek JR, et al. Measurement by a novel LC-MS/MS methodology reveals similar serum concentrations of vitamin D-binding protein in blacks and whites. Clin Chem 2016; 62:179–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Hoofnagle AN, Eckfeldt JH, Lutsey PL. Vitamin D-binding protein concentrations quantified by mass spectrometry. N Engl J Med 2015; 373:1480–2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Institute of Medicine, Food and Nutrition Board. Dietary reference intakes for calcium and vitamin D. Washington, DC: National Academies Press, 2010. [PubMed] [Google Scholar]
- 33.Ross AC, Taylor CL, Yaktine AL, et al. Dietary reference intakes for calcium and vitamin D. Washington, DC: National Academies Press,2011. [PubMed] [Google Scholar]
- 34.Binkley N, Novotny R, Krueger D, et al. Low vitamin D status despite abundant sun exposure. J Clin Endocrinol Metab 2007; 92:2130–5. [DOI] [PubMed] [Google Scholar]
- 35.Munoz N, Castellsague X, de Gonzalez AB, Gissmann L. Chapter 1: HPV in the etiology of human cancer. Vaccine 2006; 24S3:S1–10. [DOI] [PubMed] [Google Scholar]
- 36.Bouvard V, Baan R, Straif K, et al. , WHO International Agency for Research on Cancer Monograph Working Group . A review of human carcinogens—part B: biological agents. Lancet Oncol 2009; 10:321–2. [DOI] [PubMed] [Google Scholar]
- 37.R Core Team. R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, 2017. [Google Scholar]
- 38.Salazar KL, Duhon DJ, Olsen R, Thrall M. A review of the FDA-approved molecular testing platforms for human papillomavirus. J Am Soc Cytopathol 2019; 8:284–92. [DOI] [PubMed] [Google Scholar]
- 39.StataCorp. Stata statistical software: release 15. College Station, TX: StataCorp LLC, 2019. [Google Scholar]
- 40.Looker AC, Mussolino ME. Serum 25-hydroxyvitamin D and hip fracture risk in older U.S. white adults. J Bone Miner Res 2008; 23:143–50. [DOI] [PubMed] [Google Scholar]
- 41.Kramer H, Sempos C, Cao G, et al. Mortality rates across 25-hydroxyvitamin D (25[OH]D) levels among adults with and without estimated glomerular filtration rate <60 ml/min/1.73 m2: the third National Health and Nutrition Examination Survey. PLoS One 2012; 7:e47458. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Durazo-Arvizu RA, Dawson-Hughes B, Kramer H, et al. The reverse J-shaped association between serum total 25-hydroxyvitamin D concentration and all-cause mortality: the impact of assay standardization. Am J Epidemiol 2017; 185:720–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.de Sanjosé S, Diaz M, Castellsagué X, et al. Worldwide prevalence and genotype distribution of cervical human papillomavirus DNA in women with normal cytology: a meta-analysis. Lancet Infect Dis 2007; 7:453–9. [DOI] [PubMed] [Google Scholar]
- 44.Schiffman M, Clifford G, Buonaguro FM. Classification of weakly carcinogenic human papillomavirus types: addressing the limits of epidemiology at the borderline. Infect Agent Cancer 2009; 4:8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Feldman D, Wesley Pike J, Adams JS.. Vitamin D (2 volumes). Amsterdam, The Netherlands: Academic Press, 2011:2144. [Google Scholar]
- 46.Cashman KD, Hayes A, Galvin K, et al. Significance of serum 24,25-dihydroxyvitamin D in the assessment of vitamin D status: a double-edged sword? Clin Chem 2015; 61:636–45. [DOI] [PubMed] [Google Scholar]
- 47.Benard VB, Watson M, Saraiya M, et al. Cervical cancer survival in the United States by race and stage (2001–2009): findings from the CONCORD-2 study. Cancer 2017; 123:5119– 37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.International Collaboration of Epidemiological Studies of Cervical Cancer. Cervical carcinoma and reproductive factors: collaborative reanalysis of individual data on 16 563 women with cervical carcinoma and 33 542 women without cervical carcinoma from 25 epidemiological studies. Int J Cancer 2006; 119:1108–24. [DOI] [PubMed] [Google Scholar]
- 49.Calle EE, Rodriguez C, Walker-Thurmond K, Thun MJ. Overweight, obesity, and mortality from cancer in a prospectively studied cohort of U.S. adults. N Engl J Med 2003; 348:1625–38. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.