Calcium, Dairy Foods, and Risk of Incident and Fatal Prostate Cancer: The NIH-AARP Diet and Health Study

Abstract

Calcium and dairy foods in relation to prostate cancer were examined in the National Institutes of Health (NIH)-AARP (formerly known as the American Association of Retired Persons) Diet and Health Study (1995/1996–2001). Diet was assessed with a food frequency questionnaire at baseline. Multivariate relative risks and 95% confidence intervals were estimated by Cox regression. During up to 6 years of follow-up (n = 293,888), the authors identified 10,180 total prostate cancer cases (8,754 nonadvanced, 1,426 advanced, and 178 fatal cases). Total and supplemental calcium were unrelated to total and nonadvanced prostate cancer. However, a statistically nonsignificant positive association with total calcium was observed for advanced (≥2,000 vs. 500–<750 mg/day: relative risk (RR) = 1.25, 95% confidence interval (CI): 0.91, 1.71; p_trend = 0.06) and fatal (≥1,000 vs. 500–<750 mg/day: RR = 1.39, 95% CI: 0.92, 2.09; p_trend = 0.10) prostate cancer. Skim milk, but not other dairy foods, was associated with increased risk of advanced prostate cancer (≥2 vs. zero servings/day: RR = 1.23, 95% CI: 0.99, 1.54; p_trend = 0.01). In contrast, calcium from nondairy foods was associated with lower risk of nonadvanced prostate cancer (≥600 vs. <250 mg/day: RR = 0.82, 95% CI: 0.68, 0.99; p_trend = 0.04). Although the authors cannot definitively rule out a weak association for aggressive prostate cancer, their findings do not provide strong support for the hypothesis that calcium and dairy foods increase prostate cancer risk.

Abbreviations

Abbreviations
 
• AARP

  formerly known as the American Association of Retired Persons

 
• CI

  confidence interval

 
• NIH

  National Institutes of Health

 
• RR

  relative risk

The role of calcium in prostate cancer development and progression has been controversial. The calcium and prostate cancer hypothesis first emerged from a strong positive association between dairy foods and risk of prostate cancer in early case-control studies (1). However, positive findings for dairy foods from most subsequent prospective studies tended to be weak and statistically nonsignificant (2–7). Among nutrients abundant in dairy foods, calcium has been hypothesized to increase the risk of prostate cancer by down-regulating 1,25-dihydroxyvitamin D, a putative anticarcinogenic nutrient (8). In contrast, calcium may decrease prostate cancer risk by suppressing the production of parathyroid hormone, an up-regulator of insulin-like growth factor-I synthesis (9).

Epidemiologic studies have reported inconsistent findings on the association between calcium and risk of prostate cancer. Several studies found that calcium was associated with an increased risk of prostate cancer (5, 6, 10–12), whereas others found no association between the two (2, 7, 13–15). Recently, a small randomized clinical trial of 1,200 mg/day of calcium supplementation to prevent recurrence of colorectal adenomas found a statistically nonsignificant inverse association with prostate cancer in a secondary analysis (16).

The discrepant findings from previous studies may be explained by several factors. First, prostate cancer is quite heterogeneous in regard to tumor characteristics, and studies have found that the association between calcium and prostate cancer differed by tumor aggressiveness (10, 12), with a stronger association seen for advanced tumors. Second, an increased risk of prostate cancer may occur only with very high intake of calcium (i.e., ≥2,000 mg/day) but not with moderate-to-high intakes (i.e., <2,000 mg/day) (5, 10), and studies that lacked data above a critical threshold intake of calcium may have missed an association. Third, small numbers of cases may have resulted in insufficient statistical power to detect an association. Fourth, confounding factors may have affected relations in some studies.

In a large prospective study with a wide range of dietary intake and comprehensive information on potential prostate cancer risk factors, we investigated in detail the associations between calcium and dairy foods and risk of prostate cancer. We addressed whether associations differed by prostate tumor characteristics.

MATERIALS AND METHODS

Study population

In a collaborative effort, the National Institutes of Health (NIH)-AARP (formerly known as the American Association of Retired Persons) Diet and Health Study was initiated by mailing a questionnaire to AARP members aged 50–71 years in 1995–1996 and residing in one of six US states (California, Florida, Louisiana, New Jersey, North Carolina, and Pennsylvania) and two metropolitan areas (Atlanta, Georgia, and Detroit, Michigan) (17). For the 339,671 men who returned questionnaires with satisfactory dietary data, we excluded individuals whose questionnaires were completed by proxies for the intended respondent (n = 14,495); who had cancer, except nonmelanoma skin cancer, diagnosed before baseline (n = 27,261); or who had self-reported end stage renal disease at baseline (n = 626). In addition, we excluded individuals who reported extreme intakes (beyond two times the interquartile ranges of sex-specific Box-Cox log-transformed intake) of total energy (n = 2,577) and calcium from foods (n = 824). After exclusions, the analytical cohort consisted of 293,888 men.

Cancer ascertainment

We followed our cohort members annually for change of address by matching the cohort data to the National Change of Address database maintained by the US Postal Service. Information on address changes also came through receipt of US Postal Service processing of undeliverable mail, from other address change update services, and directly from participants who reported address changes when responding to study mailings such as questionnaires or newsletters. According to information from those address changes, 5 percent of participants were no longer residing in one of the eight states included in the study at the end of the follow-up.

We identified incident cases of prostate cancer (International Classification of Diseases for Oncology, Third Edition, code C619) that occurred during follow-up through December 31, 2000. Cases were identified through probabilistic linkage with the eight state cancer registry databases serving our study. These registries are certified by the North American Association of Central Cancer Registries as being 95 percent complete within 2 years of cancer occurrence. Information on prostate cancer stage and histologic grade (93 percent complete data) was also obtained from cancer registry databases. Our case ascertainment method has been described in a previous study (18). Vital status was ascertained through annual linkage of the cohort to the Social Security Administration Death Master File in the United States, follow-up searches of the National Death Index Plus for participants who matched to the Social Security Administration Death Master File, cancer registry linkage, questionnaire responses, and responses to other mailings.

Advanced prostate cancer cases were defined as those with clinical stages of T3–T4, N1, or M1 according to the American Joint Committee on Cancer's 1997 Tumor-Node-Metastasis classification system, as well as individuals who died of prostate cancer during follow-up. The remaining cases were considered nonadvanced cases. Fatal cases were those who died of prostate cancer through December 30, 2001. High-grade prostate cancer cases were defined as those with grade III by Surveillance, Epidemiology, and End Results coding, which is consistent with a Gleason score of ≥8, and low-grade prostate cancer cases were those with grade I or II (Gleason score of ≤7) (19).

Dietary and risk factor assessment

Dietary intake was assessed at baseline by use of a self-administered, 124-item food frequency questionnaire, which was an early version of the Diet History Questionnaire developed at the National Cancer Institute (20). Participants were asked to report their usual frequency of intake and portion size over the last 12 months using 10 predefined frequency categories ranging from “never” to “≥6 times per day” for beverages, from “never” to “≥2 times per day” for solid foods, and three categories of portion size. The food items, portion sizes, and nutrient database were constructed using the US Department of Agriculture's 1994–1996 Continuing Survey of Food Intake by Individuals (21, 22). The food frequency questionnaire also inquired about frequency and types of multivitamins and frequency and dose of individual calcium supplements, including calcium-containing antacids, at baseline. Calcium intake was estimated from foods and beverages only (dietary calcium) and from foods, beverages, and supplements combined (total calcium). In addition, calcium from dairy foods and nondairy foods was estimated separately. All nutrient intakes were adjusted for total energy intake by use of the residual method (23).

Dairy foods and their serving sizes were defined by the Pyramid Servings Database corresponding to the 1994–1996 Continuing Survey of Food Intake by Individuals, which utilizes a recipe file to disaggregate food mixtures into their component ingredients and assigns them to food groups. One serving of dairy food was defined on the basis of standard portion sizes developed by the US Department of Agriculture, such as 1 cup (244 ml) of milk or yogurt, 1.5 ounce (42.5 g) of natural cheese, or 2 ounces (56.0 g) of processed cheese (24). We did not consider butter as a dairy food, because it has not generally been included in the dairy food group in previous studies and it is not a major contributor to calcium intake.

The food frequency questionnaire used in our study was validated by two nonconsecutive 24-hour dietary recalls among 1,953 AARP participants (25). The median (10th–90th percentiles) of dietary calcium intake in the 24-hour recalls in men was 834 mg/day (443–1,414 mg/day). The energy-adjusted correlation coefficient of dietary calcium intake between the food frequency questionnaire and the reference method was 0.63 in men.

In the baseline questionnaire, we also inquired about race/ethnicity, educational level, marital status, smoking status, frequency of vigorous physical activity, medical history including diabetes, family history of cancers, and current body weight and height. In a subsequently mailed questionnaire in 1996–1997 (69 percent response rate), we requested information on prostate cancer screening using a prostate-specific antigen test during the past 3 years.

Statistical analysis

We used the Cox proportional hazards model (26) with age as the time metric to estimate relative risks and two-sided 95 percent confidence intervals. Person-years of follow-up time were calculated from the date the baseline questionnaire was returned until the date of prostate cancer diagnosis, death, move out of the registry areas, or end of follow-up, whichever came first. The proportional hazards assumption was evaluated and confirmed by modeling interaction terms of time and calcium and other covariates.

We used predefined categories of total calcium and dairy food intakes to maximize contrasts and to ensure comparability with other studies. The test for trend across categories of intake was performed by assigning participants the median value of their categories and entering that variable as a continuous term in the regression model. In multivariate models, we adjusted for race/ethnicity; education; marital status; body mass index; vigorous physical activity; smoking; alcohol consumption; history of diabetes; family history of prostate cancer; screening for prostate cancer using prostate-specific antigen; and intakes of tomatoes, red meat, fish, vitamin E, alpha-linolenic acid, and total energy. Additional adjustment for intakes of vitamin D or phosphorus was performed in specific multivariate models. The proportion of missing values for a covariate was generally less than 4 percent; an indicator variable for missing response was created for a covariate (27).

RESULTS

Among a total of 10,180 prostate cancer cases (total participants = 293,888), the majority were nonadvanced prostate cancer (n = 8,754), followed by advanced (n = 1,426) and fatal (n = 178) prostate cancer. The 10th and 90th percentiles of intake were 478 and 1,247 mg/day for dietary calcium, 526 and 1,530 mg/day for total calcium, and 0.3 and 2.9 servings/day for dairy foods. Approximately two thirds of dietary calcium was from dairy foods, and one third was from nondairy foods. Calcium from nondairy foods was mostly from grains (46 percent), vegetables (24 percent), and fruits (16 percent). The prevalence of calcium-containing multivitamin use was 49 percent, and the prevalence of individual calcium supplement use was 23 percent. Dietary calcium was positively correlated with dietary vitamin D (r = 0.83) and dietary phosphorus (r = 0.61), and total calcium was positively correlated with total vitamin D (r = 0.63) and with total phosphorus (r = 0.71).

Compared with men with low total calcium intake, men with high total calcium intake tended to be more educated, to be more physically active, to be screened for prostate cancer by use of prostate-specific antigen, to have a history of diabetes, and to be nonsmokers (table 1). In addition, men with high total calcium intake had higher consumption of dairy foods and vitamins D and E, but they had lower intakes of red meat and alcohol. High dairy food consumers were more likely to be non-Hispanic White, to have a history of diabetes, to be screened for prostate cancer by use of prostate-specific antigen, and to have higher intakes of vitamin D and red meat, but lower intake of alcohol, than did low dairy food consumers.

Total calcium intake was not associated with risk of total and nonadvanced prostate cancer (table 2). Comparing ≥2,000 versus 500–<750 mg/day of total calcium, the multivariate relative risk was 1.01 (95 percent confidence interval (CI): 0.90, 1.14) for total and 0.98 (95 percent CI: 0.86, 1.12) for nonadvanced prostate cancer. A nonsignificant positive association was observed for advanced prostate cancer (≥2,000 vs. 500–<750 mg/day: multivariate relative risk (RR) = 1.25, 95 percent CI: 0.91, 1.71; p_trend = 0.06) and fatal prostate cancer (≥1,000 vs. 500–<750 mg/day: multivariate RR = 1.39, 95 percent CI: 0.92, 2.09; p_trend = 0.10). Further adjustment for vitamin D attenuated the associations with advanced and fatal prostate cancer.

When associations were examined, stratified by a recent prostate cancer screening using prostate-specific antigen, similar results were observed for men with and men without prostate-specific antigen screening. Comparing ≥2,000 versus 500–<750 mg/day of total calcium intake, the multivariate relative risk for nonadvanced prostate cancer was 1.03 (95 percent CI: 0.87, 1.23) (n = 4,639 cases) for men with prostate-specific antigen screening and 0.83 (95 percent CI: 0.50, 1.38) (n = 733 cases) for men without. For advanced prostate cancer, the corresponding multivariate relative risks were 1.40 (95 percent CI: 0.91, 2.15) (n = 696 cases) for men with prostate-specific antigen screening and 1.48 (95 percent CI: 0.61, 3.60) (n = 360 cases) for men without. We could not examine the association of fatal prostate cancer by prostate-specific antigen screening because of small numbers of fatal prostate cancer cases.

The relation of calcium to prostate cancer appeared to strengthen with increasing tumor grade and stage, although confidence intervals overlapped considerably. In a comparison of ≥2,000 versus 500–<750 mg/day of total calcium, the multivariate relative risk was 1.01 (95 percent CI: 0.88, 1.16; p_trend = 0.86) (n = 7,967 cases) for low-grade prostate cancer, and it was 1.17 (95 percent CI: 0.87, 1.57; p_trend = 0.14) (n = 1,491 cases) for high-grade prostate cancer. When we further classified prostate cancer by tumor aggressiveness, the multivariate relative risk for nonadvanced low-grade prostate cancer was 1.00 (≥2,000 vs. 500–<750 mg/day: 95 percent CI: 0.87, 1.16; p_trend = 0.78) (n = 7,107 cases), and for advanced high-grade prostate cancer, it was 1.35 (≥2,000 vs. 500–<750 mg/day: 95 percent CI: 0.83, 2.21; p_trend = 0.50) (n = 480 cases).

Calcium from supplements was not related to risk of total, nonadvanced, and advanced prostate cancer, but it was suggestively related to increased risk of fatal prostate cancer (table 3). Similarly, calcium from dairy foods was not associated with total, nonadvanced, and advanced prostate cancer, but it was nonsignificantly positively associated with fatal prostate cancer. Calcium from nondairy foods showed a weak inverse association with total and advanced prostate cancer and a significant inverse association with nonadvanced prostate cancer (≥600 vs. <250 mg/day: multivariate RR = 0.82, 95 percent CI: 0.68, 0.99). In contrast, calcium from nondairy foods showed a weak positive relation with fatal prostate cancer. Additional adjustment for total vegetable intake did not materially change the association between calcium from nondairy food and nonadvanced prostate cancer.

Use of individual calcium supplements (which did not include multivitamins) was not associated with risk of prostate cancer. In a comparison of users versus nonusers of individual calcium supplements, the multivariate relative risks were 0.96 (95 percent CI: 0.91, 1.01) for total, 0.95 (95 percent CI: 0.90, 1.01) for nonadvanced, 0.98 (95 percent CI: 0.85, 1.12) for advanced, and 1.00 (95 percent CI: 0.67, 1.50) for fatal prostate cancer.

Total dairy food consumption was not independently related to risk of prostate cancer (table 4). In a comparison of ≥3 versus <0.5 servings/day of dairy foods, the multivariate relative risk was 1.00 (95 percent CI: 0.92, 1.10) for total, 1.02 (95 percent CI: 0.93, 1.12) for nonadvanced, and 0.88 (95 percent CI: 0.70, 1.12) for advanced prostate cancer. There was a statistically nonsignificant positive association between total dairy foods and risk of fatal prostate cancer, which was attenuated and became null after further adjustment for calcium or vitamin D.

Most individual dairy foods showed no relation to prostate cancer (table 5); 27, 60, and 49 percent of participants consumed whole, low-fat, and skim milk, respectively. Consumption of whole milk, low-fat milk, cheese, and yogurt was not associated with risk of prostate cancer. However, consumption of skim milk was associated with an increased risk of advanced prostate cancer (≥2 vs. 0 servings/day: multivariate RR = 1.23, 95 percent CI: 0.99, 1.54; p_trend = 0.01). The association between skim milk and risk of advanced prostate cancer became slightly stronger after additional adjustment for calcium or vitamin D, but it was attenuated and no longer statistically significant after additional adjustment for phosphorus (data not shown).

DISCUSSION

Our study, consistent with a number of previous studies (2, 7, 13–15), found no association between calcium intake and risk of total prostate cancer. In contrast, several other studies have found that calcium intake was associated with an increased risk of total prostate cancer (4–6, 10–12), particularly with very high intake of calcium (>2,000 mg/day) (5, 10); some found a direct linear relation, but others found only a threshold for increased risk. In a comparison of men with ≥2,000 versus <1,000 mg/day of dietary calcium, the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study found a linear positive association with dietary calcium (12). Tseng et al. (11), in the National Health and Nutrition Examination Survey (NHANES) Follow-up Study, observed a linearly increased risk of prostate cancer, but the range of calcium intake was narrow; the median calcium intakes were 455 and 921 mg/day in the lowest and the highest tertiles, respectively. In the Health Professionals Follow-up Study (10), for ≥2,000 versus <750 mg/day of total calcium, the multivariate relative risks of total prostate cancer for the lowest through the highest category were 0.98, 1.07, 1.03, 1.06, and 1.28 (95 percent CI: 1.02, 1.60). The Cancer Prevention Study II (5), similar to the Health Professionals Follow-up Study, also observed a threshold risk elevation for ≥2,000 versus <700 mg/day of dietary calcium, with multivariate relative risks = 1.0, 1.0, 1.0, and 1.4 (95 percent CI: 1.1, 2.0).

A possible differential effect of calcium on nonadvanced and advanced prostate cancer was observed in the Health Professionals Follow-up Study (10). The associations with calcium were stronger for advanced, fatal, or high-grade prostate cancer (≥2,000 vs. 500–749 mg/day: multivariate RR = 2.02, 2.02, and 1.89, respectively) than for nonadvanced or low-grade prostate cancer (multivariate RR = 1.13 and 0.73, respectively). These findings suggest that calcium plays an etiologic role in the later stages of prostate carcinogenesis. Other studies did not find that the association between calcium and prostate cancer differed according to tumor aggressiveness (2, 4, 14). A recent animal study also reported no effect of calcium on prostate tumor progression (28). In our study, the data revealed weak and statistically nonsignificant direct associations between calcium intake and both advanced and fatal prostate cancer.

In a small randomized clinical trial of 1,200 mg/day of calcium supplementation to prevent the recurrence of colorectal adenomas, the calcium supplementation group showed a statistically nonsignificant 15 percent risk reduction for prostate cancer that was mostly nonadvanced low-grade prostate cancer (in the calcium supplementation group: RR = 0.83, 95 percent CI: 0.52, 1.32) (n = 33 cases). When the Health Professionals Follow-up Study examined only nonadvanced low-grade prostate cancer, a similar result was observed (≥2,000 vs. 500–749 mg/day: multivariate RR = 0.79, 95 percent CI: 0.50, 1.25). Apart from chance, this statistically nonsignificant inverse association could be due to a possible beneficial effect of calcium on nonadvanced low-grade prostate cancer. Our study found that calcium from nondairy foods was associated with an approximately 20 percent lower risk of nonadvanced prostate cancer (≥800 vs. <250 mg/day: multivariate RR = 0.83, 95 percent CI: 0.68, 0.99). Calcium intake from nondairy foods may be a marker of other dietary factors or other healthy behavior related to lower risk of prostate cancer.

The association of calcium from supplements with prostate cancer has been inconsistent in previous studies. Calcium from supplements or individual calcium supplement use was weakly associated with an increased risk of prostate cancer in some studies (5, 10), but not in others (7, 15). A suggestively positive association between calcium from supplements and fatal prostate cancer in our study may be due to multivitamin use that was positively correlated with calcium intake from supplements (r = 0.71) and has been associated with a significantly increased risk of fatal prostate cancer in our cohort (29). Other potential confounders of the association are zinc, beta-carotene, selenium, and folate, which have been shown to be related to an increased risk of prostate cancer. However, additional adjustment for these nutrients in our analyses did not materially change the results.

A recent meta-analysis of 10 cohort studies (30) reported that total dairy food intake was associated with a 10 percent increased risk of total prostate cancer (lowest vs. highest category: RR = 1.11, 95 percent CI: 1.00, 1.22). Several (7, 10, 15), but not all (6, 12), studies published after the meta-analysis found a null association with total dairy food intake. Our study did not find that total dairy food intake was related to risk of total prostate cancer.

Among individual dairy foods, an increased risk of prostate cancer has been observed for intakes of milk (11, 31, 32) and cheese (6, 15). In our study, skim milk showed an increased risk of advanced prostate cancer. The significant association between skim milk and advanced prostate cancer, even after adjustment for calcium, suggests that noncalcium components in skim milk are associated with increased risk of prostate cancer. However, we cannot exclude the possibility of chance, because low-fat milk was not related to the risk of prostate cancer in our study although low-fat milk has a similar nutrient composition to that of skim milk and was commonly consumed in our cohort.

The prospective design and the availability of comprehensive data on dietary and nondietary factors allowed us to examine calcium and dairy foods across a wide range of intakes, as well as to examine calcium from different food sources. In addition, our study included more than 10,000 prostate cancer cases with information on tumor characteristics, which enabled us to examine prostate cancers by tumor aggressiveness. However, our study had limited statistical power regarding fatal prostate cancer because of the small numbers of fatal prostate cancer cases.

In our cohort, detection bias due to more frequent prostate-specific antigen screening among men with high calcium intake may have spuriously attenuated any positive association between calcium and advanced or fatal prostate cancer, because prostate cancers would have been diagnosed and treated at an early stage before a potential detrimental effect of calcium on prostate cancer progression would have occurred. However, when we examined the association of calcium with advanced prostate cancer by prostate-specific antigen screening, similar results were observed for men with and without this screening.

With regard to the pathophysiologic plausibility of the calcium/dairy-prostate cancer connection, dairy foods have been postulated to have an adverse effect on prostate cancer by increasing the concentration of insulin-like growth factor I, a potent mitogen associated with an increased risk of prostate cancer (33). Moreover, down-regulation of 1,25-dihydroxyvitamin D by a high blood calcium level has led to the hypothesis that very high intake of calcium (i.e., >2,000 mg/day) increases the risk of prostate cancer by lowering the serum 1,25-dihydroxyvitamin D concentration. However, previous studies have not consistently found an inverse association between serum 1,25-dihydroxyvitamin D and risk of prostate cancer (16, 34–36). On the other hand, another plausible mechanism, down-regulation of parathyroid hormone by calcium, may have a beneficial effect on prostate cancer. High calcium intake lowers parathyroid hormone, which increases insulin-like-growth factor I and parathyroid hormone-related protein, possible prostate cancer growth factors (9).

In conclusion, we found that total and supplemental calcium intakes were not associated with the risk of total and nonadvanced prostate cancer. However, a nonsignificant positive association with total calcium was observed for advanced and fatal prostate cancer. In contrast, calcium from nondairy foods was associated with a lower risk of nonadvanced prostate cancer. Total dairy food intake was not related to the risk of total, nonadvanced, and advanced prostate cancer; skim milk was directly associated with advanced prostate cancer, but no other dairy food was associated with advanced prostate cancer or any other subtype. Although we cannot categorically rule out a weak association for more advanced prostate cancer, our findings do not provide strong support for the hypothesis that calcium and dairy foods increase the risk of prostate cancer. Results from other large cohort studies, with adequate numbers of advanced and fatal prostate cancers, may shed further light on this question.

This research was supported by the Intramural Research Program of the National Cancer Institute, National Institutes of Health.

Cancer incidence data from the Atlanta metropolitan area were collected by the Georgia Center for Cancer Statistics, Department of Epidemiology, Rollins School of Public Health, Emory University. Cancer incidence data from California were collected by the California Department of Health Services, Cancer Surveillance Section. Cancer incidence data from the Detroit metropolitan area were collected by the Michigan Cancer Surveillance Program, Community Health Administration, state of Michigan. The Florida cancer incidence data used in this report were collected by the Florida Cancer Data System under contract to the Department of Health. Cancer incidence data from Louisiana were collected by the Louisiana Tumor Registry, Louisiana State University Medical Center in New Orleans. Cancer incidence data from New Jersey were collected by the New Jersey State Cancer Registry, Cancer Epidemiology Services, New Jersey State Department of Health and Senior Services. Cancer incidence data from North Carolina were collected by the North Carolina Central Cancer Registry. Cancer incidence data from Pennsylvania were supplied by the Division of Health Statistics and Research, Pennsylvania Department of Health, Harrisburg, Pennsylvania.

The views expressed herein are solely those of the authors and do not necessarily reflect those of the contractor or the Florida Department of Health. The Pennsylvania Department of Health specifically disclaims responsibility for any analyses, interpretations, or conclusions.

Conflict of interest: none declared.

References