Datasets:

rntc

/

pubmed-preprocessed-quality

like 0

SPSS statistics Version 22.0 (Armonk, NY: IBM Corp) and Stata version 13 (StataCorp LP, College Station, TX, USA) were used for the statistical analyses. In general data are presented as the arithmetic mean with standard deviation (SD) or 95% confidence intervals (CI).

other

A mixed model with fixed effects (streg with fe option) was used in Stata to study within-subject variations of vitamin D and related measures from second to third trimester . The model included a random intercept. To account for the considerable variation in exposure to sunlight over the year and increase the precision, we adjusted for season. Model-based serum levels in second and third trimester were estimated by using the postestimation command lincom (linear combinations of estimators). Each season was given a weight of 0.25 and this approach gave an estimate covering each season with similar weights.

other

To evaluate differences between study sites by season, we used multivariable linear regression and did separate analyses for second and third trimester. In these analyses, we adjusted for season, age, pre-pregnancy BMI, parity and pre-pregnancy physical activity. When we performed analyses for the third trimester, we also adjusted for group randomization (from the in the original RCT) to take potential treatments effects into account. In the model, we allowed for interaction between study site and season, and used likelihood ratio test to assess possible interactions. In additional analyses we adjusted for vitamin D supplementation and education.

other

The model-based serum levels in Trondheim and Stavanger and the seasonal serum levels at both study sites were estimated using the postestimation command lincom. In the model, each season was given a weight of 0.25. Furthermore, pre-pregnancy physical activity was given a weight of 0.71 (based on the proportion of women that were exercising regularly before pregnancy), and parity 0.43 (based on the proportion of women with one or more children). Continuous variables (pre-pregnancy BMI and age) were mean-centered.

other

Linear and logistic regression analyses were used to estimate the potential association between serum levels of total 25(OH)D, calculated free 25(OH)D, 1,25(OH)2D and PTH levels in second trimester and pregnancy outcomes (GDM and birthweight (BW)). The same analyses were also performed in third trimester. In the multivariable regression models, we adjusted for study site, season, age, pre-pregnancy BMI, parity and pre-pregnancy physical activity. In additional analyses, we also adjusted for education, and intake of vitamin D, calcium and fish.

other

To assess the association between total 25(OH)D and 1,25(OH)2D we used simple linear regression. In all analyses involving 1,25(OH)2D, we used the pweight function in Stata to account for the sampling scheme (the inverse of the probability of an observation being selected into the sample). In this study we did not make any adjustment for multiple testing.

other

The participants from Trondheim and Stavanger were homogeneous in terms of baseline characteristics (Table 1). Prenatally, 17 (2.0%) were underweight (BMI <18.5), 645 (76.7%) had normal weight (BMI 18.5–24.99), 141 (16.8%) were overweight (BMI 25–29.99) and 38 (4.5%) were obese (BMI ≥30) according to the classification of the World Health Organization . No one had class III obesity with BMI ≥40. A very slight increase in supplemental vitD intake (0.04 μg) was observed between second and third trimester. A total of 18% followed the recommendations of a daily intake of 10 μg vitamin D supplement in the second and third trimester (Tables 1 and 2). In third trimester, 18 women had gestational hypertension and 43 had GDM. The mean BW was 3,519 ± 540 grams (S2 Table). At the second sample collection, 58 women from Trondheim and 36 from Stavanger were lost to follow-up (Fig 1). The characteristics did not differ from the original population, although a lower proportion exercised regularly pre-pregnancy (59% vs. 71%). Serum analyses from two women in second trimester and two in third trimester could not be completed in the sub-analysis (n = 250) of 1,25(OH)2D.

review

Continuous variables are given as means ± standard deviations (SD), and categorical variables are given as (n) with percentages (%). The Norwegian authorities’ recommendations for pregnant women are a daily vitD supplement intake of 10 μg, a weekly intake of 300–450 g fish and additionally 900 mg calcium per day.

other

A slight decrease occurred in total 25(OH)D between second and third trimester (Table 4). Increasing or indifferent levels were observed in 314 (43%), whereas 410 (57%) experienced a decline. 1,25(OH)2D, PTH and DBP levels were increasing, whereas calcium, magnesium, phosphate and albumin decreased (Table 4). A decline was observed in free 25(OH)D, the percentage of free 25(OH)D (0.023% versus 0.021%, p <0.0001) and bioavailable 25(OH)D. Corrected calcium was increasing (Table 4).

other

Continuous variables are given as means ± standard deviations (SD) and categorical variables are given as numbers (n) with percentages (%). The Norwegian authorities’ recommendations for pregnant women are a daily vitD supplement intake of 10 μg, a weekly intake of 300–450 g fish and additionally 900 mg calcium per day.

other

In Table 3 crude serum values of 25(OH)D, calculated free 25(OH)D, albumin-bound 25(OH)D, bioavailable 25(OH)D, PTH, calcium, corrected calcium, magnesium, phosphate, albumin and DBP in second and third trimester are presented. The mean crude 25(OH)D levels in second and third trimester were 66.1 ± 24.8 and 64.3 ± 27.1 nmol/L, respectively. In second trimester, 232 (27%) had vitD insufficiency and 40 (5%) deficiency. In third trimester, the corresponding numbers were 246 (34%) and 52 (7%). Mean PTH concentrations were 2.8 ± 1.09 and 3.6 ± 1.51 pmol/L, respectively, in second and third trimester. In third trimester, 27 (3.7%) had PTH levels above the upper reference limit (6.9 pmol/L). Of these, 56% had vitD insufficiency or VDD. None of those with elevated PTH had 25(OH)D levels >74 nmol/L. PTH elevation was more frequent (82%) in the autumn, winter and spring. Corrected calcium, magnesium, phosphate and creatinine levels were within reference range. In a sub-analysis, 1,25(OH)2D levels ranged from 97–408 and 105–408 pmol/L, in second and third trimester. Crude mean 1,25(OH)2D levels were 199.1 (CI 192.9 to 205.2) and 229.1 (CI 220.9 to 237.3) pmol/L in second and third trimester, respectively. Calculated free 1,25(OH)2D was 833.0 (CI 806.2 to 859.8) fmol/L in second trimester and 876.7 (CI 845.9 to 907.5) fmol/L third trimester.

other

*** A mixed model with fixed effects (streg with fe option) was used in Stata. The model included a random intercept. We adjusted for season. The model-based levels in second and third trimester were estimated by using the postestimation command lincom (linear combinations of estimators).

other

Among the 250 women in the sub-analysis, the seasonally adjusted 1,25(OH)2D concentration increased from 198.9 (CI 195.6 to 202.2) pmol/L to 230.3 (CI 226.9 to 233.8) pmol/L) (change = 31.4 (CI 24.7 to 38.2) pmol/L, p <0.0001). A decline in 1,25(OH)2D concentration was observed in 26% (crude data). Among those with 25(OH)D <30 nmol/L and 25(OH)D >75 nmol/L in the third trimester, 45% and 17%, respectively experienced a decline in 1,25(OH)2D between the trimesters. Five percent of those women with falling 1,25(OH)2D had PTH elevation above the reference range, and all subjects in this group showed 25(OH)D levels <34 nmol/L (third trimester). In both second and third trimester, women with vitamin D insufficiency and deficiency displayed lower 1,25(OH)2D levels than those with adequate vitD status. In a linear regression model, each 1 nmol/L increment in the 25(OH)D concentration increased the levels of 1,25(OH)2D by 0.74 pmol/L (CI 0.53 to 0.96 pmol/L, p <0.0001), and by 0.92 pmol/L (CI 0.69 to 1.14 pmol/L, p <0.0001) in second and third trimester, respectively. Season adjusted calculated free 1,25(OH)2D increased from 832.1 (CI 818.9 to 845.1) fmol/L in second trimester to 881.7 (CI 868.3 to 895.1) fmol/L in third trimester (change = 49.7 (CI 23.2 to 76.2) fmol/L, p <0.0001)

other

Differences in serum measures between Trondheim (latitude 63°N) and Stavanger (latitude 58°N) in second and third trimester are presented in Table 5. In both trimesters, lower levels of free and total 25(OH)D and higher PTH levels were seen at the northerly latitude. Seasonal variations in all 25(OH)D measures and PTH occurred in both trimesters, while DBP and 1,25(OH)2D did not show the same seasonal pattern (Figs 2 and 3). After adjustment for education and vitD supplementation, similar results were seen. In the second and third trimester, respectively, 56 (36%) and 61 (47%) from Trondheim and 25 (40%) and 23 (51%) from Stavanger exhibited vitD insufficiency during wintertime. Of the women who were in the third trimester during the dark season, 15 (12%) from Trondheim and 3 (7%) from Stavanger had VDD.

other

(A) Seasonal variation of serum total, free and bioavailable 25(OH)D, PTH and DBP in second trimester. (B) Seasonal variation of serum 1,25(OH)2D in second trimester, in a sub-analysis including 250 women living in Trondheim, Norway. Solid squares represent women living in Trondheim, Norway (latitude 63°N) and grey dots represent women living in Stavanger, Norway (latitude 58°N). Vertical lines represent 95% confidence intervals. A multivariable linear regression analysis was used, and separate analyses were performed for second and third trimester. In analyses involving 1,25(OH)2D, we used the pweight function in Stata to account for the sampling scheme (the inverse of the probability of an observation being selected into the sample). Abbreviations: PTH, parathyroid hormone; DBP, vitamin D-binding protein.

clinical case

(A) Seasonal variation of serum total, free and bioavailable 25(OH)D, PTH and DBP in third trimester. (B) Seasonal variation of serum 1,25(OH)2D in third trimester, in a sub-analysis including 250 women living in Trondheim, Norway. Solid squares represent women living in Trondheim, Norway (latitude 63°N) and grey dots represent women living in Stavanger, Norway (latitude 58°N). Vertical lines represent 95% confidence intervals. A multivariable linear regression analysis was used, and separate analyses were performed for second and third trimester. In analyses involving 1,25(OH)2D, we used the pweight function in Stata to account for the sampling scheme (the inverse of the probability of an observation being selected into the sample). Abbreviations: PTH, parathyroid hormone, DBP, vitamin D-binding protein.

clinical case

In the simple linear regression analysis, lower total and free 25(OH)D levels in second trimester were associated with higher BW (change in BW for each 1-unit increase in the serum measure 25(OH)D and free 25(OH)D was -1.8, (CI -3.3 to -0.4) g and -7.7 (CI -13.8 to -1.6) g, respectively). In a multivariable regression model the association was almost fully attenuated (change in BW for each 1-unit increase in the serum measure 25(OH)D and free 25(OH)D was -0.5 (CI -2.1 to 1.1) g and -1.2 (CI -7.8 to 5.4) g, respectively). No association of 1,25(OH)2D and PTH with BW was found in the simple linear regression analysis (change in BW for each 1-unit increase in the serum measure 1,25(OH)2D and PTH was 0.5 (CI -0.9 to 1.9) g and 29.2 (CI -4.2 to 62.6) g, respectively) or in a multivariable regression model (change in BW for each 1-unit increase in the serum measure 1,25(OH)2D and PTH was -0.3 (CI -1.8 to 1.2) g and -5.2 (CI -39.9 to 29.6) g, respectively)

other

No associations of 25(OH)D measures, 1,25(OH)2D and PTH with GDM were observed in the logistic regression modelling (25(OH)D: crude odds ratio (OR) 1.00 (CI 0.99 to 1.01), adjusted OR 1.00 (CI 0.99 to 1.02); Free 25(OH)D: crude OR 1.00 (CI 0.95 to 1.05), adjusted OR 1.00 (CI 0.94 to 1.06); 1,25(OH)2D: crude OR 1.00 (CI 0.99 to 1.01), adjusted OR 1.00 (CI 0.99 to 1.01); PTH: crude OR 1.19 (0.92 to 1.52), adjusted OR 1.18 (CI 0.90 to 1.54)) (S3 and S4 Tables). The same analyses were also performed in third trimester, but no substantial differences were found. After adjustment for education and intake of vitD, calcium and fish, similar results were observed.

other

To our knowledge, this is the largest longitudinal study investigating several indices of vitD metabolism at two time points during pregnancy. In accordance with previous studies, hypovitaminosis D was frequent, and 246 (34%) of the well-educated, Caucasian women had vitD insufficiency and 52 (7%) VDD in the third trimester. In spite of Northern latitudes, the prevalence was lower than reported in most previous European studies of Caucasian pregnant women [9, 16, 23, 29, 30, 47–49]. In a sub-analysis (n = 250), a decline in 1.25(OH)2D was observed in half of those with VDD. This was reflected in an increased occurrence of secondary hyperparathyroidism (SHPT).

other

Few studies have addressed free and bioavailable 25(OH)D during pregnancy [21, 23, 24]. We measured DBP concentrations at two time points which allowed us to calculate free and bioavailable 25(OH)D [41, 42]. Schwartz et al. observed similar measured free 25(OH)D concentrations in pregnant women and a comparator group . A reference range for directly measured free 25(OH)D (5.3–7.7 pg/mL ≈ 13.1–19.3 pmol/L) was provided in a recent study . The levels were in the same range in our study subjects (13.6–15.3 pmol/L), although calculated free 25(OH)D has been claimed to overestimate the level [24, 50]. In accordance with previous studies, a rise occurred in DBP levels, which contribute to the decline in free and bioavailable 25(OH)D [21, 23]. A Korean study showed lower levels of calculated bioavailable 25(OH)D in pregnant than in non-pregnant women (median 1.7 ng/mL ≈ 4.3 nmol/L in second and third trimester), whereas total levels were indifferent . Median level of calculated bioavailable 25(OH)D was similar (4.4 nmol/L) in the current study. Testing of both total, free and bioavailable 25(OH)D would provide a better assessment of vitD status in conditions with changes in DBP levels like pregnancy.

other

Consistent with previous studies, a rise occurred in 1,25(OH)2D levels between second and third trimester [23, 26–28]. We also calculated free 1,25(OH)2D levels, since the free hormone is responsible for the biological actions, and observed an increment in correspondence with total 1,25(OH)2D [14, 22, 44, 51, 52].

other

A decrease in 1,25(OH)2D levels was, however, found in 45% of those with VDD in the final trimester. These women also displayed lower 1,25(OH)2D levels in both trimesters compared to those with circulating 25(OH)D >75 nmol/L. Recently, Hollis et al. suggested that maximal 1,25(OH)2D concentrations during pregnancy require 25(OH)D levels of 100 nmol/L . In the current study, only 10% of the total population reached this threshold. Most studies of 1,25(OH)2D levels in pregnancy have a small sample size, and show a large variation in third trimester (mean range 86.4–283.0 pmol/L) . In comparison, we observed 1,25(OH)2D concentration within the upper range (229.1 pmol/L), ranging from 105 to 408 pmol/L. The intestinal calcium absorption doubles during pregnancy, mainly attributed to the increase in 1,25(OH)2D [4, 26, 27, 51, 53]. The rise in 1,25(OH)2D is not driven by PTH, as a decline to the lower end of the reference level occurs [4, 26, 27]. Thus, other regulators of 1-alpha-hydroxylase as parathyroid hormone-related protein (PTHrP), placental lactogen and estradiol must account for most of the circulating 1,25(OH)2D during pregnancy [8, 26, 27]. Prolactin and placental lactogen have been proposed to compensate for the lack of vitD . PTHrP, which peaks late in pregnancy, could be involved in keeping serum calcium at an adequate level by mobilizing calcium from bone . In the current study, 4% exhibited elevated PTH levels, preferentially in months with little UVB radiation. Of these, 56% had vitD insufficiency consistent with SHPT. A similar prevalence (2%) of SHPT among Caucasian women was found in a UK study . In the current study, PTH elevation was not seen in women with 25(OH)D levels above 74 nmol/L. This complies with Kramer et al. who reported PTH suppression at 25(OH)D levels >82 nmol/L during pregnancy . The corresponding level in the non-pregnant state was 81 nmol/L . This indicates similar thresholds for vitD supplementation in pregnancy as in the non-pregnant state. This is in line with the classification of the Endocrine Society (vitD insufficiency <75 nmol/L) [3, 17].

study

The significance of the low levels of 1,25(OH)2D, as noticed in a proportion of our study subjects, is little explored. A concomitant rise in PTH and PTHrP at the end of pregnancy may pose a substantial burden on the maternal skeleton and could explain some of the cases with pregnancy-associated osteoporosis. Although 1,25(OH)2D is replaced by compensatory hormones to maintain calcium homeostasis, we postulate that this could impact the fetal and maternal skeleton adversely.

other

Several factors influence 25(OH)D levels including ethnicity, food and sun habits, latitude, altitude, season, and genetic polymorphisms [15–17, 29, 30]. Most previous studies show minor changes of 25(OH)D during pregnancy [16, 23, 26]. This is in line with our findings showing a very modest decline in 25(OH)D between second and third trimester. Maternal 25(OH)D levels seem to remain relatively stable during pregnancy, despite the increased synthesis of 1,25(OH)2D and the transplacental transfer of 25(OH)D to the fetus . Severe VDD during pregnancy was observed in a small proportion of the study subjects. This is of concern, as normal vitD levels in the neonates are reliant on adequate maternal vitD status .

other

Few studies have addressed vitD status in pregnant women at northern latitudes and the impact of small latitudinal differences . Despite minor differences in latitude and similar intake of vitD, levels of total and free 25(OH)D were lower at the northerly study site in both trimesters. VitD insufficiency was less frequent in third trimester (34%) than in a Swedish study (65%), performed at the same latitude as our southern study site . This was reflected in higher mean PTH level among the Swedish women . The difference in prevalence may be attributed to lower vitD intake in the Swedish study.

study

So far, the relationship between maternal 1,25(OH)2D levels and skeletal outcomes has not been addressed, while studies on 25(OH)D status and bone health in the offspring are diverging [8, 27]. Severe VDD in pregnancy is associated with hypocalcemia, rickets and craniotabes in the infant [4, 8, 10, 15, 17, 27]. Observational studies show a positive association between maternal calcium intake and fetal bone development, as well as postnatal bone mineral content (BMC) and bone mineral density (BMD), while RCTs show conflicting results . In a RCT from the UK, a daily supplementation of 25 μg vitD during pregnancy did not improve BMC of the infant . However, in a subgroup born in the winter season, a significant effect on BMC was seen . Longitudinal studies investigating association between maternal vitD status and BMC in offspring at 9-years of age show diverging results [8, 57]. In an Australian study (n = 341 mother-offspring pairs), maternal vitD inadequacy in second trimester was associated with reduced peak bone mass in the 20-year-old offspring, implying increased risk for osteoporosis in the future .

study

Numerous studies, including a meta-analysis have shown that VDD is associated with adverse pregnancy outcomes [1, 4, 6, 7]. The meta-analysis of observational studies, which included more than 22,000 women, concluded that maternal VDD is associated with an increased risk for GDM and lower birthweight infants . A Canadian study demonstrated an association between PTH and GDM, but not 25(OH)D and 1,25(OH)2D, and pregnancy outcomes . We found no association of vitD measures and PTH with GDM. In contrast to other studies, lower 25(OH)D was related to higher BW, but not after adjusting for potential confounding factors [6, 27]. The same relationship was observed between free 25(OH)D and BW. In previous studies reporting a positive association between maternal vitD and birthweight, a higher proportion of women had VDD compared to our study [60, 61]. Hollis et al. proposed that 25(OH)D levels should be at least 100 nmol/L to give beneficial health effects and lower risk for adverse pregnancy outcomes [20, 52, 62]. This was supported by a recent study showing that pregnant women with 25(OH)D levels ≥100 nmol/L had a 62% lower risk of preterm birth compared to those with concentrations <50 nmol/L . The fact that only 10% of our study participants reached levels ≥100 nmol/L may have reduced our ability to detect an association. Neither 1,25(OH)2D nor PTH were associated with BW. Morley et al. reported a positive relationship between PTH and BW, whereas 25(OH)D was associated with reduced intrauterine long bone growth . RCTs addressing the effects of vitD supplementation on pregnancy outcomes have shown diverging results [1, 7, 8, 27, 54]. This may be attributed to differences between studies, including the prevalence of VDD, calcium status, the vitD supplement dose, and start of the intervention [4, 7, 8, 27].

study

In the dark season, UVB-mediated synthesis of vitD is absent at northern latitudes, and vitD has to be obtained through diet and supplements [9, 15, 47]. Western-style diet has low content of vitD, and in Norway, few foods are fortified with vitD [17, 65]. In Norway, the authorities recommend a daily vitD supplement of 10 μg, a weekly intake of 300–450 g fish, and additionally 900 mg calcium daily [25, 66]. In contrast, there are no specific Swedish supplement recommendations [30, 67]. It is of concern that only 18% of the well-educated women followed the vitD supplement advices, and only half adhered to the recommendations concerning fish and calcium intake.

other

The major strengths of the present study are the large number of participants recruited all year round, a high follow-up rate, repeated sampling during pregnancy and standardized procedures for sampling. Analyses were performed concurrently, applying the same instruments and procedures. The study population was ethnically homogenous contributing to limited bias due to skin pigmentation, clothing habits and genetic polymorphisms [7, 8, 15–17]. Furthermore, the study sites were located in different geographical regions of Norway, providing an opportunity to investigate latitudinal differences.

other

The participants were well-educated Caucasian women with low-risk pregnancies, which may affect the generalizability. Serum 25(OH)D was analyzed by ECLIA (Roche), although liquid chromatography-tandem mass spectrometry (LC-MS/MS) is considered to be the golden standard . The FFQ used in this study may overestimate the intake of vitD . The calculation of free 25(OH)D and 1,25(OH)D are dependent on several factors, including accurate measurements of DBP and albumin [44, 50]. Nielsen et al. reported a high correlation between calculated free and directly measured 25(OH)D (r = >0.80), albeit calculation of free 25(OH)D levels may give an overestimation compared to direct measurement [24, 50]. In this study, several comparisons were made, thus there is an increased probability for false positive findings, and the results need to be interpreted with care.

other

Although Norwegian authorities recommend vitD supplementation and fish intake during pregnancy, we show that hypovitaminosis D in pregnancy is frequent in well-educated Caucasian women, particularly during wintertime. This was reflected in low adherence to the recommendations. Despite minor differences in latitude, levels of total and free 25(OH)D were lower at the northerly study sight at both second and third trimester. It is noticeable that almost half of those with 25(OH)D levels below 30 nmol/L experienced a decline in 1,25(OH)2D concentration between second and third trimester (sub-analysis). These women also displayed lower 1,25(OH)2D levels, as reflected in PTH elevation. The current findings are of concern as maternal vitD insufficiency has been shown to associate with lower offspring peak bone mass. Our data highlight the need for increased attention regarding vitD requirement during pregnancy among policy-makers, physicians and the general population. The authorities’ recommendations should be revisited, and strategies to ensure adherence should be implemented.

study

**DBP and 1,25(OH)2D were analyzed at Hormone Laboratory, Oslo University Hospital. Abbreviations: CV, total analytical coefficient of variation; PTH, parathyroid hormone; ECLIA, electrochemiluminescence immunoassay; DBP, Vitamin D-binding protein; RIA, radioimmunoassay.

clinical case

**In a sub-analysis of 1,25(OH)2D, 250 women from Trondheim were included. We have applied probability weights (the inverse of the probability of an observation being selected into the sample) in the statistical analysis of 1,25(OH)2D to produce estimates representative of the total Trondheim population.

other

**In a sub-analysis of 1,25(OH)2D, 250 women from Trondheim were included. We have applied probability weights (the inverse of the probability of an observation being selected into the sample) in the statistical analysis of 1,25(OH)2D to produce estimates representative of the total Trondheim population.

other

Antidepressants are the most commonly prescribed pharmacological agents for the treatment of mood disorders and are frequently prescribed off-label for the treatment of associated symptoms (e.g., insomnia). In addition, it has been suggested that adjunctive antidepressants with hormone replacement therapy are effective in the treatment of peri- and postmenopausal depressive women. However, some studies have previously reported that antidepressant use is associated with increased risk for cancer, with reproductive system and gastrointestinal cancers being the most studied malignancies.

study

A recently published meta-analysis identified a modest increase in the risk for breast and ovarian cancer with the use of antidepressants. The pooled odds ratio (OR) representing the association between antidepressant use and breast/ovarian cancer was 1.11 (95% confidence interval [CI], 1.03–1.20). Parsing mechanistic pathways potentially linking antidepressants to breast/ovarian cancer needs to consider separate lines of evidence suggesting that selective serotonin reuptake inhibitors (SSRIs) have an inhibitory effect on tumor growth. Moreover, other studies have reported no association between risk for ovarian cancer and the use of antidepressants. Several studies have reported no conclusive evidence of breast cancer risk associated with the use of SSRIs after adjusting for the degree of serotonin reuptake inhibition and duration of use.

study

Endometrial cancer is one of the leading causes of death amongst female patients with cancer worldwide. The age-standardized population incidence of endometrial cancer is 1.6 per 100,000. Few studies have investigated the effects of antidepressant use on endometrial cancer. For example, Kato et al reported that the use of antidepressants increases the risk for hormone-related cancers (i.e., breast, endometrial, and ovarian cancers) (relative risk [RR] = 1.8; 95% CI, 1.15–2.81). Fortuny et al reported no significant effect of SSRIs use (i.e., paroxetine and fluoxetine) on risk for endometrial cancer in a population-based case–control study in the United States. To our knowledge, no prior study has investigated the effect of novel antidepressants, such as serotonin-norepinephrine reuptake inhibitors (SNRIs) on endometrial cancer.

other

The National Health Insurance (NHI) program was launched on March 1, 1995 by Taiwan's government. Approximately 99.5% of the Taiwanese population was enrolled in the NHI program in December 2008. The National Health Insurance Research Database (NHIRD), derived from the original claims data of the NHI program, includes ambulatory care, hospital inpatient care, and prescription claims data between January 1, 1997 and December 31, 2008. The study was approved by the Institution Review Board of Cheng-Shan Medical University.

other

Patients with cancer, including endometrial cancer, are eligible to register with the Catastrophic Illness Registry and apply for a catastrophic illness certificate in Taiwan. The diagnosis must be confirmed by tissue pathology. The issuance of the certificate requires a diagnosis of catastrophic illness by physicians and a formal review by the Bureau of National Health Insurance, conducted by a panel of related medical experts.

other

The International Classification of Diseases, Ninth Revision, Clinical Modification, was used, and endometrial cancer was coded as 182.xx. A diagnosis of cancer was confirmed with the Catastrophic Illness Registry Dataset. The index date was operationalized as the date of the first endometrial cancer claim. The units of analysis were incident diagnosis of endometrial cancer between January 1, 1997 and December 31, 2008 and antidepressant prescription within 365 days before the index date.

other

For each endometrial cancer case, we used an incidence density sampling method and randomly selected 10 matched controls at index date. The control population of 1 million individuals was randomly selected from the NHI dataset. To be included in the control population, an individual was required to have been without a current or history of cancer on the index date. Controls were individually age-matched to the case by birth year (Fig. 1).

other

We identified antidepressants (N06A) according to the Anatomical Therapeutic Chemical classification system (table S1, Supplemental content). Antidepressants were categorized as SSRIs (i.e., citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, and sertraline), SNRIs (i.e., duloxetine and venlafaxine), tricyclic antidepressants (TCA; i.e., amitriptyline, clomipramine, dosulepin, doxepin, imipramine, maprotiline, and melitracen), monoamine oxidase inhibitors (MAOIs; i.e., moclobemide and rasagiline), noradrenergic and specific serotonergic antidepressants (i.e., mirtazapine), serotonin antagonist and reuptake inhibitor (i.e., trazodone), and norepinephrine-dopamine reuptake inhibitor (i.e., bupropion). Prescription data to proxy exposure to an antidepressant were obtained from the NHIRD. Antidepressant prescription data after the index date were excluded from analysis.

other

We adjusted for the use of pharmacological agents with potentially confounding effects (e.g., estrogen, estrogen/progesterone, aspirin, nonsteroid anti-inflammatory drugs [NSAIDs], and statins) prescribed before the index date. The presence of comorbid medical conditions (e.g., depressive disorder, anxiety disorder, type 2 diabetes mellitus [type 2 DM], hypertension, hypercholesterolemia, and obesity) was also assessed.

other

We used the SAS version 9.2 software packages (SAS Institute, Cary, NC) to carry out conditional logistic regression models to investigate the association between 7 classes of antidepressant exposure and endometrial cancer risk. In each class of antidepressants, the crude OR and the adjusted OR were stratified into 4 cumulative dosages (≧28 DDD, ≧84 DDD, ≧168 DDD, and ≧336 DDD).

other

Corrected ORs are calculated after getting adjusted for demographic data and confounding variables including depressive disorders, anxiety disorder, type 2 DM, hypertension, hypercholesterolemia and obesity, and confounding drugs. The statistical significance of associations was assessed by P value <0.05 or a 95% CI.

other

Kato et al reported that the use of antidepressants was associated with greater risk for self-reported hormone-related cancers (RR = 1.8; 95% CI, 1.15–2.81). They included 672 incident cases of hormone-related cancer, of which only 20 cases had been prescribed antidepressants at least 4 weeks before enrollment. Within this foregoing sample, 16 developed breast cancer, and 4 developed either ovarian or endometrial cancer. The number of cases with endometrial and/or ovarian cancer was not sufficient for separate analysis.

other

Several epidemiological studies have investigated the association between antidepressant use and the risk for female hormone-related cancer. Reeves et al conducted a prospective cohort study and concluded that there might be an elevated risk with SSRI use (OR = 1.16; 95% CI, 0.96–1.39). However, several other studies have reported a null association between antidepressant use and the risk for breast cancer. Harlow et al reported that pharmacological agents that operate through dopaminergic mechanism may increase relative risk for ovarian cancer (OR = 2.9; 95% CI, 1.3–6.4). A recent publication by Wu et al reported that there was no association between the risk for ovarian cancer and use of antidepressants. Studies investigating the association between antidepressant and hormone-related cancers have been inconsistent.

study

We identified 8392 cases with a diagnosis of endometrial cancer and 82,432 age-matched controls. Sociodemographic data of the study population are reported (e.g., age, income, and urbanization) in Table 1. Comorbid medical disorders, physical diseases, and exposure to potentially confounding concomitantly administered drugs are also reported in Table 1. The differences in levels of income and urbanization between cancer cases and controls were statistically significant (P < 0.001). Individuals with endometrial cancer were significantly more likely to have comorbid type 2 DM, hypertension, hypercholesterolemia, and obesity (P < 0.0001). Compared with controls, the case population was more likely to be prescribed estrogen, progesterone/estrogen, aspirin, and statins (P < 0.0001).

other

The main findings are presented in Table 2. Among individuals prescribed antidepressants, a higher percentage of individuals were prescribed SSRIs, TCAs, and MAOIs, in comparison with trazodone. A much lower percentage had been prescribed SNRIs, mirtazapine and bupropion at least 1 year before the index date. The adjusted OR of cumulative SSRI exposure ≧28 DDD was 0.98 (95% CI, 0.84–1.15); SNRI exposure ≧28 DDD was 1.14 (95% CI, 0.76–1.71). No differences in antidepressant exposure of any cumulative dose were observed between the case and control populations (the adjusted ORs of cumulative SSRI exposure ≧336 DDD were 1.19 [95% CI, 0.91–1.57]; SNRI exposure ≧336 DDD was 1.11 [95% CI, 0.53–2.33]). The exposure time effect was similar to the dosage effect (table S2, Supplemental content).

review

To the best of our knowledge, this is the first population-based study to explore the association between mechanistically dissimilar antidepressants, including novel antidepressants like SNRIs, and endometrial cancer. The results of this study indicate that there is no association between antidepressant exposure and incidence of endometrial cancer. Our results remained unchanged after adjusting for confounding factors such as comorbid psychiatric diseases, comorbid physical diseases, use of estrogen, progesterone/estrogen, aspirin, NSAIDs, and statins.

study

The primary result of our study is consistent with the findings reported by Fortuny et al who reported no association between SSRI use (i.e., paroxetine and fluoxetine) and the risk of endometrial cancer (OR = 0.9; 95% CI, 0.6–1.4). Fortuny et al included 469 endometrial cancer cases from The Estrogen, Diet, Genetics, and Endometrial Cancer Study, a population-based case–control study conducted in Northern New Jersey, United States. It was not reported however if any relationship existed between SSRI total exposure dosage and cancer risk.

other

Our first objective is to predict the minimum frequency of the resistant allele in the sheep population needed to achieve scrapie control. The second objective is to calculate, under different compliance scenarios, how long the ram selection program needs to be maintained to reach the minimum ARR allele frequency and achieve scrapie control. These two objectives are achieved by constructing a combined genetic and epidemiological model, and by using this model to predict the time development of genotype frequencies and reproduction number R0 at a national scale. Our model calculation of R0 for scrapie requires quantification of two important between-flock heterogeneities. The first one is the variation in within-flock genotype frequencies. The level of such variation is important for the prospects for scrapie control: the higher the abundance of farms with, initially, relatively low frequencies of resistant animals, the longer the breeding program may need to be sustained to reduce R0 to below unity (and, correspondingly, the higher the population-level minimum ARR allele frequency will be). In Ref. a genotyping survey of farms in the production sector, that accounts for over 90% of the Dutch sheep population, was carried out to gain information on the variation between farms with respect to genotype composition. We will use a within-flock transmission model developed in Ref. to translate the distribution of within-flock genotype frequencies into a distribution of within-flock R0 values which in turn serves as a basis for the modelling of between-flock transmission. The second important between-flock heterogeneity is of contact rates (mixing) between sheep flocks. We will use a simple model characterizing this heterogeneity with a single parameter. Due to a lack of data, the value of this parameter is unknown; our model predictions will therefore be based on assuming scenario values.

other

Recently further quantitative trait loci influencing resistance to scrapie have been identified as well as PrP gene polymorphisms at codons other than 136, 154 and 171 having a protective effect . These findings widen the range of options for the design of breeding programmes, which could be of relevance in particular in breeds with a low frequency of the ARR allele. We do not model these polymorphisms here as they were not utilised in the Dutch breeding programme.

other

Notwithstanding the strengths in our study, there were several methodological limitations. Data on smoking and lifestyle were not reported. A meta-analysis identified that cigarette smoking was significantly associated with a reduced risk of endometrial cancer, especially among postmenopausal women. Furberg conducted a prospective study and demonstrated that inactivity and high-energy intake are major risk factors for endometrial cancer. Further study to discover the relationship among antidepressants, lifestyle, and the risk of endometrial cancer is needed.

other

Antidepressants are widely used in treating symptoms such as depressed mood, anxiety, and pain. Female patients are often prescribed antidepressants for protracted periods of time. The clinical recommendation for multiple-year exposure to antidepressants, in some cases, invites the need for comprehensive characterization of safety concerns related to their exposure. The findings have important implications for clinicians to discuss with female patients about the safety of antidepressants use and endometrial cancer.

other

Amerio et al reviewed the US Food and Drug Administration (FDA) preclinical in vivo evidence to compare the carcinogenic risk between drug classes, with a focus on psychotropic drugs. Among antidepressants, 63.6% (7/11) of examined agents were associated with carcinogenicity. Specific agents associated with carcinogenicity were mirtazapine, sertraline, paroxetine, citalopram and escitalopram, duloxetine, and bupropion. Tricyclics can induce hepatic microsomal enzymes capable of enhancing estrogen metabolism, which may lead to increased gonadotropin levels. Endogenous and exogenous estrogens have known to be associated with the risk of endometrial cancer. In our study, we did not identify a moderational effect of pharmacological estrogen on the association between antidepressant and endometrial cancer.

study

There are several strengths of the study herein. First, the data were driven from national representative population, which reduces selection bias. Second, information on medication use was obtained from prescription claims data, consequently reducing recall bias. Third, confounding factors such as income, and urbanization, medications (such as estrogen, estrogen/progesterone, aspirin, NSAIDs, and statins) prescribed before the index date, and medical conditions (including depressive disorder, anxiety disorder, type 2 DM, hypertension, hypercholesterolemia, and obesity), were adjusted for in the analysis. Finally, we explored the association between all classes of antidepressants, and the cumulative dose effect of antidepressants on the risk for endometrial cancer, which provides further merit to our conclusion that a nonassociation exists.

other

We use surveillance data consisting of the scrapie test results accumulated within the Dutch active surveillance on TSEs in sheep (from 2002 onwards), and of a yearly random genotyping sample from this active surveillance (from 2005 onwards), both from the healthy-slaughter and the fallen-stock samples. Details on the sampling strategy, genotyping technique and rapid test used are given in Ref. . The test sensitivity in detecting scrapie infection in animals without ARR allele is unknown. Evaluated on scrapie cases confirmed by Western Blot of the brainstem the test sensitivity is close to 95% . However, test sensitivity in the surveillance is expected to be lower as early on in the incubation period scrapie infection has not yet propagated to the brainstem . As the rate of propagation to the brainstem is also genotype dependent, test sensitivity may therefore also be expected to depend on genotype. Detected scrapie prevalence in Dutch culled flocks gives an indication of the minimum value of the sensitivity .

study

The culled-flocks data (2003–2008) consist of scrapie genotyping results and scrapie infection test results in animals that were culled, as part of the mandatory scrapie control efforts, on flocks of origin of scrapie index cases. For details on genotyping and testing see . Immunohistochemistry (IHC) was used for confirmation of the positive cases detected using the rapid test. IHC and Western blotting were used to discriminate between classical and atypical scrapie.

other

Scrapie is a fatal infectious neurodegenerative disease for which susceptibility is associated with polymorphisms in the ovine prion protein (PrP) gene. Polymorphisms at codons 136 (A/V), 154 (R/H) and 171 (Q/R/H) largely determine resistance to scrapie with the VRQ allele being most susceptible, and the ARR allele being resistant to classical scrapie [1–3]. Based on selective breeding for resistance, national eradication programs have been implemented in several countries in Europe, including Great Britain [4–9] and The Netherlands . The intensity of selective breeding varies between countries. One of the most ambitious programs was implemented in the Netherlands, where selection of rams with the ARR/ARR genotype for breeding started in 1998 (voluntary basis) and was obligatory for all sheep farmers from October 2004 to June 2007 .

other

The attempt to eliminate scrapie in this way in The Netherlands can be seen as an instructive attempt to control a widespread infectious disease by breeding, rather than by the more usual vaccination and stamping-out strategies. Clearly, the level of compliance to the selection of scrapie resistant rams for breeding is an important determinant of the effectiveness of the program. However, full compliance may not be needed if a lower than 100% level of resistance is sufficient for control. From an epidemiological viewpoint national scrapie control can be considered achieved when the population-level basic reproduction number R0 has been reduced to below unity, the threshold value for epidemic spread. The frequency of the ARR allele for which this threshold value is attained is the minimum ARR allele frequency for scrapie control, a concept analogous to that of a critical vaccination coverage . A previous analysis has provided evidence that the Dutch scrapie control program up to now has produced both an increase in the prevalence of scrapie resistant genotypes and a reduction in scrapie transmission.

other

In 2007 a postal and genotyping survey in Dutch sheep flocks was carried out. The results were described in Ref. . From 689 farms that completed the postal survey, 168 accepted the offer to genotype (part of) their animals. A maximum of 35 ewes were blood sampled per farm, and samples were taken proportionally per birth year cohort. If farmers owned less than 35 ewes, a maximum of 5 rams could be sampled too. Samples were sent to the Central Veterinary Institute in Lelystad (Now Wageningen Bioveterinary Research, Lelystad) for analysis of the polymorphisms at the PrP gene codons 136, 154 and 171 through Taqman probe analysis. A total of 3314 sheep were genotyped, including 3207 ewes born between 1995 and 2007. For further details on this survey we refer to Ref. .

other

We model the Dutch national sheep population as a population of sheep flocks that vary in genetic content, distinguishing two levels of transmission: within a flock and between flocks. For a review of previous within-flock and between-flock scrapie transmission modelling see . The importance of taking into account the genetic variation between flocks can be illustrated as follows. Let us assume for definiteness that large within-flock scrapie outbreaks would be precluded if the ARR allele frequency in the flock exceeds 80%, and that the overall ARR allele frequency in the population is 85%. Then, if half of the flocks would have an ARR frequency of 100% and the other half one of 70%, large outbreaks were still possible in 50% of flocks, in contrast to a situation without variation, in which all flocks would have the same allele frequency of 85% and in which large outbreaks were not possible in any flock. We note that apart from the between-flock differences in genetics, another reason why it is natural to distinguish the two levels of transmission in the population is that contacts between sheep residing within one and the same flock are more intensive than between animals residing in different flocks.

other

The within-flock model calculates the within-flock reproduction number, denoted in this paper by R0w, from the genotype distribution in the flock. This is based on a model developed in Ref. , which is parameterized using genotyping and case data from Dutch flocks culled under EU statutory control measures . An initial frequency distribution of within-flock R0w values is based on both a farm genotyping survey and a genotyping sample from the active surveillance . Starting from this initial distribution, we use the within-flock transmission model to calculate how the distribution evolves in time for a given level of compliance to the ram selection program. From this, the between-flock model in turn calculates the time evolution of the between-flock R0. The latter parameter represents the population-level R0, and when it drops below unity, national scrapie control has (by definition) been achieved. The ARR allele frequency for which R0 = 1 is the minimum frequency required for scrapie control. The starting value of the population-level R0, characterizing the situation in 2008, is based on scrapie incidence data in the Dutch active surveillance. When the between-flock R0 ≤ 1, isolated within-flock outbreaks of scrapie may still occur with varying duration but no major between-flock spread will be possible. For a field study showing the success of selective breeding to control scrapie at the flock level see .

other

Our purpose is to model the flock-level scrapie transmission potential, as quantified by the within-flock basic reproduction number, in dependence of the within-flock genotype frequencies. We use an SI-type within-flock transmission model with homogeneous mixing between sheep of different genotype, in which we assume that genotypes differ both in susceptibility and in infectiousness; this model was developed in Ref. . We denote by fγ the proportion of animals in the flock that has genotype γ, by gγ the relative susceptibility of genotype γ, and by hγ the relative infectiousness of genotype γ. Finally, we denote the absolute scale of transmission by a dimensionless parameter β. Then the definition of the reproduction number leads to the following expression within-flock R0w: R0w=βQ0w,Q0w=∑γfγgγhγ.

other

This relationship can be derived by noting that fARR changes due to replacement animals having a different ARR allele frequency than the ewes they are born to; as ARR/ARR rams are selected for breeding, the frequency of non-ARR alleles in the newborns is one half of that in the ewes.

other

In the between-flock transmission model we consider a population of flocks for which the R0w is drawn from a distribution PDFt(R0w). In order to derive this distribution for the year t = 2008, the starting point for our predictive calculations, we calculate Q0w=R0w/β for each of the 168 flocks of the genotyping survey, and determine a distribution model that provides a good match to the histogram of 168 values. Subsequently, we obtain PDF2008(R0w) from this distribution and from the Weibull model distribution for β as the distribution of the product of Q0w and β.

other

This expression defines R0w as a weighted average of the product βgγhγ, with the genotype frequencies fγ as weighting factors. The values of the parameter products gγhγ used are based on setting the gγ equal to the relative scrapie risks in different genotypes as estimated from culled-flocks data in Ref. (the values of gγ used are given in Table 1 of Ref. ) and on setting the parameters hγ equal to one for all genotypes except for those with at least one ARR allele, for which hγ is set to zero. This approximation is based on the assumption motivated in that the contribution of ARR/VRQ and ARR/ARQ animals to R0w is negligible. The parameter β incorporates the variation in R0w due to causes different from the genetic content of the flock, such as lambing practice; it may therefore differ between flocks. The variation in β is described using a Weibull distribution, the two parameters of which were estimated in Ref. .

other

We calculate the effect on R0w of a breeding program based on ram selection as follows. When a flock is subject to ARR/ARR ram selection, the newborn lambs, from which replacement stock will be selected, have at least one ARR allele. It follows that, if we neglect bought-in replacement stock (as these are typically small in number ), replacement animals will contribute negligibly to R0w. Thus, assuming that all age categories are subject to the same replacement rate, the expected change in R0w between year t and year t+1 due to ram selection equals: R0w(t+1)−R0w(t)=−rR0w(t),(1) where r denotes the yearly replacement rate. The expected change in the frequency fARR of the ARR allele in the flock can also be expressed in terms of r, as follows: fARR(t+1)−fARR(t)=r2(1−fARR(t)).(2)

other

With the compliance c we denote the proportion of farms, within the farms that have an R0w above one in 2008, that comply with the ram selection program. In this model the R0w distribution for non-compliant farms is assumed to be stationary. This is a conservative approximation, as there will be some dissemination of resistant alleles into non-compliant farms when they buy-in replacement ewes from compliant flocks. As the postal survey in 2007 indicated that only 20% of Dutch sheep farms frequently purchase ewes, we expect this dissemination effect to be relatively small. Our model also neglects the effect that the culling of detected scrapie flocks, through removing susceptible alleles, will have on the R0w distribution. This effect is expected to be small due to the low yearly detection probability of affected flocks based on the arguments given in the additional file of Ref. .

other

In order to approximately account for heterogeneities of mixing between flocks that exist as a result of e.g. regionality of contacts and between-farm differences in trading of animals, breeds present on the farm, and levels of shared grazing, we introduce a single mixing parameter α. In absence of data on the mixing between Dutch flocks, more detailed modelling of the heterogeneities would introduce more parameters with unknown values. The parameter α enters in the relationship between the population-level R0 and the observed prevalence of infected farms, that we assume to be as follows: R0=α(11−i*−1)+1.(4)

study

We note that for the time evolution of s(t) it does not matter to which extent the farms with R0w already below one comply with the breeding program; only the compliance of farms with R0w above one matters. We also note that once a breeding program has run for some time, as is the case in The Netherlands at t = 2008, the average compliance of the farms with R0w above one will be less than that of farms with R0w below one, as the breeding program makes the average R0w of the compliant farms go down. We take this effect into account by a model, detailed in the SI, that calculates, from an overall compliance (the “compliance” for which we list the scenario values in the Result section), the compliance of farms which have an R0w above one in 2008.

other

For the compliant part of the population, the relationship expressed by Eq (1) implies that one year of ram selection reduces each R0w with a factor 1 −r. Therefore, starting from the R0w distribution for a given year t, one year of ram selection with a compliance c produces the following new distribution: PDFt+1(R0w)=(1−c)PDFt(R0w)+c1−rPDFt(R0w/(1−r)).

other

Here i* is the (endemic) prevalence of infected farms within the subpopulation of farms with R0w>1, and α≥1. For the case α = 1 (homogeneous mixing) the model reduces to a well-known result for the SIR model in endemic equilibrium. For α>1 it provides a simple, phenomenological expression for heterogeneous mixing, as heterogeneous mixing causes R0 to be higher than the value often calculated using the well-known result for homogeneous mixing . In an SIR model of a population of size N in which a proportion f of individuals are immunized at birth or are genetically immune against the infection, the relationship between the population-level R0 and the observed endemic prevalence of infection is given by R0=11−αi*, with α = 1/(1 − f). For sufficiently small αi* this relationship coincides with Eq (4) to a good approximation. In terms of heterogeneous mixing, the model given by Eq (4) can therefore be thought of as approximately describing a population in which a core group of size N/α is responsible for the bulk of the transmission due to high contact rates, and the remaining group of size N − N/α hardly contributes due to low contact rates. For example, a value of α = 4 can be roughly interpreted as a situation in which 25 percent of the non-resistant flocks are responsible for the bulk of between-flock transmission.

other

The value R0(2008) is calculated by applying Eq (5) to an endemic situation observed in the surveillance in 2002–2005, estimating the prevalence i* for t = 2005 from the prevalence in the active surveillance and in culled flocks and extrapolating R0(2005) to R0(2008) as detailed in the Supplementary Information.

other

In Fig 2 we present a histogram of the quantity Q0w=R0w/β as calculated with the within-flock model from the genotyping survey data in 168 flocks (black bars). The results show that there is much variation in Q0w between flocks, in line with the variation in genotype frequencies discussed in . The white bars in Fig 2 represent a model distribution fitted to the data histogram. The distribution is of exponential form with an additional probability at Q0w=0 (for details see SI).

other

As described in the Methods section, the model distribution for Q0w shown in Fig 2 and the Weibull model distribution for β estimated in Ref. together determine the model distribution for R0w. The part of this distribution relating non-zero R0w values is shown in Fig 3. According to the model, 37.3% of Dutch flocks had an R0w above one in early 2008, i.e. 37.3% of flocks were susceptible to epidemic within-flock scrapie spread. The tail of the R0w distribution in Fig 3 corresponds to flocks that, as they have highest scrapie transmission potential, would require the longest period of selective breeding to bring R0w below one. This tail is therefore an important determinant of the prospects for obtaining population-level scrapie control.

other

Model extrapolation results for R0(t) under the two scenarios of compliance c = 75% (“high compliance”) and c = 35% (low compliance) are shown in Fig 4. In both scenarios we assume that r = 0.2, which is a plausible estimate for the mean replacement rate in Dutch sheep farming .

other

Predicted R0 between Dutch flocks (circles) and ARR allele frequency fARR (squares) as a function of time; assumed is a yearly replacement rate of 20% (r = 20%). The line with open (closed) circles corresponds to α = 2 (α = 4).The dashed line indicates the critical value R0 = 1. Left panel: Compliance to ram selection of 75%; Right panel: Compliance to ram selection of 35%.

other

In order to explore the sensitivity of outcome to the uncertainty in α we use two alternative moderately heterogeneous mixing scenarios, defined by α = 2 and α = 4. We also show the time evolution of the overall ARR allele frequency in the Dutch sheep population, obtained by applying Eq (2) to the compliant part of the population. The results for α = 2 suggest that for the high-compliance scenario, scrapie control was achieved in The Netherlands by 2011, when the overall ARR frequency exceeds a minimum of approximately 63% (Fig 4A). In contrast, for a value of α = 4 results in a minimum overall ARR frequency of approximately 70%, obtained by 2014. This sensitivity analysis thus shows that both the minimum ARR frequency and the time by when control is achieved are sensitive to the uncertainty in α. The minimum frequency is also (very) sensitive to the compliance level. For a compliance of only 35% (Fig 4B) our model suggests that scrapie control is never reached, as this compliance level is insufficient for reaching R0 ≤ 1.

other

Our results suggest that with a compliance of 75% scrapie control would have been obtained approximately between 2010 and 2014 for moderately heterogeneous mixing. This prediction is consistent with recent downward trends in scrapie incidence observed in the Dutch active surveillance shown in Fig 5. The increase in fARR observed in Fig 5 in the period 2008–2013 is consistent with a compliance to selective breeding of 75%, and the low prevalence of scrapie in tested animals in recent years is a suggestive indication that scrapie control may have been achieved.

other

We have developed a model describing within-flock and between-flock scrapie transmission as well as the effect on transmission of changes in genotype frequencies due to selection of ARR/ARR rams for breeding. Using this model, we have calculated the minimum ARR allele frequency to obtain classical scrapie control in The Netherlands. The results suggest that for (overall) compliance of 75%, scrapie control is achieved in The Netherlands when the overall ARR frequency exceeds a minimum value in the range of 63 to 70 percent across scenarios assuming moderate heterogeneity of between-flock mixing. These predictions are consistent with more recent surveillance data that suggest that the current (2017) resistant allele frequency is approximately 76% percent and that current scrapie prevalence is very low.

other

By a sensitivity analysis we have shown that the model prediction for the time needed for obtaining scrapie control is dependent in particular on the heterogeneity of between-flock mixing, which usually is difficult to estimate due to a paucity of data. The stronger this heterogeneity, the slower the decline of the population-level basic reproduction number, and therefore the slower the progress towards scrapie control through selective breeding. For the Netherlands, the consistency of recent surveillance data with the model scenarios assuming moderate heterogeneity of between-flock mixing suggests that the mixing of Dutch population of sheep flocks is characterized by a weak or moderate level of heterogeneity. Another important determinant is the between-flock heterogeneity in genotype frequencies, which we have quantified using random genotyping survey data. Finally, the level of compliance to ram selection is an important determinant of the predicted minimum ARR allele frequency to obtain classical scrapie control; in particular, if the compliance is too low, scrapie control will never be reached. The approximate overall compliance to ram selection in The Netherlands is 75% as can be deduced from the observed increase in the ARR allele frequency found in random genotyping samples from the active surveillance. In our model calculation we have not addressed what would be the effect on the minimum ARR allele frequency required for scrapie control in the Dutch sheep population if there would be changes to the surveillance intensity and/or the statutory scrapie flock culling policy. As has been argued in Ref. , these latter measures are thought to have only a minor influence (of a few percent) on the scrapie transmission risks as measured by the population-level R0. Therefore, although in principle this minimum ARR frequency would rise when the number of animals tested is reduced (as has been the case in The Netherlands since January 2014) and/or when the statutory control measures were ceased in future, such a rise would be expected to be at most a few percent.

other

The attempt to eliminate scrapie by selective breeding in The Netherlands can be seen as an instructive attempt to control a widespread infectious disease by breeding, rather than by the more usual vaccination and stamping-out strategies. We hope that our modelling approach and results are also instructive to readers interested in other host-pathogen systems in which genetic changes in the population impact on pathogen spread.

other

For more than 30 years, chemotherapy of invasive urothelial carcinoma has been based on combinations of cisplatin with other cytotoxic drugs. This treatment is moderately efficacious and limited by frequent development of resistance and toxicity in the often elderly patients. Novel drugs targeting growth factor receptors or signal transduction pathways have so far not yielded significant benefits in clinical trials and have therefore not been introduced into clinical practice. Intriguingly, among all cancer types, urothelial carcinoma appears to have the highest prevalence of mutations in chromatin regulator proteins, including various components of the trithorax-like histone-modifying and SWI/SNF1 chromatin remodeling protein complexes . It is therefore reasonable to assume that epigenetic inhibitors represent an alternative approach to chemotherapy of urothelial carcinoma.

other

Many epigenetic inhibitors target the activity of enzymes modifying histones or DNA. For instance, HDAC inhibitors (HDACi) interfere with the enzymatic activity of histone deacetylases and are considered good drug candidates. Our previous comprehensive analysis of expression of different HDACs isoenzymes in UC and their suitability as therapeutic targets revealed HDAC class I enzymes as the best targets for HDACi in UC therapy . Drugs targeting specific class I HDACs like Romidepsin, Givinostat, or 4SC-202 most efficiently inhibited cell proliferation and caused cell death in UC cells. Despite their efficacy, these compounds do not seem optimal for treatment of UC on their own, because cell death occurred only partly by apoptosis and proliferation of benign urothelial control cells was also efficiently blocked [3, 4]. We therefore investigated the combination of Romidepsin with the BET inhibitor JQ1, which has been proposed to synergize with HDACi in several cancer types and to induce a canonical apoptotic response [5–8]. In pancreatic adenocarcinoma, the synergism was ascribed to several interacting mechanisms, encompassing inhibition of AKT signaling, increased STAT3 phosphorylation, and prominent induction of the CDK inhibitor p57KIP2 .

other

JQ1 is the best characterized representative of a novel class of compounds which blocks binding of chromatin proteins by targeting domains recognizing histone modifications. JQ1 specifically targets the bromodomains of transcriptional coactivators like “bromodomain and extra-terminal” (BET) proteins, especially BRD proteins, and appears to be particularly effective in cancers dependent on MYC transcription factors, which are not well druggable by other means . The best studied BET protein BRD4 has been shown to be overexpressed in UC tissues correlating with grade, progression towards metastatic disease, and poor overall survival . As an epigenetic reader of acetylation marks at histone tails, BRD4 functions as a scaffold protein linking chromatin remodeling and transcriptional regulation to cell cycle progression. A newly discovered histone acetyltransferase activity for H3K122 further contributes to chromatin decompaction and transcription activation . Inhibition of BRD4 in particular disrupts super enhancers and represses the oncogenes c-MYC and EZH2 [13, 14]. A pioneer study by Wu et al. on BRD4 in UC revealed its upregulation in cancer tissues and inhibition of cell proliferation by JQ1 in two related UC cell lines, T24 and EJ . Knockdown of BRD4 likewise inhibited proliferation of these UC cell lines. The authors ascribe these effects to inhibition of c-MYC and subsequent downregulation of EZH2.

study

In the present study, we therefore investigated whether JQ1 exerts antineoplastic effects on a broader range of UC cell lines which cover the heterogeneity of urothelial carcinoma more comprehensively. Indeed, its efficacy varied between the cell lines, and JQ1 neither suppressed clonogenic growth nor elicited pronounced apoptosis consistently. The combination of JQ1 with Romidepsin however displayed strong synergies in tumor cell growth suppression and apoptosis induction across all cell lines; concomitantly, histone acetylation was broadly enhanced. As in pancreatic adenocarcinoma cells, p57KIP2 emerged as one factor synergistically responding to the combination treatment. Surprisingly, however, p57KIP2 knockout rather enhanced apoptosis in UC cells.

other

BRD4 expression, effects of siRNA-mediated knockdown, and drug exposure were studied in a range of urothelial carcinoma cell lines (UCCs) representing the heterogeneity of urothelial carcinoma, namely, VM-Cub1, RT-112, T24, 5637, UM-UC-3, HT-1376, 639-V, BFTC-905, J-82, and SW-1710. All cell lines were regularly authenticated by STR profiling and checked for mycoplasm contamination. As normal cell controls, we used the normal urothelial cell lines HBLAK and TERT-NHUC, a culture of primary urothelial cells (NHUC) , and benign immortalized fetal kidney HEK-293 cells.

other

UCCs and HEK-293 cells were cultured in DMEM GlutaMAX-I (Gibco, Darmstadt, Germany) supplemented with 4.5 g/l d-glucose, pyruvate, and 10% FBS (Biochrom, Berlin, Germany). HBLAK cells were cultured in CnT-Prime Epithelial Culture Medium (CELLnTEC, Bern, Switzerland). TERT-NHUC cells were cultured in keratinocyte serum-free medium (Gibco) supplemented with 0.35 μg/ml N-epinephrine and 0.33 mg/ml hydrocortisone. NHUC were cultured in keratinocyte serum-free medium (Gibco) supplemented with penicillin/streptomycin, EGF, and BPE. Primary urothelial carcinoma cultures were established from fresh transurethral resectates and cultured in Epilife Medium (Gibco) supplemented with 0.5 ng/ml EGF, 25 μg/ml BPE, 1% nonessential amino acids (Invitrogen, Darmstadt, Germany), 1% ITS mix (Invitrogen), 3 mM glycine, and 10% fibroblast-conditioned medium on a collagen IV matrix . All cells were cultured at 37 °C and 5% CO2.

other

For siRNA-mediated knockdown, cells cultured in six-well plates were transfected with 8 nM BRD4 ON-TARGET plus BRD4 siRNA-SMART pool (L-004937-00-0005, Dharmacon, Freiburg, Germany) or ON-Target plus Control pool (D-001810-10-05, Dharmacon) using Lipofectamine RNAiMAX (Life Technologies, Darmstadt, Germany) and assayed 48, 72, or 120 h post-transfection.

other

Romidepsin and (+)-JQ1 were purchased from Selleck Chemicals (Munich, Germany) and dissolved in DMSO. Control cells were treated with DMSO only. Drugs were added 24 h after cell seeding. Pan-Caspase inhibitor Q-VD-Oph (SML0063, Sigma Aldrich, Hamburg, Germany) was dissolved in DMSO and used at 30 μmol/l.

other

Drug concentrations for combination treatments were adjusted for each cell line since IC50 values for both compounds varied strongly (Table 1 upper part).Table 1Treatment doses for single and combined treatment. Eight UCCs, two immortalized benign urothelial cells (TERT-NHUC, HBLAK), primary urothelial cells (NHUC), and HEK-293 cells were treated with JQ1 at a fixed range of concentrations. Cell viability was measured by MTT- assay 72 h later. IC50 values for each cell line were determined (upper part). Cell viability results were used according to the Chou-Talalay method to determine doses for the combination treatment with synergistic effects for each investigated cell line. The respective doses were applied for further functional characterization for 48 h (lower part)Urothelial carcinoma cell linesControl cell linesJQ1 IC50 [μmol/l]JQ1 IC50 [μmol/l]VM-Cub10.18TERT-NHUC0.4RT-1120.19HBLAK0.4T240.23NHUC0.2656370.39HEK-2930.26UM-UC-32.6HT-13765.2639-V6.8BFTC-90510JQ1 + Romidepsin combination dosesJQ1 [μmol/l]Romidepsin [nmol/l]Vm-Cub10.222.2UM-UC-312T240.222.2639-V1.81.6HBLAK0.40.89

study

Treatment doses for single and combined treatment. Eight UCCs, two immortalized benign urothelial cells (TERT-NHUC, HBLAK), primary urothelial cells (NHUC), and HEK-293 cells were treated with JQ1 at a fixed range of concentrations. Cell viability was measured by MTT- assay 72 h later. IC50 values for each cell line were determined (upper part). Cell viability results were used according to the Chou-Talalay method to determine doses for the combination treatment with synergistic effects for each investigated cell line. The respective doses were applied for further functional characterization for 48 h (lower part)

other

Viability of cells treated with JQ1 was measured after 72 h by 3-(4,5-dimetylthiazol-2-yl)-2,5-diphenyltetrazolium bromide dye reduction assay (MTT, Sigma Aldrich). Viability of cells after siRNA-mediated BRD4 knockdown or treatment with Q-VD-Oph was measured via total cellular ATP using CellTiter-Glo Assay (Promega, Mannheim, Germany).

other

For determination of IC50 values, JQ1 was added in defined concentration ranges. For determination of drug synergy, JQ1 and Romidepsin were used in fixed dose ratios. For each cell line, individual dose ratios were chosen based on the IC50 of the individual drugs (Table 1 upper part). For each cell line, at least five different combinations of concentrations were applied and then analyzed by the Chou-Talalay method using CompuSyn software . The final cell line-dependent concentrations for the 48-h combination treatment used for subsequent analyses are given in Table 1 (lower part).

other

For colony-forming assays, cells were seeded into six-well plates at a density of 1000 cells/well 48 h post-drug treatment and either 48, 72, or 120 h post-siRNA transfection. After 10–15 days, cells were fixed in methanol and stained with Giemsa (Merck, Darmstadt, Germany).

other

Cell cycle analyses were performed 48 h after treatment with individual drugs or their combinations and 48, 72, and 120 h post-siRNA transfections. Detached cells in supernatant and attached cells were collected and stained with buffer containing 50 μg/ml propidium iodide, 0.1% sodium citrate, and 0.1% Triton X-100 for 1 h at room temperature. To assess apoptotic cell death and necrosis, cells were incubated with Annexin V-FITC (31490013, Immunotools, Friesoythe, Germany), Annexin V binding buffer, and propidium iodide at 2 μg/ml. Flow cytometry was done using the MACSQuant Analyzer (Miltenyi Biotech, Bergisch Gladbach, Germany) and MACSQuantify software as previously described .

other

Total mRNA was isolated using the Qiagen RNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol. cDNA synthesis was performed using the QuantiTect Reverse Transcription Kit (Qiagen) with an extended incubation time of 30 min at 42 °C. qRT-PCR was performed using the QuantiTect SYBR Green RT-PCR Kit (Qiagen) and self-designed primers for the target genes and the reference gene TBP (TATA-box-binding protein) on the LightCycler 96 PCR platform (Roche). The primers used are listed in Additional file 1.

other

Total cellular protein was extracted by lysis for 30 min on ice in RIPA buffer containing 150 mmol/l NaCl, 1% Triton X-100, 0.5% deoxycholate, 1% Nonidet P-40, 0.1% SDS, 1 mmol/l EDTA, 50 mmol/l TRIS (pH 7.6), protease inhibitor cocktail (10 μl/ml, Sigma Aldrich), and phosphatase inhibitor (10 μl/ml, Sigma Aldrich). Protein concentrations were determined by bicinchoninic acid protein assay (ThermoFisher Scientific, Darmstadt, Germany). Proteins were separated in SDS-PAGE gels and then wet-blotted to polyvinylidene difluoride (PVDF) membranes (Merck Millipore, Darmstadt, Germany). Membranes were blocked by 5% non-fat dry milk or BSA in TBS-T (150 mmol/l NaCl, 10 mmol/l TRIS, pH 7.6 and 0.1% TWEEN-20), washed several times, and then incubated with primary antibodies at 4 °C overnight. After several washings with TBS-T, membranes were incubated with horseradish peroxidase-conjugated secondary antibody at room temperature for 1 h. Membranes were then developed using Super Signal West Femto (ThermoFisher Scientific) or Western Bright Quantum (Biozym, Hessisch Oldendorf, Germany). α-tubulin was used as a loading control. Antibodies are listed in Additional file 1.

other