Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Cancer Prev Res (Phila). Author manuscript; available in PMC 2014 April 21.
Published in final edited form as:
PMCID: PMC3993089

Sulindac and Sulindac Metabolites in Nipple Aspirate Fluid and Effect on Drug Targets in a Phase I Trial


Regular use of nonsteroidal anti-inflammatory drugs (NSAIDs) has been associated with reduced risk of breast cancer. Sulindac, a non-selective NSAID with both cyclooxygenase 2 (COX2) dependent and independent activities, is a candidate for breast chemoprevention. We conducted a Phase Ib trial in 30 women at increased risk for breast cancer to evaluate the breast tissue distribution of sulindac at two dose levels (150 mg q.d. and 150 mg b.i.d. for 6 weeks), using nipple aspirate fluid (NAF) as a surrogate of breast tissue drug exposure. We also explored the effect of sulindac on drug-induced biomarkers in NAF. We show that sulindac and its metabolites partition to human breast as measured by NAF levels. Sulindac intervention did not decrease 13,14-dihydro-15-keto prostaglandin A2 (PGEM), a stable derivative of prostaglandin E2, in NAF, but exposure was associated with a significant trend towards higher levels of growth differentiation factor 15 (GDF15) in NAF in women receiving 150 mg b.i.d. (p = 0.038). These results are the first to demonstrate partitioning of sulindac and metabolites to human breast tissue and the first evidence for a potential dose dependent effect of sulindac on GDF15 levels in NAF.

Keywords: NSAID, Nipple aspirate fluid, breast cancer prevention, sulindac, 13,14-dihydro-15-keto prostaglandin A2 (PGEM), growth differentiation factor 15 (GDF15)


There is considerable variation in the effect of regular use of nonsteroidal anti-inflammatory drugs (NSAIDs) and risk of breast cancer (14), though a recent meta-analysis supports a statistically significant, but modest, 12% risk reduction overall NSAID use (5). While the overall magnitude of benefit of any NSAID is arguably modest, the actual benefit is likely underestimated as drug duration, dose, type of NSAID, and possibly tumor subtype specific effects are poorly captured in the observation setting (2).

NSAIDs as chemopreventives in breast cancer are supported by work in human mammary tissues and cell culture studies where overexpression of cyclooxygenase 2 (COX2), one of two COX targets of NSAIDs, acts early in the transition from normal breast cells to malignancy (68). Further, COX2-associated prostaglandin E2 (PGE2) production has been shown to increase aromatase activity in mammary epithelial cells (9), possibly explaining the observed benefit of NSAID use in hormone receptor positive disease (2) and the observation of lower circulating estradiol levels among regular users of NSAIDs (10).

The use of NSAIDs to suppress prostanoid production by inhibiting COX2 activity is not a novel concept for prevention of epithelial cancers. Though NSAID use is now well-supported for colorectal cancer, no randomized controlled trials of NSAIDs have been conducted for the prevention of breast cancer. Extant data on the activity of commonly used NSAIDs in breast tissue following oral dosing are limited, and there are no data on the safest effective doses needed to achieve biomarker modulation in breast tissues. Accumulation in breast milk of long half-life NSAIDs (e.g., naproxen, sulindac, and piroxicam) compared with short-acting agents (e.g., ibuprofen) suggests that certain agents may be more efficacious in tissue-specific chemoprevention based on their tissue-specific dose response (11).

We conducted a phase Ib trial to evaluate the distribution of sulindac and its metabolites to the breast tissue, using nipple aspirate fluid (NAF) as a surrogate of tissue exposure, at two sulindac dose levels (150 mg q.d. and 150 mg b.i.d.). In addition, we explored breast tissue drug effects by measuring NAF levels of growth differentiation factor 15 (GDF15), an NSAID-induced pro-apoptotic protein, and 13,14-dihydro-15-keto prostaglandin A2 (PGEM), a stable derivative of prostaglandins, pre- and post-sulindac treatment.


Study design and eligibility

Thirty women (21–65 years) were randomized to 150 mg sulindac q.d. or b.i.d. for 6–7 weeks. Women were eligible if they produced ≥ 5 µl NAF at their screening visit and met at least one of the following criteria for high risk: 5-year Gail score of > 1.7% risk, medical history of lobular carcinoma in situ (LCIS) or ductal carcinoma in situ (DCIS), at least one first degree relative with history of breast cancer, or BRCA1 or BRCA2 mutation carriage untreated with oophrectomy or mastectomy. Women with a prior history of breast cancer were eligible if they retained at least one intact, unirradiated breast and were > 6 months from all therapy including hormonal therapies. Women with any sensitivity or known contraindications to NSAID use were ineligible. Baseline NAF and blood collection were taken following a 4–6 week NSAID washout. NAF and blood collection at completion of the treatment period was obtained on average within 4 hours of the last sulindac dose for all subjects (range 1–10 hours). The study was approved by the University of Arizona Human Subjects Committee. Written informed consent was obtained from all participants.

NAF collection

NAF was obtained from both breasts using self-expression with a modified breast pump under guidance of the study nurse. NAF was pooled and immediately diluted in phosphate buffered saline containing a protease inhibitor cocktail of AEBSF, leupeptin, DTT, and aprotinin. NAF was spun at 200 × g for 10 minutes; the protein-containing supernatant was aliquoted and frozen at −80°C until analysis. The total protein concentration of each NAF sample was determined using the Micro BCA™ Protein Assay Kit (Thermo Scientific, Rockford, IL). The lower and upper limits of detection for the BCA were 45 and 36,000 µg/ml, respectively.

Blood collection and serum C-reactive protein (CRP) levels

Serum was obtained from a 7cc SST Red Top following centrifugation and stored in aliquots at −80°C until processing for sulindac and metabolite measures. High sensitivity CRP was performed by Sonora Quest Diagnostics (Tucson, AZ).

Quantification of sulindac and its metabolites in serum and nipple aspirate fluid

Sulindac and its metabolites in serum and NAF were quantified using an HPLC-tandem mass spectrometry method developed in our laboratory. An aliquot of diluted serum or NAF was mixed with the internal standard solution (100 ng/mL indomethacin in acetonitrile). The mixture was acidified with 50% phosphoric acid and then extracted with a combination of ethyl acetate:hexane (1:1). The organic layer was collected and evaporated to dryness in a centrifugal evaporator. The dry residues were reconstituted with 40% acetonitrile and then injected onto the HPLC-MS system.

The HPLC-mass spectrometry system consisted of a Surveyor HPLC system and a TSQ Quantum Ultra triple quadrupole mass spectrometer (Thermo Electron, Rockford, IL). Chromatographic separation of sulindac, sulindac sulfone, sulindac sulfide, and the internal standard was achieved on a Luna C8 column (50 × 2 mm, 3 µ, Phenomenex) with a Luna C8 guard column (4 × 2 mm, Phenomenex) and a gradient of acetonitrile and water. The mass spectrometric analysis was performed with the electrospray ionization interface operated in the positive polarity mode. The selected reaction monitoring transition was: m/z 357 → 233 for sulindac, m/z 341 → 233 for sulindac sulfide, m/z 373 → 233 for sulindac sulfone, and m/z 385 → 139 for the internal standard.

For each analytical run, a fresh calibration curve was prepared in the appropriate matrix and used to determine the concentrations of sulindac and its metabolites. Quality control samples were prepared and analyzed along with the authentic samples. The serum calibration curve was linear over the range of 0.33–167 ng/ml for sulindac and sulindac sulfide and the range of 3.3–167 ng/ml for sulindac sulfone with 0.15 ml aliquots of sample volume. The NAF calibration curve was linear over the range of 0.5–250 ng/ml for sulindac and sulindac sulfide and the range of 5–250 ng/ml for sulindac sulfone with 0.1 ml aliquots of sample volume. Satisfactory results of the within-day and between-day assay precision and accuracy were obtained with both the coefficient of variation and the percent difference between measured and theoretical concentrations < 15% over the calibration curve range.

PGEM as a surrogate marker of PGE2 (1214)

PGEM levels were detected using the Prostaglandin E Metabolite EIA Kit (Cayman Chemicals, Ann Arbor, MI) following validation of a modified protocol for NAF. To validate this assay for NAF, eligibility samples collected prior to randomization were pooled to determine the minimal detectable levels and to assess linear performance and between-sample and -run variation. Standards and controls were prepared following the manufacturer’s instructions ( The lower and upper limits of detection of this assay were 1.50 and 220.5 pg/mL, respectively for NAF. All paired samples were run in a single plate. Data were considered valid if the intra-assay CV was ≤ 5%. Ten percent of samples failed intra-assay validity and were thus re-analyzed.

GDF15 analysis

GDF15 content in NAF was determined using the ELISA Builder Human GDF15 DuoSet ELISA (R&D Systems, Minneapolis, MN), as described previously (15). To establish assay conditions for NAF and assess linearity, eligibility samples were diluted 1:20–40 at collection and run in a series of additional dilutions brought up in Reagent Diluent. The assay was conducted according to the manufacturer’s instructions. To control for inter-assay variation, samples from a single participant were assayed on the same plate. All samples and controls were run in duplicate. NSAID-induced and non-induced human colorectal cancer cell line HCT116 lysates were run at two dilutions on every plate as positive controls. The lower limit of detection was 78 pg/mL. The intra- and inter-plate CVs were < 5% for all control samples and < 8 % for within plate duplicates for the NAF samples.

Statistical analysis

The total number of participants randomized was 30; this constitutes the sample size for the intention-to-treat analysis. The sample size 30 (15 per dose group) was determined to allow detection of a standard effect size of 1.06 with 80% power based on a two-sided independent t-test at a significance level of 5% while testing if changes differ significantly by dose. There were 15 randomized to 150 mg sulindac (Clinoril®) daily (q.d) and 15 to 150 mg twice daily (b.i.d.) for 6–7 weeks. Values below the detectable limits of any assay were set to zero for statistical analyses. All serum samples were available for serum drug analysis. Due to technical reasons, NAF levels of drug and metabolites were not measured for 4 participants, resulting in a sample size of 26 (13 in each dose group) for these variables. NAF PGEM and GDF15 concentrations were adjusted by total NAF protein levels to correct for volume and dilution differences between samples. The sample size for change in NAF PGEM and GDF15 was 21 and 22, respectively as protein levels were not detectable in all samples, which reduced the sample size in the paired analyses. The sample size for the pairs for PGEM is one less than GDF15 as one NAF was exhausted before accurate measures of PGEM could be obtained. All serum samples were available for hsCRP analysis.

A two-sided Wilcoxon rank-sum test was used to test if the change in sulindac/sulindac metabolite concentration differed by dose and to test if biomarker concentrations changed from baseline to end-of-study or differed by dose. Furthermore, correlations between drug and metabolite or biomarker levels were determined using Spearman rank correlation tests. These secondary analyses were not corrected for multiple comparisons with a significance level set at 0.05.


Study participants and nipple aspirate fluid production among eligible women

A total of 197 women were contacted for participation in the study with 46 enrolled and 30 randomized to agent. Of the 197 contacted, 67 women refused and an additional 84 were found to be ineligible at pre-study screen. NSAID related risk was the most common reason for ineligibility. Of the 46 women enrolled and screened for NAF production, 14 failed to produce NAF (~30%), one was lost to follow up and one was excluded for concomitant medicine usage not identified prior to enrollment. Each of the 30 randomized participants successfully completed the study and provided information at the end of the study. The average age of the subjects was 45.2 years with no significant differences for menopausal status between the two treatment groups. Baseline and end-of-study paired NAF protein levels were strongly correlated (rho = 0.65, p = 0.0001). There were no significant differences in the mean or median total protein values pre and post sulindac after correcting for NAF dilution (59.2 vs. 60.1 mg/ml and 40.3 vs. 37.8 mg/ml, respectively).

Serum levels of sulindac and metabolites

The mean and median serum levels of sulindac and each of the metabolites at end-of-study are shown in Table 1. All three compounds were detectable in the serum of all subjects (n = 30) after sulindac treatment for 6 weeks. The mean serum sulindac, sulindac sulfide, and sulindac sulfone concentrations were 2.95, 3.89, and 1.95 µg/ml, respectively. Dosing at 150 mg b.i.d. resulted in an average of 35, 120, and 77% increase in serum sulindac, sulindac sulfide, and sulindac sulfone concentrations, respectively, compared with 150 mg q.d. Serum sulfone concentrations were significantly higher in higher dose group (p = 0.05), whereas serum sulfide levels did not significantly differ by dose group (p = 0.06). The level of sulindac in serum was strongly correlated with the levels of sulfide (rho = 0.83, p < 0.0001) and less strongly with the levels of sulfone (rho = 0.46, p = 0.01) in serum.

Table 1
Serum concentration of sulindac and sulindac metabolites, overall and by dose

Nipple aspirate levels of sulindac and metabolites

Of the NAF samples, 26 pairs (pre- and post-drug) were evaluated for levels of sulindac and its metabolites. Mean and median NAF levels of the parent compound and each of the metabolites at end-of-study are shown in Table 2. The NAF drug and metabolite concentrations ranged from undetectable to 1157.0, 340.1, and 157.0 ng/ml for sulindac, sulindac sulfide, and sulindac sulfone, respectively, after 6 weeks of sulindac treatment. NAF sulindac, sulfide, and sulfone levels were undetectable in 10/26, 11/26, and 23/26 of the post-drug samples, respectively. Six of the 10 NAF samples for which sulindac and sulfide were undetectable also had low or undetectable protein levels, suggesting the samples may have been overly dilute to detect drug. The mean (and median) NAF sulindac and sulindac sulfide concentrations were 95.8 µg/ml (39.8) and 86.9 µg/ml (46.3), respectively. For NAF sulindac sulfone, there were too few samples with detectable levels to be informative. There were no statistically significant differences between sulindac (p = 0.56) or sulfide (p = 0.77) levels in NAF by dose. There were no significant correlations between serum levels of sulindac, sulfide or sulfone and NAF drug/metabolite levels (all p > 0.05). However, similar to the strong correlation observed between serum sulindac and sulfide concentrations, there was also a significant positive correlation between NAF levels of sulindac and the sulfide metabolite (rho=0.69, p < 0.0001).

Table 2
Nipple aspirate concentration of sulindac and sulindac sulfide, overall and by dose*

NAF PGEM at baseline and change with exposure to sulindac

Mean and median total NAF PGEM levels at baseline were 27.8 and 11.5 ng/mg total protein, respectively, with a range of 0–198.7 ng/mg total protein. As shown in Table 3, NAF PGEM levels were non-significantly decreased overall with treatment (p=0.519). Change in PGEM did not differ by dose assignment (Figure 1A & B). Of the directly measured drug and drug metabolite concentrations in serum and NAF, serum sulindac and its metabolites were all non-significantly correlated with a decrease in NAF PGEM levels (data not shown).

Figure 1
Change in nipple aspirate biomarker levels of PGEM and GDF15. Subject-level change in PGEM normalized by total protein is shown for (A) 150 mg q.d. sulindac and (B) 150 mg b.i.d. sulindac. Subject-level change in GDF15 normalized by total protein is shown ...
Table 3
Nipple aspirate concentration of PGEM and GDF15, pre and post sulindac and change (post – pre) by dose.

NAF GDF15 at baseline and change with exposure to sulindac

Mean and median total NAF GDF15 levels normalized to total protein at baseline were 0.82 and 0.45 ng/mg total protein, respectively. GDF15 concentrations at baseline ranged from 0–2.93 ng/mg total protein. As shown in Figure 1C & D, NAF GDF15 levels did not change in the 150 mg q.d. dose group (p = 0.300) but non significantly increased in the 150 mg b.i.d dose group (p = 0.097) As shown in Table 3, this difference between the dose groups achieved significance (p = 0.038). There were no statistically significant associations between serum levels of sulindac, sulindac sulfide, or sulindac sulfone levels and GDF15 concentrations in NAF (data not shown). Similarly, we observed no association between the concentration of sulindac in NAF or its sulfide metabolite and GDF15 levels.

Serum CRP at baseline and change with exposure to sulindac

Our initial attempts to assess CRP in NAF as a recognized NSAID response biomarker and positive control biomarker failed due to the lack of a linear performance of CRP ELISA assays for NAF samples. Western blot analyses of CRP in NAF suggest multiple uncharacterized species binding to the anti-CRP antibodies, limiting their use in this study (data not shown). As a substitute, we evaluated serum high sensitivity (hs)CRP levels as a surrogate marker of the systemic anti-inflammatory response of sulindac and used diagnostic laboratory obtained values for hsCRP. For the entire study group, serum hsCRP levels decreased by 0.59 mg/L in response to sulindac treatment but did not reach statistical significance (p = 0.15). Among subjects with values considered above normal at baseline (>1.0 mg/L; n = 15), a statistically significant decrease of 1.76 mg/L was observed (p = 0.02) (Figure 2). The decrease in serum hsCRP levels was significantly greater in women who received 150 mg b.i.d. than those in the lower dose group (p = 0.03), though our sample size became limiting. There were no statistically significant associations between serum levels of sulindac, sulfide, or sulfone levels and hsCRP in serum (data not shown).

Figure 2
Change in CRP according to baseline level


Our study showed, for the first time, that sulindac and its sulfide metabolite can be detected in about half of the evaluated NAF samples after sulindac treatment, suggesting the partitioning of sulindac and the sulfide metabolite to human breast tissue. In contrast, the sulfone metabolite was only detectable in 3 of 26 NAF samples (11.5%). Limited studies have examined the breast tissue distribution of sulindac. Sulindac is known to be secreted in rat milk; concentrations in milk were 10 – 20% of those levels in plasma (16). Kapetanovic et al. (17) determined the plasma and mammary tissue concentrations of sulindac and sulindac sulfone in rats. The maximum plasma concentrations of sulindac and sulfone were around 4 times higher than the maximum mammary tissue concentrations. In our study, the drug and sulfide concentrations in NAF were significantly lower than those detected in serum. The median NAF sulindac, sulfide, and sulfone concentrations were 39.8µg/ml, 46.3µg/ml, and undetectable, respectively. Compared with the average serum drug and metabolite concentrations of 2.94 µg/ml sulindac, 3.90 µg/ml sulfide, and 1.95 µg/ml sulfone, the values suggest about a 100-fold difference between NAF and serum levels. It is likely that the breast tissue concentrations of sulindac and its metabolites are higher than those detected in NAF because the secretion of these compounds from the fatty breast tissue to NAF may be limited by their high lipophilicity (18). This suggests that in the case of sulindac and possibly other lipophilic agents, NAF levels may not represent drug levels in the breast tissue.

Our study also explored the effects of sulindac on the putative sulindac drug targets PGEM and GDF15. These are considered exploratory endpoints because the study was powered only for the primary endpoint. We did not observe a consistent change in NAF PGEM levels with sulindac treatment. Sauter et al. (19, 20) determined the effects of celecoxib intervention at 400 mg b.i.d. or 200 mg b.i.d. on PGE2 levels in high risk women and women with breast cancer. They showed that celecoxib at 200 mg b.i.d. taken on average for 2 weeks had no effect on NAF PGE2 levels. Celecoxib at 400 mg b.i.d. significantly decreased NAF PGE2 levels in postmenopausal high-risk women and women with newly diagnosed breast cancer but did not result in consistent changes in NAF PGE2 in high-risk premenopausal women. Our study is limited by the small sample size to further stratify the analysis by menopausal status, though higher median PGEM levels were noted in premenopausal women. Collectively, these results suggest that suppression of NAF PGE2 may be observed only at higher doses of COX inhibitors or under conditions with less hormonal variation.

Our study showed the presence of GDF15 in NAF and a trend for induction of GDF15 in NAF with the higher dose of sulindac. This finding is consistent with the role of GDF15 as an NSAID inducible molecule (21) and with in vitro (22) and limited in vivo studies (23) supporting a strong drug dose-dependence on expression. These results support a direct effect of sulindac exposure on known molecular targets of sulindac in NAF.

In addition, our study assessed the effects of sulindac intervention on serum hsCRP levels. For hsCRP we were able to demonstrate a much stronger suppressive effect of sulindac in women whose hsCRP values were above a ‘risk’ relevant cutpoint of 1 mg/L, a cutpoint associated with cardiovascular disease (24). This finding suggests that sulindac drug effects may be detected only in individuals with abnormal expression of the biomarker of interest.

In conclusion, our results demonstrate that sulindac and its metabolites partition to human breast as measured by NAF levels and show trends of changes in one of the drug effect biomarkers in NAF. Follow-up chemoprevention trials are warranted to evaluate the effects of sulindac on biomarkers of tumorigenesis in the breast tissue.


Grant support: Funding for the study was provided by the National Cancer Institute, Division of Cancer Prevention, contract no. N01CN35158

We thank Betsy C. Wertheim for her assistance in careful review and editing of the manuscript. We thank Donna Vining, RN for her accrual of participants and Melissa May and Wade M. Chew for their technical support for the biomarkers.


1. Gill JK, Maskarinec G, Wilkens LR, Pike MC, Henderson BE, Kolonel LN. Nonsteroidal Antiinflammatory Drugs and Breast Cancer Risk: The Multiethnic Cohort. Am. J. Epidemiol. 2007;166:1150–1158. [PubMed]
2. Terry MB, Gammon MD, Zhang FF, Tawfik H, Teitelbaum SL, Britton JA, Subbaramaiah K, Dannenberg AJ, Neugut AI. Association of Frequency and Duration of Aspirin Use and Hormone Receptor Status With Breast Cancer Risk. JAMA. 2004;291:2433–2440. [PubMed]
3. Gierach GL, Lacey JV, Jr, Schatzkin A, Leitzmann MF, Richesson D, Hollenbeck AR, Brinton LA. Nonsteroidal anti-inflammatory drugs and breast cancer risk in the National Institutes of Health-AARP Diet and Health Study. Breast Cancer Res. 2008;10:R38. [PMC free article] [PubMed]
4. Harris RE, Beebe-Donk J, Alshafie GA. Reduction in the risk of human breast cancer by selective cyclooxygenase-2 (COX-2) inhibitors. BMC Cancer. 2006;6:27. [PMC free article] [PubMed]
5. Takkouche B, Regueira-Mendez C, Etminan M. Breast Cancer and Use of Nonsteroidal Anti-inflammatory Drugs: A Meta-analysis. J Natl Cancer Inst. 2008 Oct 15;100(20):1439–1447. [PubMed]
6. Crawford YG, Gauthier ML, Joubel A, Mantei K, Kozakiewicz K, Afshari CA, Tlsty TD. Histologically normal human mammary epithelia with silenced p16(INK4a) overexpress COX-2, promoting a premalignant program. Cancer Cell. 2004;5:263–273. [PubMed]
7. Berman H, Zhang J, Crawford YG, Gauthier ML, Fordyce CA, McDermott KM, Sigaroudinia M, Kozakiewicz K, Tlsty TD. Genetic and Epigenetic Changes in Mammary Epithelial Cells Identify a Subpopulation of Cells Involved in Early Carcinogenesis. Cold Spring Harbor Symposia on Quantitative Biology. 2005;70:317–327. [PubMed]
8. Singh-Ranger G, Salhab M, Mokbel K. The role of cyclooxygenase-2 in breast cancer: review. Breast Cancer Research and Treatment. 2008;109:189–198. [PubMed]
9. Diaz-Cruz ES, Shapiro CL, Brueggemeier RW. Cyclooxygenase Inhibitors Suppress Aromatase Expression and Activity in Breast Cancer Cells. J Clin Endocrinol Metab. 2005;90:2563–2570. [PubMed]
10. Hudson AG, Gierach GL, Modugno F, Simpson J, Wilson JW, Evans RW, Vogel VG, Weissfeld JL. Nonsteroidal anti-inflammatory drug use and serum total estradiol in postmenopausal women. Cancer Epidemiol Biomarkers Prev. 2008;17:680–687. [PubMed]
11. Albert KS, Gernaat CM. Pharmacokinetics of ibuprofen. Am J Med. 1984;77:40–46. [PubMed]
12. Giagoudakis G, Markantonis SL. Relationships between the concentrations of prostaglandins and the nonsteroidal antiinflammatory drugs indomethacin, diclofenac, and ibuprofen. Pharmacotherapy. 2005;25:18–25. [PubMed]
13. Cheng L, Kelly RW, Thong KJ, Hume R, Baird DT. The effect of mifepristone (RU486) on the immunohistochemical distribution of prostaglandin E and its metabolite in decidual and chorionic tissue in early pregnancy. J Clin Endocrinol Metab. 1993;77:873–877. [PubMed]
14. Kankuri E, Vaali K, Korpela R, Paakkari I, Vapaatalo H, Moilanen E. Effects of a COX-2 Preferential Agent Nimesulide on TNBS-Induced Acute Inflammation in the Gut. Inflammation. 2001;25:301–310. [PubMed]
15. Tanno T, Bhanu NV, Oneal PA, Goh S-H, Staker P, Lee YT, Moroney JW, Reed CH, Luban NLC, Wang R-H, Eling TE, Childs R, Ganz T, Leitman SF, Fucharoen S, Miller JL. High levels of GDF15 in thalassemia suppress expression of the iron regulatory protein hepcidin. Nat Med. 2007;13:1096–1101. [PubMed]
16. Østensen ME. Safety of non-steroidal anti-inflammatory drugs during pregnancy and lactation. Inflammopharmacology. 1996;4:31–41.
17. Kapetanovic IM, Krishnaraj R, Martin-Jimenez T, Yuan L, van Breemen RB, Lyubimov A. Effects of oral dosing paradigms (gavage versus diet) on pharmacokinetics and pharmacodynamics. Chem Biol Interact. 2006;164:68–75. [PubMed]
18. Bjarnason I, Scarpignato C, Takeuchi K, Rainsford KD. Determinants of the short-term gastric damage caused by NSAIDs in man. Aliment Pharmacol Ther. 2007 Jul 1;26(1):95–106. [PubMed]
19. Sauter ER, Qin W, Schlatter L, Hewett JE, Flynn JT. Celecoxib decreases prostaglandin E2 concentrations in nipple aspirate fluid from high risk postmenopausal women and women with breast cancer. BMC Cancer. 2006;6:248. [PMC free article] [PubMed]
20. Sauter ER, Qin W, Hewett JE, Ruhlen RL, Flynn JT, Rottinghaus G, Chen YC. Celecoxib concentration predicts decrease in prostaglandin E2 concentrations in nipple aspirate fluid from high risk women. BMC Cancer. 2008;8:49. [PMC free article] [PubMed]
21. Eling TE, Baek SJ, Shim M, Lee CH. NSAID activated gene (NAG-1), a modulator of tumorigenesis. J Biochem Mol Biol. 2006;39:649–655. [PubMed]
22. Baek SJ, Kim K-S, Nixon JB, Wilson LC, Eling TE. Cyclooxygenase Inhibitors Regulate the Expression of a TGF-{beta} Superfamily Member That Has Proapoptotic and Antitumorigenic Activities. Mol Pharmacol. 2001;59:901–908. [PubMed]
23. Kim JH, JC, Rhee K-H, Yoon J-H, Kwon SHo, Kun KS, Chang Il WL, Hyun CJ, Kim JK-S. Cyclooxygenase inhibitors induce apoptosis in sinonasal cancer cells by increased expression of nonsteroidal anti-inflammatory drug-activated gene. Intl. J. Cancer. 2008;122:1765–1773. [PubMed]
24. Mitka M. Panel endorses limited role for CRP tests. JAMA. 2003;289:973–974. [PubMed]