Nonsteroidal anti-inflammatory drug use and breast cancer risk in a European prospective cohort study

Abstract

Experimental studies have shown a protective effect of nonsteroidal anti-inflammatory drugs (NSAIDs) on breast cancer development. However, results from epidemiological cohort studies are less consistent. Our objective was to assess the association between NSAID use and breast cancer risk within the European Prospective Investigation into Cancer and Nutrition (EPIC).

EPIC is a prospective cohort study initiated in 1992 in 10 European countries. Self-reported information on NSAID use at baseline has been collected in five EPIC countries. Multivariable Cox regression models were used to estimate hazard ratios for the association of NSAID use with breast cancer incidence with adjustment for potential confounders. We also assessed effect modification by breast cancer risk factors and examined the associations within specific breast cancer subtypes.

Among the 140,981 women included in the analysis, 7% were regularly using NSAIDs at baseline. During a median follow-up time period of 13 years, 7,379 incident breast cancer cases were diagnosed (816 in situ and 6,563 invasive). There were no statistically significant associations between NSAID use and breast cancer risk, overall and by subtypes. However, a statistically significant interaction was observed for invasive cases between NSAID use and ever use of menopausal hormonal therapy (MHT) among postmenopausal women [MHT users: HR_{NSAID use}=0.84 (0.73 – 0.96); non MHT users: HR_{NSAID use} = 1.08 (0.93 – 1.25); P_interaction= 0.05].

Our results indicate potential effect modification of MHT use on the association between use of NSAIDs and breast cancer risk which deserves in-depth investigation in studies with accurate data on both NSAID and MHT use.

Introduction

Breast cancer is the most frequently diagnosed cancer and the most common cause of cancer death among women worldwide ¹. In past decades, evidence from both experimental and observational studies has accumulated suggesting that chronic inflammation could promote breast cancer development ². Therefore, nonsteroidal anti-inflammatory drugs (NSAIDs), a class of drugs commonly used to treat pain, fever, and inflammation, have been proposed as potential chemo-preventive agents for breast cancer ³. Although some experimental studies have shown a protective effect of NSAIDs on breast cancer development ⁴, epidemiological evidence is less consistent. Indeed, recent meta-analyses have found a statistically significant decreased risk of breast cancer among NSAID users in case-control studies but not in cohorts where NSAID use was assessed prior to breast cancer onset ⁵. It also remains unclear whether the potential association between NSAID use and breast tumour would vary by breast cancer subtypes and by known breast cancer risk factors. We therefore examined the associations between regular NSAID use and breast cancer risk among women enrolled in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort, overall and by breast cancer subtypes, and according to various breast cancer risk factors.

Materials and Methods

EPIC cohort

EPIC is a multi-center prospective cohort study designed to investigate the relationships between diet, nutrition, metabolic factors and cancer and other chronic diseases. The study enrolled more than 521,000 participants, mostly aged 35–70 years, between 1992 and 2000, from 23 regional and national research centers located in 10 Western European countries: Denmark, France, Germany, Greece, Italy, The Netherlands, Norway, Spain, Sweden, and The United Kingdom.

At baseline, participants completed questionnaires on diet and lifestyle including reproductive and medical history, exogenous hormone use, occupation, education and physical activity. Participants were also invited to a study center for blood collection, anthropometric measurements, and blood pressure measurement. Extensive details on the standardized recruitment procedures are provided elsewhere ⁶.

Study Population

Female participants in EPIC were eligible for inclusion in this study if they had no history of cancer prior to recruitment, and no missing date of cancer diagnosis or censoring. Only the five countries that have provided data on NSAID use were included in the current analysis: The Netherlands (Utrecht center), France (all centers), Denmark (all centers), Germany (all centers), and United Kingdom (Norfolk center). Women were further excluded if exposure was missing at baseline, and if they were diagnosed with non-epithelial breast cancer. A total of 140,981 women were therefore included in the analyses.

Follow-up

In the Netherlands, Denmark and the United Kingdom, incident breast cancer cases were identified using record linkage with cancer and pathologies registries. In France and Germany, cases were identified through regular direct contacts with participants and their next of kin or through national cause-of-death registries. All reported breast cancer cases were then systematically verified against clinical and pathological records. Vital status was collected from regional or national mortality registries. Cancer incidence data were classified according to the International Classification of Diseases, 10^th Revision (ICD-10). Women were followed from study entry until the date of last known contact, cancer diagnosis, death, emigration or the end of the follow-up period (between 2008 and 2014 depending on the center), whichever came first.

Exposure assessment

Data on NSAID use were collected at baseline by questionnaire. The following questions were asked to the participants:

• -“Are you currently taking anti-inflammatory drugs at least three times per week?” in France.
• -“Have you ever taken aspirin continuously for 3 months or more?” in The United Kingdom.
• -“During the past 4 weeks, did you take any regular medications? If yes, please specify the medication.” in Germany.
• -“Do you take aspirin daily?” in The Netherlands.
• -“Have you used more than one painkilling tablet/month during the last year? If yes, how many pills containing aspirin and how many pills containing ibuprofen: none; 2–3/month; 1–2/week; 3–6/week; 1–3/day; 4–5/day; 6+/day?” in Denmark.

We defined “regular use” of NSAID by combining participants responding “yes” to the question in France, in the United Kingdom, and in the Netherlands, those who cited NSAID and/or aspirin as regular medications in Germany, and people responding at least “3 pills per week “ for using aspirin or ibuprofen in Denmark.

Statistical analysis

Comparison of participants’ characteristics according to NSAID use have been assessed by chi-2 test. Associations between regular NSAID use and breast cancer risk were evaluated using Cox proportional hazard models to estimate hazard ratios (HR) and 95% confidence intervals (CI), overall and by subtypes (invasive / in situ, invasive ductal / invasive lobular, invasive oestrogen receptor (ER)-positive / invasive ER-negative, invasive progesterone receptor (PR)-positive / invasive PR-negative, invasive ER/PR-positive / invasive ER/PR-negative). Heterogeneity between breast cancer subtypes was evaluated by competing risk analyses, with women who developed the competing breast cancer subtypes censored at the time of occurrence and cases with missing information on the studied subtype excluded from the corresponding analysis ⁷. All models were stratified by age in one-year intervals and by study center to prevent violations of the proportional-hazards assumption. Of the breast cancer risk factors listed in Table 1, only those that modified the HR estimate between NSAID use and breast cancer risk by more than 10% were included in the models. The final adjusted models therefore included: body mass index (BMI) (continuous), educational level (primary/no schooling, technical/professional/secondary, longer education, missing), menopausal hormone therapy (MHT) use (never, former, current, missing), age at first full term pregnancy (nulliparous, ≤24 years, >24 years, missing), and alcohol intake (non-drinker, >0-3 g/d, >3-12 g/d, >12-24 g/d, >24-60 g/d, >60 g/d, missing). A complete case analysis that excluded women with missing covariates was conducted and yielded similar results (data not shown). We also examined whether the NSAID use and breast cancer associations varied according to BMI (<25 kg/m², ≥25 kg/m²), waist/hip ratio (<0.85, ≥0.85), level of physical activity (Cambridge index: inactive, moderately inactive, moderately active, active), educational level (primary/no schooling, technical/professional/secondary, longer education), use of contraceptive pills (never, former, current and never, ever), full term pregnancy (no, yes), age at first full term pregnancy (<24 years, ≥24 years) and number of full term pregnancies (1, ≥2) among parous women, breastfeeding among parous women (no, yes), menopausal status (premenopausal, postmenopausal), MHT use (never, former, current and never, ever) among postmenopausal women, type of MHT used (oestrogen alone, oestrogen-progestagen) and alcohol intake (non-drinker, > 0–3 g/d, >3–12 g/d, >12 g/d). Effect modification was evaluated by using likelihood ratio tests to compare models with and without cross-product interaction terms. Two-sided P values of less than 0.05 were considered to be statistically significant. All analyses were carried out using the SAS software system (SAS institute, INC., Cary, NC), version 9.4.

Table 1. Selected participant characteristics according to NSAID use at baseline.

Characteristics	Total, N	Regular use of NSAIDs, n (%) 1
		No	Yes
Total	140,981	131,074 (93)	9,907 (7)
Age at recruitment, missing N=0 ≤ median (53 years) > median (53 years)	66,923 74,058	64,027 (49) 67,047 (51)	2,896 (29) 7,011 (71)
Menopausal status, missing N=0 Premenopausal Postmenopausal	33,865 107,116	32,833 (25) 98,241 (75)	1,032 (10) 8,875 (90)
Country, missing N=0 France United Kingdom The Netherlands Germany Denmark	61,388 12,505 10,812 27,908 28,368	57,710 (44) 11,405 (9) 10,413 (8) 26,500 (20) 25,046 (19)	3,678 (37) 1,100 (11) 399 (4) 1,408 (14) 3,322 (34)
Educational level, missing N=4,037 primary/no schooling, technical/professional/secondary, longer education	29,732 72,287 34,925	26,483 (21) 67,506 (53) 33,354 (26)	3,249 (34) 4,781 (50) 1,571 (16)
Family history of breast cancer, missing N=67,265 No Yes	67,122 6,594	63,068 (91) 6,227 (9)	4,054 (92) 367 (8)
BMI (WHO categories), missing N=0 <18.5 (Underweight) 18.5 – 24.9 (Normoweight) 25 – 29.9 (Overweight) ≥ 30 (Obese)	3,378 85,601 37,774 14,228	3,192 (2) 80,762 (62) 34,564 (26) 12,556 (10)	186 (2) 4,839 (49) 3,210 (32) 1,672 (17)
Physical activity, missing N=967 Inactive Moderately inactive Moderately active Active	24,695 50,930 39,094 25,295	22,164 (17) 47,448 (36) 36,875 (28) 23,697 (18)	2,531 (26) 3,482 (35) 2,219 (23) 1,598 (16)
Alcohol intake, missing N=2,131 Non-drinker >0 – 3 g/d >3 – 12 g/d >12 – 24 g/d >24 g/d	14,929 36,189 46,489 23,552 17,691	13,768 (11) 33,433 (26) 43,474 (34) 22,035 (17) 16,453 (13)	1,161 (12) 2,756 (28) 3,015 (31) 1,517 (16) 1,238 (13)
Age at menarche, missing N=2,959 ≤13 years > 13 years	83,712 54,310	78,157 (61) 50,183 (39)	5,555 (57) 4,127 (43)
Use of contraceptive pills, missing N=1,449 Never Former Current	51,034 82,965 5,533	46,653 (36) 77,730 (60) 5,388 (4)	4,381 (45) 5,235 (54) 145 (1)
Use of MHT2,missing N=3,028 Never Former Current	57,390 13,924 32,774	53,012 (56) 12,535 (13) 29,903 (31)	4,378 (51) 1,389 (16) 2,871 (33)
Full term pregnancy, missing N=4,554 No Yes	15,819 120,608	14,727 (12) 112,147 (88)	1,092 (11) 8,461 (89)
Age at first full term pregnancy3, missing N=489 ≤ 24 years > 24 years	66,084 54,035	60,958 (55) 50,753 (45)	5,126 (61) 3,282 (39)
Number of full term pregnancies3, missing N=22 1 ≥ 2	23,776 96,810	22,134 (20) 89,992 (80)	1,642 (19) 6,818 (81)
Breastfeeding3, missing N=2,489 Never Ever	23,247 94,872	21,769 (20) 88,138 (80)	1,478 (18) 6,734 (82)

¹Variations between users and non-users of NSAIDs have been assessed by chi-2 test. All the p-value were < 0.0001, except for family history of breast cancer (p=0.12), full term pregnancy (p=0.60), and number of full term pregnancies (p=0.46).

²Among postmenopausal women

³Among parous women.

Results

Of the 140,981 women included in the current analysis, 9,907 (7%) were regular NSAID users at baseline (Table 1). Characteristics of participants according to NSAID use at baseline are shown in table 1.

During a median follow-up time period of 13.1 (SD=3.8) years, 7,379 incident breast cancer cases were diagnosed (816 in situ and 6,563 invasive). Among those, 5,360 invasive breast cancer cases had information on ER status (4,405 ER-positive, 955 ER-negative) and 4,290 on PR status (2,806 PR-positive, 1,484 PR-negative). In multivariable models, there was no statistically significant association between regular NSAID use and breast cancer risk overall (Table 2). Compared with women who were not regular users at baseline, the HR for regular users of NSAIDs was 0.99 (95% CI= 0.90 to 1.08). Risk estimates did not vary significantly by histological subtypes or by hormone receptor status of the tumour (all P_{heterogeneity} ≥0.36, data not shown) but there was a borderline significant heterogeneity between in situ and invasive breast cancers [HR_insitu=1.27 (0.98 – 1.65); HR_invasive=0.96 (0.87 – 1.05); P_{heterogeneity}=0.05].

Table 2. Associations between regular use of NSAIDs and the risk of breast cancer, overall, by breast cancer subtypes, and stratified by MHT use.

Breast Cancer	Regular use of NSAIDs
	All Women (N=140,981)	Postmenopausal women (N=104,088)
		Never use of MHT	Ever use of MHT
	N exposed cases HR1 (95% CI)	N exposed cases HR2 (95% CI)	N exposed cases HR2 (95% CI)	Pinteraction3
Overall	540	217	255
	0.99 (0.90 – 1.08)	1.09 (0.95 – 1.26)	0.87 (0.76 – 0.99)	0.06
Invasive	477	194	225
	0.96 (0.87 – 1.05)	1.08 (0.93 – 1.25)	0.84 (0.73 – 0.96)	0.05
In situ	63 1.27 (0.98 – 1.65)	23 1.25 (0.82 – 1.92)	30 1.18 (0.81 – 1.73)	0.99
Invasive Ductal	351 0.97 (0.87 – 1.08)	141	168
		1.09 (0.90 – 1.28)	0.87 (0.74 – 1.02)	0.16
Invasive Lobular	74	24	38
	0.98 (0.77 – 1.24)	0.96 (0.63 – 1.46)	0.86 (0.62 – 1.21)	0.92
Invasive ER+	307	118	150
	0.93 (0.82 – 1.04)	1.02 (0.85 – 1.24)	0.82 (0.69 – 0.97)	0.19
Invasive ER-	72	34	29
	1.05 (0.82 – 1.34)	1.39 (0.97 – 2.00)	0.77 (0.53 – 1.13)	0.04
Invasive PR+	181	73	83
	0.94 (0.81 – 1.10)	1.11 (0.87 – 1.41)	0.78 (0.62 – 0.98)	0.09
Invasive PR-	92	33	45
	0.91 (0.74 – 1.13)	0.94 (0.66 – 1.34)	0.80 (0.59 – 1.08)	0.69
Invasive ER+PR+	168 0.91 (0.78 – 1.07)	67	79
		1.06 (0.82 – 1.36)	0.77 (0.61 – 0.97)	0.16
Invasive ER-PR-	44 0.91 (0.68 - 1.24)	19	19
		1.14 (0.71 – 1.83)	0.70 (0.43 – 1.11)	0.23

¹Hazard Ratio stratified for recruitment center and age at baseline, and adjusted for BMI (continuous), educational level (primary/no schooling, technical/professional/secondary, longer education, missing), MHT use (never, former, current, missing), age at first full term pregnancy (nulliparous, ≤ 24 years, > 24 years, missing) and alcohol intake at recruitment (non-drinker, >0-3g/d, >3-12g/d, >12-24g/d, >24-60g/d, >60g/d, missing).

²Hazard Ratio adjusted for all the above but MHT use.

³Likelihood ratio tests comparing models with and without cross-product interaction terms of NSAID x MHT use.

The association between regular NSAID use and breast cancer risk did not vary significantly by breast cancer risk factors (all P_interactions between 0.10 and 0.96) or by country P_{heterogeneity}=0.95; data not shown), however there was a statistically significant interaction by MHT use (Table 2 and Figure 1). Regular NSAID use was inversely associated with overall breast cancer risk among postmenopausal women who had ever used MHT [HR_MHT =0.87 (0.76 – 0.99)] but not among never MHT users [HR_noMHT = 1.09 (0.95 – 1.26); P_interaction=0.06]. The test for interaction reached statistical significance for invasive breast cancer cases [HR_MHT=0.84 (0.73 – 0.96); HR_noMHT = 1.08 (0.93 – 1.25); P_interaction= 0.05], but was not statistically significant for in situ cases [HR_MHT=1.25 (0.82 – 1.92); HR_noMHT = 1.18 (0.81 – 1.73); P_interaction =0.99]. A similar pattern of associations by regular NSAID and ever MHT use was observed for all other invasive breast cancer subtypes examined, although the test for interaction only reached statistical significance among invasive ER- cases (data not shown). Figure 1 shows the results of the joint analysis of the effects of regular NSAID use and ever MHT use on invasive breast cancer risk among postmenopausal women. Ever MHT users had a significant increased risk of invasive breast cancer compared to non-MHT/non-NSAID users [HR_{NO NSAID-MHT} = 1.38 (1.30 – 1.47)]. However, the risk estimate was significantly attenuated among women who also regularly used NSAIDs [HR_NSAID-MHT=1.19 (1.03– 1.37)]. For ER- cases, no significant increased risk was observed among women who had used both MHT and NSAIDs [HR_NSAID-MHT=1.07 (0.73– 1.58)]. The association between regular NSAID use and breast cancer risk (all subtypes) did not vary significantly between former and current MHT users and between different types and durations of MHT used at baseline (data not shown).

Figure 1. Associations between regular NSAID use, ever MHT use among postmenopausal women, and the risk of invasive (A), invasive ER+ (B), and invasive ER- (C) breast cancers.

¹ Hazard Ratio stratified for recruitment center and age at baseline, and adjusted for BMI (continuous), educational level (primary/no schooling, technical/professional/secondary, longer education, missing), age at first full term pregnancy (nulliparous, ≤ 24 years, > 24 years, missing) and alcohol intake at recruitment (non-drinker, >0-3g/d, >3-12g/d, >12-24g/d, >24-60g/d, >60g/d, missing).

Discussion

In this prospective cohort study there was no association between regular NSAID use and breast cancer risk (overall and by tumour subtypes). However, regular NSAID use was associated with a lower risk of invasive breast cancer diagnosis among MHT ever users.

Of the previous prospective studies that examined the NSAID-breast cancer association, only a few found a statistically significant inverse association between NSAID use with breast cancer ⁵. These inconsistent findings could be due to differences in the accuracy of the exposure assessment (measured at baseline or updated during follow-up), differences in the exposure period definition (ever use, regular use, long-term use, etc.), or in the exposure subtypes (all NSAIDs, all NSAIDs but aspirin, low-dose aspirin, aspirin, ibuprofen, COX-2 inhibitors, etc.). The associations between breast cancer and NSAID use could be limited to specific breast cancer subtypes. In fact, stronger inverse associations with NSAID use have been found for ER+ and/or PR+ breast cancer in some studies ⁵. In our study, we did not observe differences according to breast cancer subtypes but the number of cases in some subtypes was small and our power to detect differences may be have limited. The NSAID-breast cancer association could also be limited to certain sub-population. In particular, our results suggest that the reported associations may depend on MHT use.

Estrogen-progestagen MHT use is a well-established risk factor for breast cancer ⁸ and any evidence of interaction between MHT and NSAID use in breast cancer risk would be of importance for public health. To our knowledge, only eight studies have examined the modifying effect of MHT use on the association between regular NSAID use and breast cancer risk. Seven of these studies did not observe any significant effect modification ^9–15 and only one study reported an inverse association between NSAID use for at least five years and breast cancer risk among MHT users ¹⁴. In the other six studies, risk estimates associated with NSAID use among MHT users were not reported. One other study showed a statistically significant positive association with regular/extra-strength aspirin use and breast cancer risk among MHT users for ten years or more ¹⁶. Nevertheless, this result should be considered cautiously because the number of exposed cases among MHT users for 10 years or more was very low (n=25).

It has been suggested that NSAIDs could reduce breast cancer risk through the inhibition of the cyclooxygenase-2 (COX-2) enzymes ¹⁷. Inactivating COX-2, that has been found to be overexpressed in breast cancer tissue when compared to normal tissue, may reduce the downstream synthesis of pro-inflammatory prostaglandins (PG), particularly PGE₂, a potent mitogen. This reduction results in inhibition of angiogenesis, proliferation and in the promotion of apoptosis ¹⁸. There is also accumulating evidence that the inhibition of COX-2 and PGE₂ is associated with the suppression of estrogen synthesis via a decrease in aromatase activity ¹⁷. Consistently, cross-sectional investigations among postmenopausal women, who were not recently exposed to exogenous hormones, suggest that current NSAID use may be associated with lower circulating estradiol levels ¹⁹. However, results of a recent randomized controlled clinical trial have shown no effect of intake of 325 mg/day aspirin over 6 months on estradiol levels, among postmenopausal women who were not exposed to exogenous hormones ²⁰.

To date, the potential effect of the combination of NSAIDs and MHT on breast tissues has not been evaluated. However, MHT could modify exposure to various hepatically-derived proteins including many cytokines due to the hepatic first pass effect ^{21, 22}. MHT may also reduce insulin sensitivity which would be expected to increase breast cancer risk ²³. NSAIDs can reduce insulin resistance ²⁴ and it is therefore possible that NSAID use in MHT users reduces breast cancer risk by targeting some aspects of the insulin signaling pathways. An in vitro study has demonstrated that the combination of progestin and NSAIDs synergistically inhibits cell growth and induces apoptosis in the ovarian epithelium ²⁵. To our knowledge, there are no published studies examining the possible effect of the combination of exogenous estrogens and NSAIDs on the cancer development.

The main strength of this study is the use of a large European population-based prospective cohort, followed-up for a median period of 13 years. The prospective design minimizes recall bias and the large number of participants and incident breast cancer cases allowed us to examine interactions of NSAID use with breast cancer risk factors. Nevertheless, because of the number of comparisons evaluated, the effect modification by MHT may be a chance finding. The number of cases, although large overall, remained nevertheless limited for some breast cancer subtypes. Thus, results for ER/PR status have to be considered with caution and replicated. As in many other cohorts, the main limitation of our study is the poor accuracy and heterogeneity in the assessment of NSAID use across countries. First, we used self-reported data, and NSAID exposure was assessed only once at baseline which could lead to a misclassification of longer term exposure. Second, we had no information on the type, frequency, duration, and dosage of NSAID use. Thus, we were not able to characterize lifetime NSAID use to examine a dose-response relationship or to study the specific effect of each NSAID on breast cancer development. A further limitation of our study is the large number of missing data for family history of breast cancer, however, in previous studies it did not appear to modify the relationship between NSAID use and breast cancer risk ¹². Furthermore we performed a sensitivity analysis excluding women with family history of breast cancer in the three EPIC countries with data available and estimates were virtually unchanged.

Conclusion

In this analysis of 140,981 women, NSAID use was associated with decreased invasive breast cancer risk in MHT ever users only. Our results require confirmation in other prospective cohorts that have collected accurate data on NSAID and MHT use, including type, dosage, frequency and duration of use.

Novelty and Impact.

Results from epidemiological studies on nonsteroidal anti-inflammatory drug (NSAID) use and breast cancer risk are not consistent. In the EPIC prospective cohort we showed that NSAID use was associated with a decreased risk of invasive breast cancer in MHT ever users. The interaction between MHT use, NSAIDs and breast cancer risk deserves replication and further investigation of the underlying mechanisms. If confirmed, it may modify the risk/benefit ratio for NSAID use among MHT users.