Breast Cancer Research and Treatment

, Volume 150, Issue 2, pp 395–403

Smoking and survival in female breast cancer patients

  • Alicia Padron-Monedero
  • Stacey L. Tannenbaum
  • Tulay Koru-Sengul
  • Feng Miao
  • Damien Hansra
  • David J. Lee
  • Margaret M. Byrne
Epidemiology

DOI: 10.1007/s10549-015-3317-3

Cite this article as:
Padron-Monedero, A., Tannenbaum, S.L., Koru-Sengul, T. et al. Breast Cancer Res Treat (2015) 150: 395. doi:10.1007/s10549-015-3317-3

Abstract

The purpose of this study was to determine if smoking affects survival in female breast cancer patients, both overall and stratified by race, ethnicity, and socioeconomic status. We linked data from the 1996–2007 Florida cancer data system, the Florida Agency for Health Care Administration, and the U.S. census. Inclusion criteria were females ≥18 years, diagnosed with breast cancer, and residing in Florida (n = 127,754). To analyze the association between smoking and survival, we performed sequential multivariate Cox proportional hazard regression models with progressive adjustment for main confounders. Compared to never smokers, worse survival was found in current (hazard ratio 1.33; 95 % CI 1.28–1.38) and former smokers (1.09; 1.06–1.13). Those who smoked <1, 1–2, and >2 packs/day had worse survival (HR 1.28; 1.20–1.36; HR 1.40; 1.33–1.47 and 1.70; 1.45–1.99, respectively) (p for linear trend <0.001), than never smokers. Among Whites, current and former smokers had worse survival (HR 1.38; 1.33–1.44 and HR 1.11; 1.07–1.15, respectively) than never smokers. Worse survival was also found for current and former smokers (HR 1.34; 1.29–1.40 and HR 1.10; 1.06–1.15, respectively) compared with never smokers among non-Hispanics; similarly, worse survival was found among current Hispanic smokers (HR 1.13; 1.01–1.26). The association was not significant for Blacks. Current smoking is associated with worse survival in White breast cancer patients and through all socioeconomic status categories and ethnicities compared to never smoking. Former smoking is associated with worse survival in White and non-Hispanic females. Blacks had similar survival regardless of smoking status. Nonetheless, all female breast cancer patients should be advised to quit smoking.

Keywords

Female breast cancer Smoking Survival Health inequalities 

Abbreviations

SES

Socioeconomic status

FCDS

Florida cancer data system

AHCA

Agency for Health Care Administration

HR

Hazard ratio

CI

Confidence interval

Introduction

Breast cancer is the most common cancer among U.S. women, after non-melanoma skin carcinoma, and the second leading cause of cancer death [1]. Approximately 232,340 new cases of invasive breast cancer and 39,620 breast cancer deaths are expected to occur among U.S. women in 2013 [2]. From 2006–2010, the incidence rates for White, African Americans, and Hispanics in the US were 127.3, 118.4, and 91.1 per 100,000, and mortality rates were 22.7, 30.8, and 14.8, per 100,000, respectively [2]. Moreover, breast cancer is the leading cause of cancer death in Hispanic women [1].

Recent studies have suggested that ever smoking may increase the risk for breast cancer, [3] but, to date, there is no conclusive evidence that tobacco use influences breast cancer mortality in women [3]. A limited number of studies found a link between smoking and survival in female breast cancer patients [4, 5, 6, 7, 8, 9]. However, these studies did not adjust for co-morbidities that can have a considerable impact on female breast cancer survival; and smokers have more co-morbidities than non-smokers [3].

Disparities by race in female breast cancer survival have been described in the literature, with Blacks demonstrating worse survival than Whites [9, 10, 11, 12, 13, 14]. However, the possible effect of smoking has scarcely been studied on different racial, ethnic, and socioeconomic groups; only one study analyzed racial differences [8] and as far as we could discern, no study assessed this association among ethnicity or socioeconomic status (SES). Because smoking is a modifiable behavior, protocols for breast cancer smokers in tertiary prevention and clinical management could be of benefit, particularly for subpopulations at higher risk for poorer outcomes. Thus, the objective of this study was to determine if smoking is associated with survival in female breast cancer patients and to assess possible associations by race, ethnicity, and SES.

Methods

Study design and population

Our data were obtained from the linkage of two population-based databases, the Florida cancer data system (FCDS) and the Agency for Health Care Administration (AHCA) with the U.S. census to form an extensive dataset of Floridian women diagnosed and treated for breast cancer from 1996–2007. In the state of Florida and as required by law, the FCDS registry collects population-based information (demographics, diagnosis, clinico-pathological, and therapeutic data) using unique identifiers to protect patient confidentiality. All Florida patient encounters at hospitals and freestanding ambulatory surgical and radiation clinics are documented in the AHCA database. We used de-identified data from a large database, and we had no participant interaction.

Inclusion criteria for this study were female patients diagnosed with breast cancer in the state of Florida during the years 1996–2007 (n = 177,058). Patients younger than 18 years of age (n = 12), non-Florida residents (n = 9,277), and with carcinoma in situ (n = 29,055) were excluded. We also excluded those with missing values for race, ethnicity, or SES (n = 10,960) resulting in a total sample size of 127,754. Because of the large number of unknown smoking status (20.5 % of the sample), we did not exclude these patients to avoid bias in our results because it is possible that the unknown/missing smoking status cases are fundamentally different from cases for which status is known. For example, cases that are missing cancer registry data are typically Black and more impoverished [15]. Finally, we excluded from the regression models (but include in the descriptive tables) 10,751 individuals with unknown status for co-morbidities; this is because the regression models would not converge if they were included. Thus, the final analytical sample included 117,003 individuals.

Study variables

Overall survival, the primary outcome variable, was defined as the time from diagnosis to death or last treatment encounter. We followed the cohort for an additional three-year period until 2010 to more fully capture survival. Self-reported smoking status was classified as current, former, never, or unknown at the time of diagnosis. In a separate analysis, we also classified smokers by intensity (never, <1, 1–2, >2 packs smoked/day; excluding former smokers).

Our main sociodemographic variables of interest were race (White, Black, Native American, Asian, Pacific Islander, Asian Indian or Pakistani, other for descriptive information; and White Black, other for inferential statistics), Ethnicity (Hispanic or non-Hispanic), and SES. SES was categorized from the U.S. census tract-level information on percent of households at the tract level living below the federal poverty line. Each tract was categorized as being in the lowest (≥20 %), middle-low (≥10 and <20 %), middle-high (≥5 and <10 %), or highest (<5 %) SES group. Other sociodemographic variables were age at diagnosis (shown descriptively as continuous and categorical variables [≤40, 41–50, 51–65, and >65 years] and used in regression models as a continuous variable) and marital status (married, divorced/separated/widowed, or never married). Women were considered to be obese if their body mass index (calculated as weight in kilograms divided by squared height in meters) was >30. Obesity, alcohol abuse, and the rest of the Elixhauser co-morbidities were categorized as yes, no, or unknown. The number of Elixhauser co-morbidities were also summed as an aggregated variable called “Co-morbidities” and coded as 0, 1–2, 3–4, and >4 for descriptive purposes.

The cancer stage at diagnosis was categorized as (1) localized, (2) regional lymph node only, (3) regional direct extension with or without lymph node involvement, (4) distant metastasis, or (5) unknown or unstaged according to the Surveillance, Epidemiology, and End Results (SEER) Program’s summary stage 2000 and summary stage 1977. We used the SEER 2000 classification and only applied the SEER 1977 classification when we encountered missing data.

The histology of the tumor was divided into the following categories “CA ductal,” “CA lobular,” and “Other” that included the rest of the histologic diagnoses provided. We grouped differentiation grade into the following categories: “well differentiated,” “moderately differentiated,” “poorly differentiated,” and “undifferentiated.” We added another category as “unknown” when no information was collected about the differentiation grade. The ER and PR hormonal receptor status were both classified as “positive,” “negative,” and “unknown,” and the treatments (chemotherapy, radiation therapy, surgery, and hormonal treatment) were all classified as “yes,” “no,” or “unknown.”

Statistical analyses

The association between categorical variables was examined using Chi-square tests. Log rank tests were used to examine the unadjusted association of smoking status and overall survival. Cox proportional hazard regressions were used to calculate survival hazard ratios controlling for confounders.

Using Cox models to assess the association between smoking status and survival, we constructed four block sequential models: (1) univariate model (Model A); (2) multivariate model adjusting for all sociodemographic variables, including race, ethnicity, SES, age, and marital status (Model B); (3) Model B plus tumor stage, histology, differentiation grade, ER/PR status, chemotherapy, radiation, surgery, and hormonal treatment (Model C); and (4) Model C plus obesity, alcohol abuse, and the other Elixhauser co-morbidities as individual variables (Model D). We tested for possible interactions between smoking and race, smoking and ethnicity, and smoking and SES in Cox survival models.

As the fully adjusted model (Model D) showed that current smokers had a significantly increased risk compared to never smokers, we replicated the sequential logistic regression models to assess the association between smoking intensity (<1pack/day, 1–2 packs/day, >2 packs/day) and survival in breast cancer patients using as the reference group those who never smoked (excluding former smokers).

Finally, to assess the association between smoking status and survival in breast cancer patients for different subpopulations, we stratified our sample by race (White, Black, or other), ethnicity (Hispanic or non-Hispanic), and SES (lowest, middle-low, middle-high, or highest) and re-ran the fully adjusted models for each.

Because the patients being treated at the same facility are not independent observations due to clustering at the facility level, we used robust standard errors for all analyses. Statistical significance was assessed at p ≤ 0.05. The SAS statistical software version 9.2 (SAS Institute Inc., Cary, NC) was used to perform all analyses. The study was approved by the Institutional Review Board of the University of Miami and the Florida Department of Health.

Results

The demographic and clinical characteristics of female breast cancer cases by self-reported smoking status are depicted in Tables 1 and 2. Of the 127,754 patients included in the analysis, the mean age at diagnosis was 63.5 years and 90.4 % were White, 8.5 % Black, and the rest belonged to other races. The mean follow-up was 4.9 years with a range of 0–15.0 years; 29.8 % of the patients of our sample (38,054 patients) died during follow-up.
Table 1

Demographics characteristics by smoking status of female breast cancer patients in Florida

 

Smoking status

 

All

Never

Former

Current

Unknown

n

Column %

n

Row %

n

Row %

n

Row %

n

Row %

Total sample

127,754

100

62,361

48.8

23,365

18.3

15,793

12.4

26,235

20.5

Deaths

38,054

29.8

17,333

45.5

6849

18

5,015

13.2

8857

23.3

Age at diagnosis (years)

 Mean (SD)

63.5 (14.3)

63.7 (14.7)

65.0 (12.7)

58.2 (12.7)

64.9 (14.9)

 Median (25th%, 75th%)

64.0 (52.75)

65.0 (52.75)

66.0 (56.75)

58.0 (48.68)

66.0 (54.76)

 Min, max

19,105

19,105

22,100

19,100

20,104

Age at diagnosis (years)

 ≤40

7161

5.6

3880

54.2

729

10.2

1184

16.5

1368

19.1

 41–50

19,943

15.6

9668

48.5

2802

14.1

3719

18.6

3754

18.8

 51–65

40,398

31.6

18,697

46.3

7677

19

6174

15.3

7850

19.4

 >65

60,252

47.2

30,116

50

12,157

20.2

4716

7.8

13,263

22

Race

 White

115,506

90.4

55,064

47.7

22,042

19.1

14,621

12.7

23,779

20.6

 Black

10,820

8.5

6,395

59.1

1,210

11.2

1,100

10.2

2,115

19.5

 Native American

60

0

31

51.7

7

11.7

11

18.3

11

18.3

 Asian

656

0.5

460

70.1

41

6.3

34

5.2

121

18.4

 Pacific Islander

50

0

32

64

9

18

1

2

8

16

 Asian Indian or Pakistani

212

0.2

140

66

15

7.1

5

2.4

52

24.5

 Other

450

0.4

239

53.1

41

9.1

21

4.7

149

33.1

Ethnicity

 Non-Hispanic

115,514

90.4

54,563

47.2

22,161

19.2

14,753

12.8

24,037

20.8

 Hispanic

12,240

9.6

7798

63.7

1204

9.8

1040

8.5

2198

18

SESa

 Lowest

14,381

11.3

7683

53.4

1965

13.7

1884

13.1

2849

19.8

 Middle-low

37,382

29.3

18,296

48.9

6346

17

5066

13.6

7674

20.5

 Middle-high

48,356

37.9

22,860

47.3

9353

19.3

5857

12.1

10,286

21.3

 Highest

27,635

21.6

13,522

48.9

5,701

20.6

2986

10.8

5426

19.6

Marital status

 Never Married

14,111

11

6531

46.3

2018

14.3

2351

16.7

3211

22.8

 Married

69,433

54.3

35,180

50.7

13,180

19

7710

11.1

13,363

19.2

 Divorced/Separated/Widowed

40,567

31.8

19,203

47.3

7607

18.8

5308

13.1

8449

20.8

 Unknown

3643

2.9

1447

39.7

560

15.4

424

11.6

1212

33.3

aSES, Socioeconomic status

Table 2

Clinical characteristics of female breast cancer patients in Florida

 

All

Smoking status

Never

Former

Current

Unknown

n

Column  %

n

Row %

n

Row %

n

Row %

n

Row %

Co-morbidities

 None

10,763

8.4

5106

47.4

1516

14.1

983

9.1

3158

29.3

 1–2

18,156

14.2

9078

50

3033

16.7

2309

12.7

3736

20.6

 3–4

25,737

20.1

13,074

50.8

4496

17.5

2978

11.6

5189

20.2

 >4

73,098

57.2

35,103

48

14,320

19.6

9523

13

14,152

19.4

Obesity

 Yes

14,273

11.2

7222

50.6

2829

19.8

1570

11

2652

18.6

 No

102,730

80.4

50,047

80.3

19,022

81.4

13,250

83.9

20,411

77.8

 Unknown

10,751

8.4

5092

8.2

1514

6.5

973

6.2

3172

12.1

Alcohol abuse

 Yes

2736

2.1

668

24.4

541

19.8

1066

39

461

16.8

 No

114,267

89.4

56,601

90.8

21,310

91.2

13,754

87.1

22,602

86.2

 Unknown

10,751

8.4

5092

8.2

1514

6.5

973

6.2

3172

12.1

SEER

 Localized

75,269

58.9

37,653

50

14,787

19.6

9223

12.3

13,606

18.1

 Regional, direct extension ± lymph nodes

5981

4.7

3160

52.8

950

15.9

860

14.4

1011

16.9

 Regional, lymph nodes only

27,781

21.7

14,475

52.1

5,137

18.5

3804

13.7

4,365

15.7

 Distant

5519

4.3

2640

47.8

1065

19.3

925

16.8

889

16.1

 Unknown

13,204

10.3

4433

33.6

1426

10.8

981

7.4

6364

48.2

Histology

 Ductal CA

90,076

70.5

45,067

50

16,643

18.5

11,578

12.9

16,788

18.6

 Lobular CA

19,247

15.1

9438

49

3858

20

2397

12.5

3554

18.5

 Other

18,431

14.4

7856

42.6

2,864

15.5

1818

9.9

5893

32

Grade

 Well differentiated

18336

14.4

9023

49.2

3884

21.2

2265

12.4

3164

17.3

 Moderately differentiated

41944

32.8

20,474

48.8

8250

19.7

5390

12.9

7830

18.7

 Poorly differentiated

34,994

27.4

18,034

51.5

6205

17.7

4701

13.4

6054

17.3

 Undifferentiated

1986

1.6

1072

54

341

17.2

275

13.8

298

15

 Unknown

30,494

23.9

13,758

45.1

4685

15.4

3,162

10.4

8889

29.1

ER Status

 Positive

26,905

21.1

13,544

50.3

5620

20.9

3218

12

4523

16.8

 Negative

7334

5.7

3854

52.5

1306

17.8

964

13.1

1210

16.5

 Unknown

93,515

73.2

44,963

48.1

16,439

17.6

11,611

12.4

20,502

21.9

PR Status

 Positive

22,008

17.2

11,052

50.2

4,629

21

2679

12.2

3648

16.6

 Negative

11,647

9.1

6034

51.8

2204

18.9

1436

12.3

1973

16.9

 Unknown

94,099

73.7

45,275

48.1

16,532

17.6

11,678

12.4

20,614

21.9

Chemotherapy

 No

85,453

66.9

41,840

49

15,960

18.7

9680

11.3

17,973

21

 Yes

34,993

27.4

17,859

51

6679

19.1

5408

15.5

5047

14.4

 Unknown

7308

5.7

2662

36.4

726

9.9

705

9.6

3215

44

Radiation therapy

 No

70,771

55.4

35,052

49.5

11,169

15.8

8441

11.9

16,109

22.8

 Yes

51,948

40.7

25,677

49.4

11,675

22.5

6949

13.4

7647

14.7

 Unknown

5035

3.9

1632

32.4

521

10.3

403

8

2479

49.2

Surgery

 No

4,245

3.3

1699

40

540

12.7

449

10.6

1557

36.7

 Yes

121,722

95.3

60,601

49.8

22,805

18.7

15,323

12.6

22,993

18.9

 Unknown

1787

1.4

61

3.4

20

1.1

21

1.2

1685

94.3

Hormonal treatment

 No

96,339

75.4

47,513

49.3

16,417

17

12,060

12.5

20,349

21.1

 Yes

24,282

19

12,355

50.9

6100

25.1

3045

12.5

2782

11.5

 Unknown

7133

5.6

2493

35

848

11.9

688

9.6

3104

43.5

The self-reported smoking status distribution by race showed that a larger proportion of Blacks (59.1 %) compared to Whites (47.7 %) were never smokers. A greater percent of Hispanic patients (63.7 %) were never smokers than non-Hispanics (47.2 %). A majority of patients in the lower SES were never smokers (53.4 %) compared with the percent of never smokers in the highest SES category (48.9 %).

Table 2 shows clinical characteristics of cancer patients in Florida. Many of the patients who were alcohol abusers were also current smokers (39.0 %), but only 11.0 % of obese individuals were current smokers. Patients with greater than 4 comorbid conditions were more likely to be current smokers. In addition, those with distant disease at diagnosis were more likely to be current smokers (16.8 %) with a lower percentage of those with localized disease being current smokers (12.3 %).

In the Cox proportional hazard models, significant associations were found for smoking status and worse survival (Table 3). Progressive adjustment from Model B (adjusted for age, race, ethnicity, SES, and marital status) to Model C (additionally adjusted for SEER, histology, differentiation grade, ER, PR, chemotherapy, radiation therapy, surgery, and hormonal treatment) changed the hazard ratio (HR) for current smokers (HR 1.44; 95 % CI = 1.38-1.50 and HR 1.43; 1.36–1.50, respectively) and for former smokers (HR 1.07; 1.04–1.11 and 1.10; 1.07–1.14, respectively). In the fully adjusted model, Model D (after additionally adjusting for co-morbidities), and compared to those who never smoked, current smokers and former smokers maintained worse survival (HR 1.33; 1.28–1.38 and HR 1.09; 1.06–1.13, respectively).
Table 3

Cox proportional hazard regression models for breast cancer overall survival by smoking status and smoking intensity (packs/day)

 

Model A

Model B

Model C

Model D

 

HR (95 %CI)

p value

HR (95 %CI)

p value

HR (95 %CI)

p value

HR (95 %CI)

p value

Smoking Status

 Never

1

 

1

 

1

 

1

 

 Current

1.15 (1.11, 1.19)

<0.001

1.44 (1.38, 1.50)

<0.001

1.43 (1.36, 1.50)

<0.001

1.33 (1.28, 1.38)

<0.001

 Former

1.04 (1.01, 1.08)

0.009

1.07 (1.04, 1.11)

<0.001

1.10 (1.07, 1.14)

<0.001

1.09 (1.06, 1.13)

<0.001

 Unknown

1.37 (0.96, 1.96)

0.082

1.37 (0.97, 1.93)

0.075

1.08 (1.02, 1.14)

0.008

1.05 (0.98, 1.12)

0.190

Smoking amount (pack/day)

 Never

1

 

1

 

1

 

1

 

  <1

1.03 (0.97, 1.09)

0.330

1.25 (1.18, 1.33)

<0.001

1.34 (1.27, 1.42)

<0.001

1.28 (1.20, 1.36)

<0.001

  1–2

1.18 (1.13, 1.23)

<0.001

1.51 (1.44, 1.59)

<0.001

1.52 (1.44, 1.60)

<0.001

1.40 (1.33, 1.47)

<0.001

  >2

1.69 (1.47, 1.95)

<0.001

2.11 (1.85, 2.41)

<0.001

1.89 (1.62, 2.21)

<0.001

1.70 (1.45, 1.99)

<0.001

 p linear trend

1.10 (1.07, 1.12)

<0.001

1.24 (1.21, 1.27)

<0.001

1.24 (1.21, 1.27)

<0.001

1.19 (1.16, 1.22)

<0.001

Model A Univariate: smoking only

Model B Multivariate: smoking + age, race/ethnicity/SES, marital status

Model C Multivariate: Model B + SEER, histology, differentiation grade, ER, PR, Chemotherapy, Radiation, Therapy, Surgery, Hormonal Treatment

Model D Multivariate: Model C + co-morbidities

In the fully adjusted models, we also observed that obese women had a decreased risk of mortality compared to non-obese, with an HR of 0.83 (0.80–0.87). Female alcohol abusers had an increased death risk with an HR of 1.09 (1.01–1.18) compared to the non-alcohol abusers/unknown ones (data not shown).

In an analysis restricted to current and never smokers, we determined the effect of smoking intensity (Table 3). In the fully adjusted model, compared to never smokers, those who smoked 1, 1–2, and >2 packs/day had worse survival (HR 1.28; 1.20–1.36, HR 1.40; 1.33–1.47 and HR 1.70; 1.45-1.99) with a significant linear dose–response (p for linear trend <0.001).

In the fully adjusted model D (data not shown), Black females had worse survival (HR 1.32; 1.25–1.39) compared to Whites, and Hispanic females had better survival than non-Hispanics (HR 0.92; 0.87–0.98). With lowest SES category as the reference, individuals from the middle-low, middle-high, and highest SES had a progressively better survival (HR 0.94; 0.90–0.98, HR 0.88; 0.84–0.92 and HR 0.81; 0.76–0.85, respectively) (data not shown).

We found significant interactions between smoking status and ethnicity and SES on survival. In the stratified analysis (Table 4), we found that Whites who were current or former smokers had worse survival compared with never smokers (HR 1.38; 1.33–1.44 and HR 1.11; 1.07–1.15, respectively). For Blacks, there was no difference in survival between never smokers and current smokers, with an HR of 1.10 (0.98, 1.23), or between former smokers and never smokers, with an HR of 1.02 (0.90–1.14).
Table 4

Cox proportional hazard regression models for overall survival by race/ethnicity/SESa

Modelsb

 

Smoking status

  

Never

Former

Current

Unknown

Reference

HR (95 %CI)

p value

HR (95 %CI)

p value

HR (95 %CI)

p value

Race

 A

White

1.00

1.11 (1.07, 1.15)

<.001

1.38 (1.33, 1.44)

<.001

1.06 (1.00, 1.14)

0.064

 B

African American

1.00

1.02 (0.90, 1.14)

0.799

1.10 (0.98, 1.23)

0.099

0.97 (0.86, 1.10)

0.677

 C

Other

1.00

1.28 (0.86, 1.89)

0.226

1.43 (0.73, 2.80)

0.300

0.91 (0.60, 1.38)

0.647

Ethnicity

 D

Non-Hispanic

1.00

1.10 (1.06, 1.15)

<.001

1.34 (1.29, 1.40)

<.001

1.05 (0.98, 1.12)

0.189

 E

Hispanic

1.00

1.01 (0.85, 1.19)

0.941

1.13 (1.01, 1.26)

0.033

1.07 (0.92, 1.23)

0.388

Socioeconomic Statusc

 F

Lowest

1.00

1.07 (0.97, 1.19)

0.170

1.15 (1.05, 1.26)

0.003

1.11 (1.00, 1.23)

0.042

 G

Middle-low

1.00

1.11 (1.04, 1.17)

<.001

1.32 (1.23, 1.41)

<.001

1.05 (0.96, 1.14)

0.290

 H

Middle-high

1.00

1.12 (1.06, 1.19)

<.001

1.41 (1.32, 1.50)

<.001

1.03 (0.95, 1.12)

0.487

 I

Highest

1.00

1.04 (0.96, 1.13)

0.311

1.36 (1.24, 1.49)

<.001

1.05 (0.96, 1.15)

0.293

aPlease note that each row depicts a separate model

bAll models are adjusted by age at diagnosis, marital status, co-morbidities, SEER, histology, differentiation grade, ER status, PR status, chemotherapy, radiation therapy, surgical treatment, and hormonal treatment

Models A, B, and C are additionally adjusted by Ethnicity and SES

Models D and E are additionally adjusted by Race and SES

Models F, G, H, and I are additionally adjusted by Race and Ethnicity

cSocioeconomic Status was categorized from the U.S. Census tract-level information on percent of households in the neighborhood living below the poverty index. Each tract was grouped by lowest (≥20 %), middle-low (≥10 and< 20 %), middle-high (≥5 and <10 %), or highest (% <5 %)

Looking at associations with ethnicity, current and former non-Hispanic smokers had worse survival compared with non-Hispanic never smokers (HR 1.34; 1.29–1.40 and HR 1.10; 1.06–1.15, respectively). Likewise, differences in survival were found between Hispanics who were current smokers compared to never smokers with an HR of 1.13 (1.01–1.26), but not between former and never smokers (HR 1.01; 0.85–1.19).

Compared to never smokers, worse survival was observed for current smokers in the highest (HR 1.36; 1.24–1.49), middle-high (HR 1.41; 1.32–1.50), middle-low (HR 1.32; 1.23–1.41), and lowest SES (HR 1.15; 1.05–1.26). In former smokers compared to never smokers, worse survival was only found in middle-high and middle-low SES categories (HR 1.12; 1.06–1.19 and HR 1.11; 1.04–1.17, respectively), but not in the highest or lowest SES categories.

Discussion

Our study revealed some interesting findings on the associations of survival, smoking status, smoking intensity, race/ethnicity, and SES status. Our multivariate models revealed that after adjusting for demographic, clinico-pathological characteristics, treatments, and extensive co-morbidities, overall breast cancer patients who were current or former smokers had worse survival than never smokers. Other studies have found significant associations between smoking and mortality in female breast cancer patients [4, 5, 6, 7, 8, 9]. However, these studies did not include extensive co-morbidities in the models. This is important as smokers tend to have more co-morbidities which have a significant impact on female breast cancer survival [3, 16, 17]. Our results showed that the association of worse survival remained significant even after adjusting for these comorbid conditions. The association also remained significant after adjusting for tumor stage as disease spread has been shown to be more extensive in smokers [18].

We found an inverse association of survival with smoking intensity, that is, the dose–response to larger quantities of smoking in breast cancer patients remained significant in the fully adjusted model. At least 2 pack/day smokers with breast cancer had a 70 % worse survival after controlling for co-morbidities and all other demographic and clinical characteristics. Other authors have also found significant associations between the amount of smoking and breast cancer mortality, but they did not adjust for numerous co-morbidities, so our study gives a unique perspective to this association [4, 5].

Stratifying by race, our fully adjusted regression models showed that White current and former smoking breast cancer patients had worse survival than never-smoking White patients. However, among Black patients, there was no survival advantage for those who never smoked compared with current and former smokers. Like White patients, current smoking non-Hispanic and Hispanic breast cancer patients also displayed worse survival compared with their never-smoking respective category. Although racial and ethnic disparities in female breast cancer survival have been previously described [10, 11, 12, 13], only one study has assessed the association between smoking and survival in female cancer patients for the different races separately [8], and none has assessed this association among different ethnicities. Yu et al. did not find significant associations between smoking and mortality for White patients [8]; this discrepancy with our results may be due to our adjustment for numerous potential confounders.

Our results suggest a different susceptibility for breast cancer survival in relation to smoking for the different races. A genetic perspective may help provide a better understanding of the underlying relationship between smoking and breast cancer survival and explain our findings of racial differences; this line of research was recommended by the surgeon general [3], as it is possible that the associations between risk for breast cancer and active smoking could differ according to different phenotypes. A possible modification of the risk for breast cancer incidence in relation to smoking for the NAT2 genotype has been already suggested [3]. The surgeon general report warrants that future research should explore the risk of smoking on breast cancer in genetically defined subgroups [3], and our results could be consistent with this line of research in terms of breast cancer survival.

As far as we know, this was the first study to analyze the association between smoking and survival in female breast cancer patients by neighborhood SES. For our stratified SES models, current smokers in each category of SES had worse survival than their never-smoking SES counterparts, controlling for all other variables and co-morbidities. However, former smokers in the lowest and highest SES categories were equally likely to survive, while those in the middle-low and middle-high SES had worse survival compared to never smokers in their same respective SES category. As we had no information on when former smokers quit or for how long they had smoked, it is difficult to interpret these findings.

Limitations

We had several limitations in this study. We did not have access to HER information which limited our ability to assess a possible relationship between smoking and both HER-2 receptor and triple-negative breast cancers on survival. In addition, the number of women with unknown ER and PR status was very high; this is because little information on either status was collected by FCDS prior to 2004. We assessed smoking status at diagnosis, so we cannot be certain whether quitting after diagnosis would benefit the patient’s prognosis. Furthermore, we did not have the time span of those who quit smoking. Smoking status was self-reported, so we expect to have some response bias [19]. On the other hand, current smokers appear to have a worse prognosis than former smokers, so the effect of smoking on survival appears to decrease after quitting smoking regardless of how long prior to diagnosis the patient quit.

Strengths

The greatest strength of this study was our ability to enhance the FCDS data with the Florida ACHA data, and link it to the U.S. Census to form a very large dataset of 127,754 patients. We were also able to adjust for an extensive number of confounding variables including 31 co-morbidities and stage of disease at presentation, allowing us to identify the independent effect of smoking on breast cancer patients’ survival as well as a dose–response.

Conclusions

In conclusion, current smoking is strongly associated with an increased risk of mortality in most populations of female breast cancer patients with a linear dose–response. Patients should be advised to quit smoking as part of the tertiary prevention protocols for all female breast cancer patients. Based on results from our data, those who will mostly benefit from these recommendations are White patients.

Acknowledgments

This study was funded by the James & Esther King Florida Biomedical Research Program (Grant 10 KG-06).

Compliance with ethical standards

This manuscript complies with the current laws of the country in which the research was performed. The study was approved by the Institutional Review Board of the University of Miami and the Florida Department of Health, and has been performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. For this type of study, formal consent is not required.

Conflict of interest

The authors declare that they have no conflict of interest.

Copyright information

© Springer Science+Business Media New York 2015

Authors and Affiliations

  • Alicia Padron-Monedero
    • 1
    • 2
  • Stacey L. Tannenbaum
    • 3
  • Tulay Koru-Sengul
    • 1
    • 3
  • Feng Miao
    • 3
  • Damien Hansra
    • 1
  • David J. Lee
    • 1
  • Margaret M. Byrne
    • 1
    • 3
  1. 1.Department of Public Health SciencesUniversity of Miami Miller School of MedicineMiamiUnited States
  2. 2.Department of EpidemiologyComunidad de MadridMadridSpain
  3. 3.Sylvester Comprehensive Cancer CenterUniversity of Miami Miller School of MedicineMiamiUnited States

Personalised recommendations