European Journal of Epidemiology

, Volume 27, Issue 11, pp 857–866

Survival from breast cancer in relation to access to tertiary healthcare, body mass index, tumor characteristics and treatment: a Hellenic Cooperative Oncology Group (HeCOG) study

Authors

  • Paraskevi Panagopoulou
    • Department of Hygiene, Epidemiology and Medical StatisticsAthens University Medical School
  • Helen Gogas
    • 1st Department of Medicine, “Laiko” General HospitalUniversity of Athens
  • Nick Dessypris
    • Department of Hygiene, Epidemiology and Medical StatisticsAthens University Medical School
  • Nikos Maniadakis
    • Department of Health Services Organisation and ManagementNational School of Public Health
  • George Fountzilas
    • Department of Medical Oncology, “Papageorgiou” General HospitalAristotle University of Thessaloniki School of Medicine
    • Department of Hygiene, Epidemiology and Medical StatisticsAthens University Medical School
CANCER

DOI: 10.1007/s10654-012-9737-z

Cite this article as:
Panagopoulou, P., Gogas, H., Dessypris, N. et al. Eur J Epidemiol (2012) 27: 857. doi:10.1007/s10654-012-9737-z
  • 339 Views

Abstract

Apart from tumour, treatment and patient characteristics at diagnosis, access to healthcare delivery may as well play a significant role in breast cancer prognosis. This study aimed to assess the additional impact exerted on survival by travel burden—a surrogate indicator of limited access to healthcare- expressed as geographical distance and/or time needed to reach the tertiary healthcare center from the patient’s residence. Between 1997 and 2005, 2,789 women participated in therapeutic clinical trials conducted by the Hellenic Cooperative Oncology Group. The effect of geographical distance and travel time between patient’s residence and treating hospital on survival was estimated using Cox proportional hazards regression adjusting for age, menopausal status, tumour size/grade, positive nodes (number), hormonal receptor status, HER2 overexpression, surgery type/treatment protocol as well as for body mass index >30 kg/m2. More aggressive tumour features, older treatment protocols and modifiable patient characteristics, such as obesity (HR: 1.27) adversely impacted on breast cancer survival. In addition, less studied indicators of access to healthcare, such as geographic distance >350 km and travel time >4 h were independently and significantly associated with worse outcomes (HR = 1.43 and 1.34 respectively). In conclusion, to address inequalities in breast cancer survival, improvements in access to healthcare services related to increased travel burden especially for patients of lower socioeconomic status should be considered, more than ever at times of financial crisis and independently of already known modifiable patient characteristics.

Keywords

Breast cancerSurvivalPrognosisHealthcare accessHealthcare inequalitiesObesitySocioeconomic status (SES)

Introduction

Survival from breast cancer has been shown to be strongly associated with a wide variety of factors including tumor characteristics such as tumor size, histology and grade, stage at diagnosis [1], hormonal receptor status and lymph node involvement as well as appropriateness of treatment [2]. Additionally, biological, molecular and genetic markers such as Human Epidermal Growth Factor Receptor 2 (HER2) overexpression, Ki67, p53 and bcl-2 mutations have gained interest as possible predictors of response to treatment [3, 4]. The role of lifestyle factors such as alcohol and tobacco use, physical activity and consumption of fruits and vegetables [57] has also been examined without definite results.

Obesity, a modifiable characteristic, is linked to increased risk for breast caner. The attributable to obesity, and hence preventable fraction of breast cancers in the UK in 2010 amounts to 8.7 % [8]. Several studies, including a recent metaanalysis [9], have also examined the effect of obesity on breast cancer prognosis [1012]. They support a rather positive association both for pre- and post-menopausal women [13] which seems to vary according to specific tumor and treatment characteristics. Adverse disease features, hormonal influences, and comorbidities [14] have been stipulated as underlying pathophysiologic mechanisms for this unfavourable effect.

Another poor prognostic indicator for breast cancer seems to be low socioeconomic status (SES) [15, 16] which has been associated with delayed diagnosis, choice of treatment and more premature death [17, 18]. A low SES is in general related with limited healthcare access [19], whereas a key element in conceptualizing geographic access to healthcare is travel burden [20]. In fact, it is possible that the interaction between increased travel burden and low SES, which is more common in remote areas, may further limit physical accessibility and use of healthcare services resulting in worst health outcomes [21]. Consequently, travel burden could emerge as a proxy variable for both geographic access to healthcare and low SES. Travel burden, expressed as geographic distance from or travelling time to a healthcare facility or both, has been associated with disparities in the use of cancer prevention services [22],with treatment decisions for various types of cancer -including, digestive tract [23], lung [24] and breast cancer [25, 26]- and with an overall detrimental effect on cancer prognosis [27]. Nevertheless, the studied populations, the methodologies used and the cut-off values varied widely therefore no definite results can be reached. Additionally, very few published studies have explored the differential impact of SES and access to the healthcare on survival.

The present study aimed to investigate the effect of limited access to healthcare and possibly low SES on the overall survival of patients with early stage, operable breast cancer among women who participated in treatment protocols across Greece, using gross indicators of travel burden namely geographical distance and travel time between place of residence and treating hospital, taking into account the effect of specific tumor, treatment and patient related variables, with an emphasis on obesity.

Materials and methods

Study population

Since 1997, the Hellenic Cooperative Oncology Group (HeCOG) has recruited a total of 2,823 patients with histologically confirmed, node positive and high risk node negative epithelial breast cancer, participating in successive study protocols (HE10/97, HE10/00, HE10/04, HE10/05) in six major Greek cities (Athens, Thessaloniki, Larissa, Ioannina, Patras, Chania). Details on these phase III clinical trials have been published elsewhere [2832] (“Appendix”).

Data collection

Between May 1st, 1997 and November 7, 2008, 2,823 women with breast cancer were enrolled in the four protocols and were actively followed till May 31st, 2011 for their vital status and cause of death. For 34 patients, no information was available on either survival (N = 24) or time of death (N = 10) and they were excluded from the study. Thus, a total of 2,789 patients [age range: 21–79 years, mean age (±SD): 52 (±12) years] were included in the analyses. Twenty-six women, who died from a cause not related to breast cancer, were considered as alive till the date of death for the purposes of the present study.

All patients had undergone partial or modified radical mastectomy followed by selected chemotherapy regimens and/or hormone therapy and radiotherapy according to current treatment guidelines. A standardised questionnaire was used to record important patient characteristics, namely, age, menopausal status and somatometrics (height and weight) at diagnosis, tumour characteristics, notably, size, grade, number of positive nodes, hormonal (oestrogen and progesterone) receptor status, HER2 overexpression, type of surgery and treatment protocol.

To calculate the corresponding travel distance between the patient’s place of residence and the treating hospital, electronic maps and related software were utilized (http://maps.google.com). For 98.7 % of cases either the complete address was available or the city of residence was known and so the mean travel distance from the city centre to the destination hospital was obtained from the aforementioned software as an approximation. For the remaining 1.3 % of cases for whom no data were available, travel distance was estimated as the weighted mean of available distances for each destination hospital, separately (n = 37) [33, 34]. For the latter scenario, the number of patients who lived in each city as the proportion of the total number of patients who were admitted to each hospital destination was used as a weight.

Statistical analyses

Associations of BMI in three categories (<25, 25.0–29.9, 30+ kg/m2) by available patient and disease characteristics were examined using the Chi-square test. Univariate distributions of outcome (dead or alive) by patient characteristics (age, BMI and menopausal status) and disease features (tumor grade and size, number of positive nodes, hormonal receptor and HER2 status), type of treatment protocol and type of operation, as well as by two gross indicators of access to healthcare (geographical distance and time needed to travel to the treating hospital) were explored using log-rank tests; the Kaplan–Meier curves were also derived by BMI categories and the access to healthcare variables. Lastly, the Cox proportional hazards regression model was used to assess their adjusted effect on survival. Given that the “geographical distance” and “time needed to travel” variables are intercorrelated, they were alternatively introduced in the model of core variables. We opted to examine both covariates, despite the high expected co-linearity in order to account for discrepancies as in the case of geographically proximal areas in which the mountainous geography of the country leads to different travelling times or in the case of geographically proximal islands that avail frequent airline connection in contrast to those that are only connected via maritime link. The impact of the hormonal receptor and HER2 overexpression status (in three categories: triple negative; HER2-positive; HER2-negative) was also explored alternatively to the “hormonal receptor status” variable in a subset of cases for whom information was available. SAS was used for the analyses (SAS v9, SAS Institute Carry, NC, USA).

Ethical considerations

All clinical protocols were approved by local regulatory authorities and fulfilled the criteria of the Helsinki Declaration. Each patient provided a study-specific written informed consent at the time of initial recruitment before registration in the study. The clinical trials HE10/97, HE10/00 and HE10/05 were included in the Australian New Zealand Clinical Trials Registry (ANZCTR) and the allocated Registration Numbers were respectively: ACTRN12611000506998, ACTRN12609001036202 and ACTRN12610000151033, respectively.

Results

Table 1, shows the distribution of the 2,789 breast cancer patients in three BMI categories by patient, disease and treatment characteristics. As expected, obesity was positively associated with most tumour characteristics (tumor size, number of positive nodes, other than triple negative status), albeit the association with tumour grade did not reach statistical significance in this dataset. Additionally, obese patients were older and post-menopausal when compared to leaner patients.
Table 1

Distribution of the 2,789 breast cancer cases by patient, disease and treatment characteristics as well as by body mass index (BMI) category

Variable

BMI (kg/m2)

P value

<25.0

25.0–29.9

30.0+

N

%

N

%

N

%

Grade

 1 and 2

489

53

545

51.5

404

50.1

0.23

 3, undifferentiated

434

47

514

48.5

403

49.9

 

Size (cm)

 ≤2.0

359

38.9

364

34.4

254

31.5

0

 2.1–5.0

474

51.4

577

54.5

464

57.5

 

 >5.0

90

9.7

118

11.1

89

11

 

Positive nodes (number)

 0–3

450

48.8

466

44

323

40

0

 4+

473

51.2

593

56

484

60

 

Hormonal receptor status

 Negative

216

23.4

258

24.4

194

24

0.74

 Positive

707

76.6

801

75.6

613

76

 

Triple negativea

 Yes

102

13.1

109

12.8

95

13.5

0

 HER2 positive

278

35.6

317

37.2

192

27.3

 

 HER2 negative

401

51.3

426

50

416

59.2

 

Type of operation

 Modified radical mastectomy

558

60.5

687

64.9

471

58.4

0.44

 Partial mastectomy

365

39.5

372

35.1

336

41.6

 

Age (years)

 <40

251

27.2

96

9.1

44

5.4

0

 41–49

349

37.8

288

27.2

142

17.6

 

 50–59

180

19.5

333

31.4

257

31.9

 

 60+

143

15.5

342

32.3

364

45.1

 

Menopausal status

 Pre/perimenopausal

644

69.8

461

43.5

211

26.2

0

 Postmenopausal

279

30.2

598

56.5

596

73.8

 

Total number of cases

923

33.1

1,059

38

807

28.9

 

a453 cases missing

BMI body mass index, HER2 Human Epidermal Growth Factor Receptor-2

The average follow-up was 5.72 (±2.67) years with a median period of 5.08 years, ranging from 0 to 13.3 years). During the follow-up, 507 women (18.2 %) died on account of breast cancer and the mean time to death was 4.43 (±2.35) years. In Table 2, the distribution of the survival status by sociodemographic/lifestyle, disease/treatment characteristics and by the two indicators of access to healthcare, namely “geographical distance” and “time needed to travel to the treating hospital”, is shown. Longer survival was positively and significantly associated with less aggressive grade and smaller tumours, fewer positive nodes, positive hormonal receptors and lack of triple negative status, as well as with partial mastectomy and more recent therapeutic protocols. Survival was marginally associated with menopausal status (P = 0.06) but not with age at diagnosis (P = 0.44) in this dataset. Regarding the main variables of interest, obesity exerted a statistically significant adverse impact on survival and so did an over 300 km long geographical distance between the patients’ place of residence and the treating hospital and a longer than four hours travel time. When geographical distance was examined as an ordinal variable a significant effect on survival was also noted for distances over 350 km. These results are also shown in three Kaplan–Meier survival curves of breast cancer patients by BMI, travel distance and travel time (P values derived from log-rank tests = 0.003, 0.03 and 0.01, respectively) (Fig. 1).
Table 2

Distribution of 2,789 breast cancer cases by patient, disease and treatment characteristics, and by survival status

Variable

Alive

Dead

P value

N

%

N

%

Grade

 1 and 2

1,225

85.2

213

14.8

0

 3 and undifferentiated

1,057

78.2

294

21.8

 

Size, cm

 ≤2.0

866

88.6

111

11.4

0

 2.1–5.0

1,203

79.4

312

20.6

 

 > 5.0

213

71.7

84

28.3

 

Positive nodes

 0–3

1,127

91

112

9

0

 4+

1,155

74.5

395

25.5

 

Hormonal receptor status

 Negative

522

78.1

146

21.9

0.01

 Positive

1,760

83

361

17

 

Triple negativea

 Yes

243

79.4

63

20.6

0.01

 HER2-positive

646

82.1

141

17.9

 

 HER2-negative

1,064

85.6

179

14.4

 

Age

 <40 years

318

81.3

73

18.7

0.44

 41–49

653

83.8

126

16.2

 

 50–59

621

80.7

149

19.4

 

 60+

690

81.3

159

18.7

 

Menopausal status

 Pre/perimenopausal

1,096

83.3

220

16.7

0.06

 postmenopausal

1,186

80.5

287

19.5

 

BMI, kg/m2

 <25.0

776

84.1

147

15.9

0.01

 25.0–29.9

863

81.5

196

18.5

 

 30.0+

643

79.7

164

20.3

 

Study protocol

 HE/97

401

67.2

196

32.8

0

 HE/00

875

78.3

242

21.7

 

 HE/04

86

88.7

11

11.3

 

 HE/05

920

94.1

58

5.9

 

Type of mastectomy

 Modified radical

1,339

78

377

22

0

 Partial

943

87.9

130

12.1

 

Geographical distance—dichotomous (residence to hospital, km)

<300

2,149

83.2

462

17.7

0.01

 300+

133

74.7

45

25.3

 

Geographical distance—ordinal (residence to hospital, km)

 <50

1,475

82.1

322

17.9

0.04

 50–149

449

82.4

96

17.6

 

 150–249

177

83.9

34

16.1

 

 250–349

97

82.2

21

17.8

 

 350+

84

71.2

34

28.8

 

Time needed to travel (residence to hospital, hours)

 <4

2,151

82.4

460

17.6

0

 4+

131

73.6

47

26.4

 

Total number of cases

2,282

81.8

507

18.2

 

aMissing 453 cases

BMI body mass index, HER2 Human Epidermal growth factor Receptor-2

https://static-content.springer.com/image/art%3A10.1007%2Fs10654-012-9737-z/MediaObjects/10654_2012_9737_Fig1_HTML.gif
Fig. 1

Kaplan-Meier survival curves for breast cancer cases by a BMI categories b Travel distance c Travel time

Table 3 presents unconfounded Hazard Ratios (HR) and 95 % Confidence Intervals (95 % CIs) for survival derived from proportional hazards models, confirming those from the univariate analysis. Indeed, the probability of death was substantially higher and statistically significant for women with grade 3 or undifferentiated tumours (33 %), with larger size tumours, with more than four positive nodes (HR: 2.47) and with negative hormonal receptors (HR: 1.25) or triple negative status (HR: 1.38). By contrast, those who sustained partial mastectomy and were enrolled in more recent treatment protocols had a significantly better survival. A substantial 27 % higher probability of death was noticed among women with BMI at diagnosis over 30 kg/m2 compared to women with normal BMI. Finally, the detrimental effect of an increasing geographical distance between residence and treating hospital was noted. Specifically, when examined as dichotomous variables distances longer than 300 km and a travelling time longer than 4 h were associated with a probability of death reaching 37 and 34 %, respectively. Additionally, when geographical distance was also examined as an ordinal variable, distances longer than 350 km were associated with a 43 % higher probability of death.
Table 3

Hazard Ratios (HR) and 95 % confidence intervals (95 % CIs) for survival derived from proportional hazards models by a number of study variables

Variable

Category or increment

HR

95 % CI

P value

Core model

Grade

1 and 2

Baseline

1.11

1.6

0

3 and undifferentiated

1.33

   

Size, cm

≤2.0

Baseline

1.26

1.95

0

2.1–5.0

1.57

1.39

2.48

0

>5.0

1.86

   

Positive nodes

0–3

Baseline

1.99

3.07

0

4+

2.47

   

Hormonal receptors

Positive

Baseline

1.03

1.53

0.03

Status

Negative

1.25

   

Age

10 years more

0.99

0.86

1.15

0.93

Menopausal status

Pre/perimenopausal

Baseline

0.81

1.49

0.54

Postmenopausal

1.1

   

BMI, kg/m2

<25

Baseline

0.87

1.35

0.48

25–29.9

1.08

1

1.62

0.04

30+

1.27

   

Type of mastectomy

Partial

Baseline

1.06

1.6

0.01

Modified radical

1.3

   

Study

HE/97

1.42

1.03

1.95

0.03

HE/00

1.43

1.06

1.94

0.02

HE/04

1.43

0.75

2.74

0.28

HE/05

Baseline

   

Alternatively or additionally introduced variables in the core model

Model 1: “triple negative” status alternatively introduced to “hormonal receptors status” variable

 Triple negative (missing 453 cases)

HER2-positive

Baseline

0.8

1.26

0.98

HER2-negative

1

1.02

1.86

0.04

Yes

1.38

   

Model 2: additionally introduced variable

 Geographical distance—dichotomous (residence to hospital, km)

<300

Baseline

1

1.87

0.04

300+

1.37

   

 Geographical distance—ordinal (residence to hospital, km)

<50

Baseline

0.79

1.25

0.94

50–149

0.99

0.55

1.12

0.18

150–249

0.79

0.52

1.26

0.35

250–349

0.81

1

2.04

0.05

350+

1.43

   

Model 3: alternatively to the “geographical distance” introduced variable

 Time needed to travel (residence to hospital, hours)

<4

Baseline

1

1.82

0.05

4+

1.34

   

BMI body mass index, HER2 Human Epidermal Growth Factor Receptor-2

Discussion

In this large collaborative study from Greece, we investigated the effect of gross indicators of access to healthcare as well as that of obesity on the survival of women with early-stage breast cancer after controlling for known confounders such as tumor characteristics and mode of treatment. The study confirmed the adverse impact of increased BMI on breast cancer survival with a 27 % increase in mortality observed among obese women. Over and beyond the impact of obesity, however, limited access to healthcare, as expressed through the less studied gross indicators of increased travel burden, namely geographical distance and travel time, was independently associated with increased mortality and the respective effects were even higher (43 and 34 % respectively). Finally, travel burden emerged as an indirect indicator of access to healthcare in the context of low socioeconomic status.

The effect of geographical access to healthcare on health outcomes, expressed as distance from a tertiary hospital, has been examined in a few studies mostly in relation to preventive services [22] and treatment decisions [23, 24], whereas data regarding its relation to breast cancer prognosis are scarce. Travel burden expressed as driving distances or travelling times from patient’s residence to healthcare provider was found to be significantly associated with early diagnosis of breast cancer and with time to treatment for colorectal cancer in the US [22]. Distance and travel time were also associated with treatment decisions (e.g. radiotherapy after surgery) for breast [25] and lung cancer [24] in the US and the UK. In Northern England, an inverse association between travel time and treatment take-up was shown for services provided in specialised centres involving longer than average patient journeys [26]. Regarding breast cancer it was also shown that treatment decisions were influenced by travel burdens such as travel duration and expenses, and travelling by car or boat during winter months [35, 36]. Such treatment decisions are inevitably associated to outcome and prognosis. Regarding survival per se, distance to the nearest cancer centre was considered a potential survival predictor for digestive tract cancer in France [23], whereas a study from Scotland provided evidence that the increased cancer mortality in rural patients is compounded by a travel time longer than 3 h [27]. Finally, researchers in less developed countries have suggested that a decentralized health policy regarding cancer care aiming at reduced travel times is required in order to decrease inequalities [37]. Some of these studies have examined the effect of distance in relatively well-defined areas such as a city, province or state whereas others examine it at a countrywide level taking into account different potential disparities [38], hence the variation in the results. In accordance with these studies the present study also showed that a geographical distance between the patient’s residence and the treating hospital longer than 350 km or a travelling time over 4 h were related to significant increases in mortality reaching 43 and 34 %, respectively, independently from several well established predictors of survival, including obesity, which is itself a barrier to obtaining health care [39]. Consequently, geographical distance might represent a barrier to optimal cancer care for some patients and may contribute to the rural–urban disparity in breast cancer survival that has been previously reported [40]. In Greece all inhabitants enjoy a universal coverage health insurance system, therefore disparities in access to healthcare are mainly due to barriers limiting more its physical accessibility than its acceptability and affordability [41]. The Organization for Economic Co-operation and Development (OECD) has documented a profound disparity of the gross domestic product per capita between urban and rural areas in Greece among other countries [42]. This fact lends support to the hypothesis that increased travel burden for people of remote areas can be considered as a proxy variable for low SES and one of the causes of worst health outcomes. Indeed, recent studies have shown that socioeconomic factors are likely to influence survival from childhood leukemia at least in some socio-cultural contexts in Greece [43] and in the UK [44] whereas prior studies have reported an association between low SES and poor breast cancer prognosis [17, 45]. The present study not only confirms the previously described association but also offers a novel insight proposing the increased travel burden as a proxy indicator for low SES for patients from remote areas.

The survival inequality observed in the present study may be due to less timely detection of complications or recurrences among patients residing longer distances from the treating hospitals despite the fact that all participants completed their treatment and follow-up regardless of their home location.

In accordance with previous research, this study also confirmed in a Greek population that lower grade and smaller tumors, fewer positive nodes, positive hormonal receptors in contrast to triple negative tumors, partial mastectomy and more recent therapeutic protocols had better prognosis [1, 5]. The study additionally confirmed the association of obesity with several tumor and patient characteristics such as tumor size, number of positive nodes, menopausal status and age but not with histological grade, hormonal receptors and type of operation [10, 12, 46]. Moreover, the previously shown association between obesity and a worse breast cancer-specific survival both for pre- and post-menopausal women (HR = 1.33) was also evident in our study which showed that the probability of death for obese women was of the same order of magnitude (HR = 1.27) [10, 47, 48]. This detrimental effect has been attributed to the association of obesity with more aggressive and larger tumors leading to later-stage diagnosis [49]. In the present study however the finding remained significant even after adjustment for these variables. Poor survival of obese patients has also been attributed to several other factors including: the relative underdosing, common in patients with body surface area exceeding 2 m2 [50], the elevated estrogen levels produced by excess adipose tissue and/or the decreased levels of sex hormone binding globulin [51] as well as the potentially different response of obese women to anticancer treatment. Additionally, it has been reported that hormone receptors and HER2 overexpression, as well as their combinations (e.g. triple negative tumors) may affect prognosis differently for obese and lean women, but studies are still controversial [47, 5255]. It is also likely, that inferior survival might be due to comorbidities common in obese women, such as cardiovascular or renal conditions [1], whereas hyperinsulinemia, insulin resistance and other obesity-associated factors (e.g. IGF-1, adipocytokines) may be implicated in breast cancer recurrence [56, 57].

This is a relatively large study that includes a significant fraction of a rather culturally and ethnically homogeneous group of Greek women diagnosed with early-stage breast cancer during the past 15 years. The patients were enrolled in study protocols and were followed by the same physicians, reducing thus the possibility of bias due to differences in treatment or follow-up practices. Finally, data on a wide range of prognostic factors was available and analysed. However, since all data were derived from medical records information for modifiable lifestyle factors was available only for obesity but not for diet or physical activity. Most importantly, this is one of the few studies providing insight to the effect of travel burden on breast cancer survival. This issue is of particular significance for countries with geographic make-ups preventing easy travelling and access to specialized healthcare (like mountainous mainland and dispersion of the population to small islands) and becomes of prime importance at times of financial crises [58]. If data on individual socioeconomic proxies had been available, we would have been able to disentangle, to the extent possible, the direct effect of SES indicators from the impact exerted on survival by differential access to tertiary healthcare. Unfortunately, information on SES is missing in this dataset. Nevertheless, travel burden has emerged as a potential surrogate indicator of SES since it is for the unprivileged that an increased geographical distance and/or travel time may represent an additional barrier to healthcare. Finally, patients participating in clinical trials are considered to have a better overall survival than non-protocol eligible individuals. Thus, it cannot be inferred with certainty how generalisable the findings would be, if patients of all stages upon diagnosis were included in the analyses and not only those eligible for the respective protocols.

In conclusion, over and beyond the already known impact of inherent or modifiable characteristics like obesity on breast cancer survival, limited access to healthcare as expressed through the less studied gross indicators of geographical distance and travel time, may play an even more important negative role, especially in the setting of a low socioeconomic status and at times of financial crises.

Ethical standards

The present study complies with the current laws of the Greece.

Conflict of interest

All authors declare that they have no conflict of interest.

Supplementary material

10654_2012_9737_MOESM1_ESM.docx (15 kb)
Supplementary material 1 (DOCX 14 kb)

Copyright information

© Springer Science+Business Media Dordrecht 2012