Breast cancer guidelines—Physicians' intentions and behaviors

Abstract

Guidelines are written to define what a physician should do, and networks set up to provide every patient with good practice. However, is willingness to treat according norms enough to actually implement it? Between 1997 and 2003, 4,533 women with invasive, noninflammatory, nonmetastatic breast cancer have been treated within the framework of a regional network (R2C). The rate of implementation of 5 consensual norms was assessed. The rate of “abnormal” management regarding surgical re-excision for inadequate margin was found to be 12.6%. The main explanatory variable was patient age >70 years (OR = 4.05). For nodal exploration, the sampling quality threshold was set at 10. Mean rate of lack of compliance was 25.2%. The 2 main explicative factors were surgeon's experience and women's age. The observed rate of “insufficient” irradiation dose was 18.2%. The main explanatory variables were age (with a gradient) and a negative nodal status. Concerning adjuvant chemotherapy, the rate of no treatment (despite consensual indication) was 16.0%. Again, the main explicative factor was age (with a gradient). Women's age appears to be a major explanatory variable predicting lack of physician's compliance with consensual norms. Besides the age of the women, a “better” prognosis (negative nodal status and pT ≤ 20 mn) is often associated with lack of compliance. It is not clear, however, if it's the rules that do not fit the clinical situation of aging patients or the physicians who are not aware of the benefit of consensual disease management for aging patients. © 2006 Wiley-Liss, Inc.

Access to high standard of care will be achieved by a 2-step process: define “the” best treatment, and make it available for all. However, this is a theoretical and idealized view and it is of major importance to check in the real world which population is really offered the actual treatment that complies with good medical practice.

Regardless of the debate on how to define the “recommended” treatment from a theoretical perspective,1, 2 local implementation and local assessment is a critical but often dismissed step. This has been described for long3, 4 but is iteratively rediscovered.5, 6 Therefore, the diffusion of good medical practice depends on the quality of the norm7 and on local “tailoring”.8

Access to medical care is an issue of horizontal equity (all and every person with the same condition should be offered the same treatment). Many factors impede such a valuable goal from being attained, among which cost, information, empowerment and actual availability or access.9

In France it may be assumed that cost is not (yet?) an issue. Indeed, different health insurance organizations cover almost all persons legally living in France. For a serious condition (such as cancer), the treatment is free of charge. However, practical difficulties and convenience might limit access to health services.10 Two innovations might remove these brakes: a technological one, “telemedicine,” and an organizational one “the networks”.11 In the latter, a group of physicians covering a geographical area share technical resources, consensual assessment of disease and patients, consensual treatment protocol, database, quality control and risk management processes.

In the Region (a French equivalent of the Länders in Germany, the Provinces in Canada, and the States in US) “Provence-Alpes-Côte d'Azur,” such a network has been set up in 1997. It started with the commitment of 227 physicians (409 in 2003), all involved in cancer treatment, with as a first step the definition of shared objectives, internal rules and funding process. The second step was to define “local” good medical practices for specific conditions (among which, first line management of breast cancer) based on different published guidelines (ASCO statement, Cochrane database, French SOR…). This negotiated process is seen as critical as it should facilitate appropriation and commitment of actors, medical management being built on a limited set of validated options and not imposed from the outside “Freedom is obedience to a self-imposed law of reason”.12

With such a favorable background (self-decided and approved rules for treatment, voluntary adhesion, same geographical area and cultural background), it was expected that compliance with our standard would be high and that we would observe no or few heterogeneity in patient's management.

We were also very concerned with the population of our aging patients for whom we assumed less knowledge and skill at gaining information, and less empowerment. Norms are all the more valuable tools for this group of patients.

An audit of physician's compliance with their own norms defined in an organized network was therefore set up to answer 2 questions with regard to breast cancer treatment (i) is the rate of compliance with our defined standard of care high enough? (ii) are older patients more at risk of being treated outside the recommendation?

Material and methods

Network

The regional network “R2C” was set up in 1997. The aim was to offer patients access to local management of cancer according to a treatment standardized with regard to gravity. Physicians who want to participate in this network have to agree with the rules of the network, to apply the recommended treatment (if available and relevant) and to share medical files (while respecting French legal rules about medical information and confidentiality). The French Commission on Data Processing and Liberties approved the constitution of the shared database, which was used for this report (No = 535,590 October 14, 1997).

Patients

According to the French Register,13 15,380 breast cancers had occurred between 1997 and 2003 in the geographical area covered by the network, while 6,533 (42.5%) had been treated (at least partially by surgery, radiotherapy, chemotherapy or hormone therapy) in the R2C network. Thus, almost 40% of breast cancers in this area had theoretically been managed according to the R2C guidelines.

Every patient treated during this period for a nonmetastatic, noninflammatory, invasive breast cancer has been included in the survey. The date of the diagnostic process (surgical or transcutaneous biopsy) was the entry criterion. Local recurrent (less than 7 years) and prevalent nodal metastatic cancers were removed from analysis.

Guidelines

Our guidelines were mainly based on the document released by the French National Network of Regional Cancer Clinics about nonmetastatic breast cancer management. The methodology used was the so-called “Standard Options and Recommendations.”14 Norms had been set up according to “evidence based medicine”. The R2C guidelines were approved in 1997 and updated in 2000 (with changes about chemotherapy and hormone therapy). For prognosis staging 3 levels were defined: Stage 1 (not deserving adjuvant chemotherapy), Stage 2 and Stage 3.

Endpoints

Five criteria have been assessed with regard to compliance with the norms:

• Number of lymph nodes sampled in standard surgical assessment. Sentinel node processes were not analyzed for this criterion (10 or more versus less).

• Surgical re-excision for inadequate margins (Yes versus No).

• No adjuvant chemotherapy carried out despite consensual indication (Yes versus No).

• No adjuvant hormone therapy carried out despite consensual indication (Yes versus No).

• Where relevant, the level of irradiation used (60 grays or more versus less).

Statistical analysis

The explained variables are the endpoints described earlier. The associated factors (explanatory variables) tested belong to the following 4 main categories:

• Tumor characteristics: Date of diagnosis (year), pTNM, SBR grade, hormonal receptors status, free margin (5 mm).

• Patient's characteristics (only women were assessed): Age, menopause status, personal history of previous cancer.

• Physician's characteristics: Age, gender, location of practice (population covered), level of oncology practice, work in a clinic with a chemotherapy unit, involvement in the R2C network practice.

• Treatment characteristics: Chemotherapy (Y/N), radiotherapy (Y/N and dose), hormone therapy (Y/N).

Statistical methodology

The first step was to assess compliance rate (number of patients treated or assessed according to the norm/number of patients treated or assessed). If this rate was above or equal to 90%, no further analysis was carried out. When the rate was below this threshold, a further stepwise logistic regression analysis was carried out to determine the explicative factors.

Results

Between 1997 and 2003, 4,533 women with invasive, noninflammatory, nonmetastatic breast cancer were treated within the framework of the R2C network. Main patient's characteristics are presented in Table I. Mean age was 58 years (Range 24–97 years) and 21.95% of patients were aged 70 years or older. Forty percent (N = 1,833) of the women were treated for every intervention by a member of the R2C network, while 60% (N = 2,715) of our sample was treated not for every intervention (pathological analysis, surgery, radiotherapy, chemotherapy or homonotherapy) by members of this network.

Table I. Main Characteristics of the Sample (N = 4,533 Women)

Characteristics	N	%
Age
<50	1,229	27.1
50–59	1,196	26.4
60–69	1,113	24.5
70+	995	22.0
Unknown	0	0.0
Tumor size (mm)
<20	2,635	58.1
≥20 to <50	1,353	29.8
≥50	476	10.5
Unknown	69	1.5
Regional lymph node
pN0	2,730	60.2
pN+	1,625	35.8
Unknown	178	3.9
Grade
I	1,232	27.2
II	1,853	40.9
III	1,103	24.3
Unknown	345	7.6
Estrogen receptor
Positive	3,529	77.9
Negative	821	18.1
Unknown	183	4.0

The rate of “abnormal” management of surgical re-excision for inadequate margin was found at 12.6%. The main explanatory variable was age above 70 years (odds ratio 4.05) (Table II).

Table II. Results of Multivariate Analysis—Re-Excisaion

Explanatory variables	Coefficient (β)	CI 95%	p (Wald χ ^2)
Variable: re-excision for inadequate margin not carried out despite consensual indication. Initial sample: N = 4,533. Women who underwent initial tumorectomy: N = 3,998. Women who should have undergone surgical re-excision for inadequate margin: N = 1,238. Lack of compliance with the norm: 12.6% (N = 156). R2C is the name of the Cancer Network.
Age ≥70 years vs. younger	4.05	2.74–5.99	<0.0001
pT ≤ 20 mm vs. greater	1.81	1.13–2.89	0.014
Surgeon not member of R2C vs. member	1.58	1.11–2.25	0.013

For nodal exploration, the sampling quality threshold was set at ≥10. In our survey, the rate of lack of compliance was 25.2%. The 2 main explicative factors (Table III) were surgeon's experience (as assessed by yearly activity) and women's age.

Table III. Results of Multivariate Analysis—Quality of Nodal Exploration

Explanatory variables	Coefficient (β)	CI 95%	p (Wald χ ^2)
Variable: Quality of nodal exploration (≥10 sampled nodes). Initial sample: N = 4,533. Women who underwent nodal exploration: N = 3,516. Data available: N = 3,493. Lack of compliance with the norm: 25.2% (N = 879). * Identified chemotherapy unit. R2C is the name of the Cancer Network.
Surgeon's activity: ≤30 operations per year vs. greater	1.50	1.24–1.80	<0.0001
Age ≥ 70 year vs. younger	1.45	1.21–1.74	0.0002
Surgeon not member of R2C vs. member	1.34	1.10–1.69	0.0069
Hospital with an oncologic department*	1.22	1.00–1.50	0.049

The observed rate of “insufficient” irradiation (<60 grays) was 18.2%. Here again, the 2 main explanatory variables were patient's age (with a gradient of impact with increasing age) and the fact that the radiotherapist was not a member of the R2C network (Table IV).

Table IV. Results of Multivariate Analysis—Quality of Radiotherapy

Explanatory variables	Coefficient (β)	CI 95%	p (Wald χ^2)
Variable: Quality of radiotherapy (>60 Gy). Initial sample: N = 4,533. Women who underwent breast radiotherapy: N = 2,987. Data on dose available: N = 2,938. Lack of compliance with the norm: 18.2% (N = 536). R2C is the name of the Cancer Network.
Radiotherapist not member of R2C vs. member	9.77	7.74–12.35	<0.0001
Age (years)			<0.0001
≥80	2.54	1.45–4.44
70–79	1.55	1.13–2.22
60–69	1.69	1.31–2.18
≥59 (reference)	1
Nodal status − vs. +	1.38	1.07–1.77	0.013
Year of treatment: 1997–98 vs. later	1.53	1.15–2.06	0.004

Adjuvant chemotherapy was expected to be delivered to 2,719 patients. In reality, the rate of missing treatment was 16.0%. Here again, the main explicative factor was age (with a gradient of impact with increasing age) (Table V).

Table V. Results of Multivariate Analysis—Adjuvant Chemotherapy

Explanatory variables	Coefficient (β)	CI 95%	P (Wald χ^2)
Variable: Adjuvant chemotherapy not carried out despite consensual indication. Initial sample: N = 4,533. Women who should have undergone adjuvant chemotherapy: N = 2,719. Lack of compliance with the norm: 16.0% (N = 434). * Defined in R2C network as a prognosis staging leading to a different type of chemotherapy.
Age (years)			<0.0001
≤80	24.30	7.1–83.15
70–79	6.27	4.16–9.46
60–69	3.04	2,03–4.56
≤59 (reference)	1
Nodal status: − vs. +	4.33	2.95–6.34	<0.0001
Histological Grade I or II vs. III	1.97	1.37–2.83	0.0003
pT ≤ 20 mm vs. greater	1.60	1.17–2.20	0.0037
ER + vs. −	1.71	1.14–2.56	0.0094
Level II vs. III*	1.87	1.06–3.30	0.03

The rate of missing adjuvant hormone therapy was 13.1%. Age was not an explicative factor (Table VI). In Table VII all significant factors, are summarized.

Table VI. Results of Multivariate Analysis—Hormone Therapy

Explanatory variables	Coefficient (β)	CI 95%	p (Wald χ^2)
Variable: Hormone therapy not carried out despite consensual indication. Initial sample: N = 4,533. Women who should have undergone adjuvant hormone therapy: N = 2,566. Lack of compliance with the norm: 13.1% (N = 337).
Nodal status − vs. +	1.58	1.20–2.08	0.0012
pT ≤ 20 mm vs. greater	1.51	1.15–1.99	0.0032
Year of treatment
1997	16.37	9.90–27.06
1998	9.37	6.27–13.99
1999	6.54	4.40–9.71
2000	3.44	2.34–5.07
2001	2.03	1.31–3.15
2002	1.50	0.86–2.63
2003 (reference)	1

Table VII. Analysis of the Explanatory Variables Found Among all Outcome Variables Explored

Outcome variables	Characteristics of the patient	Characteristics of the tumour	Characteristics of the organization	Time frame
R2C is the name of the Cancer Network.
Lack of re-excision for inadequate margin	Age ≥70 years	pT ≤ 20mm	Surgeon not member of R2C	–
Nodal exploration of <10 sampled nodes	Age ≥ 70 years	–	Surgeon's activity <30 per year	–
			Surgeon not member of R2C
			Hospital without chemotherapy unit
Breast radiotherapy <60 Gy	Older age (gradient)	Negative nodal status	Radiotherapist not member of R2C	Year of treatment 1997–98
Expected (consensual) adjuvant chemotherapy not carried out	Older age (gradient)	Negative nodal status	–	–
		SBR I or II
		pT ≤ 20 mm
		ER+
		“Better” prognosis
Expected (consensual) adjuvant hormonotherapy not carried out	–	Negative nodal status	–	Earlier year (gradient)
		pT ≤ 20 mm

For women treated not for every intervention in the R2C network the mean lack of compliance is 21.1% while for women always treated in the R2C network, the mean rate of lack of compliance is 13.5% (OR1,7 IC_95% 1.55–1.88).

Discussion

Concerning the physicians who participated in the network, no survey or data about their motivation(s) to be involved in the R2C network is available, and there is no comparison with physicians who didn't want to participate. However, the main objective of this survey was not to compare the physicians involved in the network with nonmembers, but instead to measure the rate of compliance of physicians to their own defined norms for disease management, to evaluate the gap between intention and behavior. For the same reason, the issue of representativeness of the studied population of patients (the affected women) might not be relevant, as such, given the objective of our survey. However, mean patient age in our study was 58 years while in France mean age at diagnosis is 61 years.13 Furthermore, the aim of the survey was not to test the validity or the efficacy of the norms, or to determine whether the rules were optimal, but if they were actually followed by those who mainly defined them. Compliance with some practices was not tested in that survey because they were at a very high level of compliance at the start of this network. For instance, the rate of nodal assessment for invasive breast cancer in women aged 69 years or younger and of radiotherapy for breast conservative surgery was already above 99% in 1997.

In this “population-based” survey, the rates of noncompliance with consensual norms ranged from 12.6 (for surgical re-excision for inadequate margin) to 25.2% (for ≥10 sampled lymph nodes). It thus appears that even self-imposed, rational, validated rules are not enough. However, it is difficult in such a study to distinguish between the treatment proposed to the patient that should be the consensual norms, and that which she consented to submit to (the observed one).

These high rates of noncompliance were unexpected, and indeed the design of this survey was expected to demonstrate a low rate of “abnormal” breast cancer management. First, the guidelines were simple rules for frequent conditions; second, the assessment of compliance was made on the files, physicians sent voluntarily to the designated R2C data manager.

What can be learnt from our data? Among the 5 disease management processes tested, we founded 20 explicative factors (Table VII), among which patient's age was in 4 cases associated with largely increased odds ratios and an increasing rate of dysfunction with increasing women's age. This is really a critical and very important issue since in our network women with breast cancer and aged over 70 years accounted for 20.9% of the cases. Well-known prognosis characteristics of the tumors (T and N) and the organization of care were also found as explicative factors on 4 and 3 occasions, respectively, while time frame was found only twice.

This last point, i.e., low impact of time frame, led us to revisit the issue of impact of a “normative” network. The fact that being a nonmember of the network was indeed associated with noncompliance (for re-excision, nodal sample and radiotherapy dose) could at first sight raise the hypothesis that implementation of the network should progressively steer disease management to higher quality (according to norms). If it were the case, time frame should almost always feature among explicative factors. Since it is not the case, we made an alternative hypothesis: it is not the network that imposes norms but, conversely, it is the physicians who were willing to implement norms who formed the network and who were beforehand (and before the setting up of the network) likely to apply the rules.

We believe that some, or many of the observed lack of compliance, are in fact because of the rational clinical practice adjusted to the clinical cases.

Hobbes believed in absolute rule, Rousseau however thought it was a kind of perversion of the social contract. Who was wrong? Are practitioners right when they don't comply with rules and therefore are rule makers (the same!) wrong, and do rules need to be changed particularly for older patients, or, alternatively, are rule makers right and should practitioners be trained and better informed? This is a really critical issue to address for aging patients.

Should decision be case-based (casuistic) or rule-based? The answer may not be that simple. A good physician is definitively not the one with low compliance to rules but also not the one with a 100% compliance rate. Rule makers, besides guidelines, should also set a targeted area of compliance (80%, 90%...) because rules cannot integrate the complexity and diversity of all situations. In our experience, a single disease with a relative homogeneous cultural background for both physician and women, we observed a trade off between on one hand precise guidelines that could reach a high level of compliance (above 90%) but will be complex and not “user friendly” and, on the other hand simpler rules which will aim at 80% of compliance.

Acknowledgements

The authors acknowledge the help and commitment of the physicians of the following cancer clinics: Centre Hospitalier du Pays d'Aix, Centre Hospitalier d'Arles, Clinique La Casamance, Centre Hospitalier de Briançon, Centre Hospitalier de Gap, Centre Hospitalier de Hyères, Centre Hospitalier de Draguignan, Hôpital Paul Desbief, Polyclinique de Furiani, Hôpital Ambroise Paré, Centre Hospitalier de La Ciotat, Centre Hospitalier de Martigues, Clinique “l'Etoile”, Institut Paoli-Calmettes, Centre Hospitalier de Toulon “Font-Pré”, Centre Hospitalier de La Seyne, Centre de Radiothérapie “Saint Louis” de la Croix Rouge.