Compliance with clinical practice guidelines for breast cancer treatment: a population-based study of quality-of-care indicators in Italy.

0.36 MB·English

by  Sacerdote, Carlotta

Sacerdoteetal.BMCHealthServicesResearch2013,13:28
http://www.biomedcentral.com/1472-6963/13/28
RESEARCH ARTICLE Open Access
Compliance with clinical practice guidelines for
breast cancer treatment: a population-based
study of quality-of-care indicators in Italy
Carlotta Sacerdote1,6*, Rita Bordon2, Sabina Pitarella2, Maria Piera Mano2, Ileana Baldi1, Denise Casella2,
Daniela Di Cuonzo1, Alfonso Frigerio3, Luisella Milanesio3, Franco Merletti1, Eva Pagano1, Fulvio Ricceri4,
Stefano Rosso2, Nereo Segnan2, Mariano Tomatis2, Giovannino Ciccone1, Paolo Vineis4,5 and Antonio Ponti2
Abstract
Background: It has been documentedthat variations exist inbreast cancer treatment despite wide dissemination
ofclinical practice guidelines.The aim ofthis population-based study was to evaluate the impact of regional
guidelines (Piedmont guidelines,PGL) for breast cancer diagnosis and treatment onquality-of-care indicators in the
Northwestern Italian region of Piedmont.
Methods: We included two samples of women aged 50–69 years with incident breast cancer treated inPiedmont
before and after theintroduction ofPGL: 600in 2002 (pre-PGL) and 621in 2004 (post-PGL). Patients were randomly
selected among all incident breast cancer cases identifiedthrough thehospital discharge records database. We
extracted clinical data onbreast cancer cases from medical charts and ascertainedvital status through linkage with
townoffices. We assessed compliance with 14 quality-of-care indicators from PGLrecommendations, before and
after their introduction inclinical practice.
Results: Among patientswith invasive lesions, 77.1% (N = 368) and 77.5% (N = 383)in the pre-PGL and post-PGL
groups, respectively,received breast conservative surgery (BCS) as a first-line treatment. Following BCS, 87.7%
received radiotherapy in 2002, compared to 87.9%in 2004. Ofall patients atmedium-to-high risk of distant
metastasis, 65.5% (N =268) and 63.6%(N =252) received chemotherapy in2002 and in 2004,respectively. Among
the117 patients with invasive lesions and negative estrogen receptor status in2002, hormonal therapy was
prescribed in 23 ofthem (19.6%). The incorrect prescription ofhormonaltherapy decreased to 10.8%(N = 10)
among the 92 estrogen receptor-negativepatients in2004 (p< 0.01).
Compliance with PGL recommendations was already high inthe pre-PGL group, although some quality-of-care
indicators did not reach thestandard. In thepre/postanalysis, 8out of 14 quality-of-care indicators showed an
improvement from 2002 to 2004,but only 4out of 14 reached statistical significance. We did not find any change
inthe risk of mortality inthe post-PGLversus the pre-PGL group (adjusted hazard ratio 0.94, 95%CI0.56–1.56).
Conclusions: These results highlight the need to continue to improve breast cancer care and to measure
adherence to PGL.
Keywords: Breast cancer, Guideline adherence, Population-based, Evidence-based medicine, Quality of care
*Correspondence:[email protected]
1CancerEpidemiologyUnit,SanGiovanniBattistaHospital,CPOPiemonte
andUniversityofTurin,Turin,Italy
6CancerEpidemiologyUnit,SanGiovanniBattistaHospital,ViaSantena7,
10129,Torino,Italy
Fulllistofauthorinformationisavailableattheendofthearticle
©2013Sacerdoteetal.;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsoftheCreative
CommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,and
reproductioninanymedium,providedtheoriginalworkisproperlycited.
Sacerdoteetal.BMCHealthServicesResearch2013,13:28 Page2of12
http://www.biomedcentral.com/1472-6963/13/28
Background start of the post-PGL period was 1.5 years after the
Evidence-based guidelines serve as a tool to ensure that introduction of PGL, as at this time they were likely to
patients receive treatment based on the best available havebeenimplemented inPiedmonthospitals.
evidence. In 1995, Sainsbury et al. postulated that the We did not include prevalent breast cancer cases,
improvement of clinical practice in breast cancer treat- i.e., women who had been previously hospitalized for
mentcould increase5-year survivalbyupto10%[1]. breast cancer and were recorded in the Piedmont
In the Piedmont Region (Northwestern Italy, popula- Cancer Registry and/or were in the HDR database
tion 4.25 million), clinical practice guidelines (PGL) for between1998and2002.
the treatment of breast cancer were first released in July Forthepurposesofthisstudy,breastcancerwasdefined
2002 and disseminated to all relevant clinicians and by the following International Classification of Diseases 9th
other stakeholders [2,3] Furthermore, as from 1996, the Revision – Clinical Modification (ICD9-CM) codes: 174,
Piedmont Region has been covered bya breast screening carcinoma of the breast; 233.0, in situ carcinoma of the
programforallresident women aged50–69years. breast;238.3,neoplasmofuncertainbehaviorofthebreast;
It has been well documented that there is considerable or 239.3, neoplasm of unspecified nature of the breast, in
variation in breast cancer treatment despite wide dis- anypositionoftheHDRdatabase.
semination of clinical practice guidelines [4]. Variations Fifty-eight patients from the randomly-selected pre-
are usually related not only to patient characteristics, PGL group and 71 patients from the randomly-selected
such as age and educational level, but also to geographic post-PGL group were not eligible for inclusion as they
area of residence and hospital and physician characteris- had benign lesions, prevalent lesions, recurrences, other
tics[5]. cancers, no surgical treatment, or unavailable clinical
Although several studies have examined the different records. After exclusion, 542 patients were left in the
surgical and medical breast cancer treatments employed pre-PGL group and 550 in the post PGL-group, and
following the publication of clinical practice guidelines, were included inthefollowinganalyses (Figure 1).
to our knowledge few reports have included a compari- In situ carcinomas of the breast were over-sampled in
son with clinical practice prior to guideline publication, both groups. To do this, the same inclusion criteria were
and even fewer have examined the impact of guidelines applied, but were not restricted to the first 6 months of
inclinical practiceatapopulationlevel[6,7]. 2002 and 2004. Instead, all women with in situ carci-
To evaluate the real impact of PGL on breast cancer noma (ICD9-CM code 233.0) who were surgically
treatment in Piedmont, we collected data from the treated at any time in 2002 and 2004 were included. The
medical charts of women with breast cancer aged totalnumberofoversampledpatientswithinsitulesions
50–69 years to assess compliance with 14 quality-of-care was 121 in 2002 and 108 in 2004. This over-sampling
indicators, based on PGL recommendations, before and was only used in the analyses concerning the indicators
after the introduction of PGL in clinical practice. Fur- for patients with carcinoma in situ (CIS) in order to
thermore, we explored the use of post-surgical medical increasethepowerofthestudy.
treatment, including radiotherapy, chemotherapy and The accuracy of the method adopted to identify inci-
hormonaltherapyinthetwo periods, according to FIGO dent breast cancer cases was validated using data from
stage, lymph node involvement and hormone receptor the Piedmont Cancer Registry, which covers about 20%
status.Finally,wecomparedthesurvivalratesofpatients of the regional population [8], as a gold standard. The
treated beforeandafterthe introductionofPGL. sensitivity of the algorithm was 76.7% for breast cancer
andthepositivepredictive valuewas92.6%.
Methods Follow-up of patients from both groups was performed
Populationanddatasources through linkage to different databases using two hierar-
The Piedmont hospital discharge records (HDR) data- chicallinkagekeys.TheHDRdatabasewasusedtoidentify
base was used to identify female patients with incident all subsequent relevant hospitalizations (2002–2005) for
breast cancer, aged 50–69 years, residing in Piedmont other surgical treatments (on the same, or other lesions),
and surgically treated in 2002 and 2004 at regional medical complications and chemotherapy sessions. Radio-
hospitals. In 2002 we identified 1,764 female patients, of therapywasassessedthroughlinkagewiththeradiotherapy
whom 866 underwent surgical treatment in the first outpatient record database (which also includes extra-
6 months of 2002. Among these 866 patients, 600 were regional radiotherapy records), and hormonal therapy was
randomly selected for this study (pre-PGL group). In assessed though linkage to the pharmaceutical prescription
2004, 1,777 female patients were identified, of whom record database (which includes all drug prescriptions
905 underwent surgical treatment in the first 6 months reimbursedtopatientsbythepublichealthsystem).
of 2004. Among those 905 patients, 621 were randomly All clinical records of any surgical or radiotherapy
selected for this study (post-PGL group) (Figure 1). The hospitalization identified for patients in the HDR
Sacerdoteetal.BMCHealthServicesResearch2013,13:28 Page3of12
http://www.biomedcentral.com/1472-6963/13/28
Breast cancer cases Breast cancer cases
Piedmont Region 2002 Piedmont Region 2004
Age 50-69 years Age 50-69 years
N incident cancers=1764 N incident cancers=1777
N prevalent cancers=111 N prevalent cancers=117
Surgery in first 6 months of 2002 Surgery in first 6 months of 2004
N incident cancers=866 N incident cancers=905
Randomly-selected study sample: Randomly-selected study sample:
N=600 N=621
(Invasive lesions =521 (Invasive lesions =559
In situ =25 In situ =28
Neoplasm of uncertain behaviour =35 Neoplasm of uncertain behaviour =22
Neoplasm of unspecified nature =19) Neoplasm of unspecified nature =12)
Data from clinical records evaluation:
Not eligible:
30 Benign lesion 32
Study population 17 Prevalent lesions 32 Study population
N=542 2 Recurrences 4 N=550
2 Other cancers 1
0 No surgery 2
7 Not available clinical records 0
Main lesion:
Invasive Microinvasive In situ Invasive Microinvasive In situ
lesions lesions lesions lesions lesions lesions
N=468 N=9 N=65 N=479 N=15 N=56
86.3% 1.7% 12.0% 87.1% 2.7% 10.2%
Figure1Studypopulationofwomenwithincidentbreastcancerwhounderwentsurgeryin2002and2004.
database were retrieved. The clinical data were Mainoutcomemeasures
extracted from medical charts by two breast cancer Fourteen quality-of-care indicators were chosen to evalu-
screening technicians supervised by a gynecologist ate the impact of PGL (Table 1). We explored the use of
and an epidemiologist, working independently of the the sentinel lymph node (SLN) technique by surgical unit
practitioners caring for patients in different hospitals. annualcaseload,tomeasuretheintroductionofthispro-
It was impossible to extract data blinded to the year cedure. Finally, we investigated post-surgical medical
of treatment since the data were obtained directly treatment, including chemotherapy among patients with
from the patients’ records. To control the extracted invasive lesions at low/medium to high risk of distant
data, a random 10% sample was rechecked by the metastasis(accordingtotheGoldhirschscheme)[10],and
supervisors, blinded to the previous decisions. The hormonaltherapyprescribedaccordingtoestrogenrecep-
collected data were entered into a database previously torstatus.
used for clinical audit (the Audit System on Quality
of Breast Cancer Treatment), developed by a multi- Statisticalanalyses
disciplinary team from the European Breast Cancer The differences in distribution between post-PGL and
Network [9]. pre-PGL breast cancer cases according to patient, tumor
The ascertainment of vital status was carried out and surgical unit characteristics and in the rates of
through linkage with town offices, identifying the date patients receiving post-surgical medical treatment were
of death and allowing for the retrieval of the death assessedbytwo-wayChisquareTestorFisherExactTest.
certificate to identify the specific causes of death The 14 quality-of-care indicators (Table 1) were ana-
(2002–2010). All procedures concerning death certi- lyzed as dichotomous variables by multivariable logistic
ficates, data collection and coding were applied regression models, using the period (post-PGL versus
uniformly to both groups. pre-PGL) as the main effect and controlling for
Sacerdoteetal.BMCHealthServicesResearch2013,13:28 Page4of12
http://www.biomedcentral.com/1472-6963/13/28
Table1PiedmontRegionclinicalpracticeguidelinesquality-of-careindicators
Levelofevidence(AHRQ Quality-of-careindicators Function Quality-of-
gradeofunderlying care
recommendation) standards
B %ofmalignantlesionswith Calculatestheproportionofpatientswithinvasiveorinsitu ≥70%
cytologicalorhistologicalpre- lesionswithapre-operativecytologicalorhistologicaldiagnosis
operativediagnosis (C5orB5),outofthetotalnumberofpatientswithinvasiveorin
situlesionswhounderwentsurgicaltreatment.
A %BCSinpT1,unifocal Calculatestheproportionofpatientsdiagnosedwithinvasive ≥80%
lesionsofapathologicalsizeof≤20mm(pT1,microinvasive
included),notclinicallymulticentricormultifocal,whowere
treatedwithBCS.
B %BCSwithfreemargins(>1mm) CalculatestheproportionofBCS(lastBCSifmorethanone)for /
invasiveorinsitulesionswhichensuredclearmargins(distance
>1mmfromthelesion),outofthetotalnumberofBCS
performed.
/ %singlesurgeryafterdiagnosis Calculatestheproportionofpatientswhosefirstsurgical /
treatmentwasnotfollowedbyfurtherlocaloperationsthatwere
requiredduetoincompleteexcision(excludingfailedbiopsies),
outofthetotalnumberofpatientswhoweresurgicallytreated
forlocalizedinvasiveorinsitulesionswithapositiveor
suspiciouscytologicalorhistologicalpre-operativediagnosis.
C %frozensectioninlesions≤10mm Calculatestheproportionofpatientssurgicallytreatedfor ≥95%
invasivelesions(excludingmicroinvasivelesions)ofamaximum
pathologicalsize≤10mmforwhichtherewasnofrozensection,
outofthetotalnumberofpatientswiththesamediagnosis.
A %ofpatientswithinvasivelesions Calculatestheproportionofpatientswithinvasivelesionswho /
treatedwithaxillaryclearanceor weretreatedwithaxillarydissectionorSLNtechnique,outofthe
SLNtechnique totalnumberofpatientswithinvasivelesions.
C %ofpatientstreatedwithaxillary Calculatestheproportionofpatientswithinvasivelesionswho ≥95%
clearancewith>9lymphnodes weretreatedwithaxillaryclearance(levelI-III),excluding
sampling,andfromwhomatleast10lymphnodeswereexcised,
outofthetotalnumberofpatientswithinvasivelesionswho
weretreatedwithaxillaryclearance.
C %withNOdissectionamongCIS CalculatestheproportionofpatientsdiagnosedwithCISornot ≥95%
patients otherwisespecifiedinsitulesions(microinvasivecancerexcluded)
onwhomnoaxillarydissectionwasperformed(notevenlevelI)
outofthetotalnumberofpatientswiththisdiagnosiswhowere
surgicallytreated.
C %correctSLNidentification Calculatesthe%ofSLNidentifiedoutofthetotalofidentified ≥90%
SLNinpatientswithinvasivelesions,whoweretreatedwithSLN
technique.
C %histopathologicalgrading Calculatestheproportionofpatientswithinvasivelesions ≥95%
available (excludingmicroinvasivecancer)whoweresurgicallytreatedand
forwhommeasuringwasprovided,outofthetotalnumberof
patientswithinvasivelesionswhounderwentsurgicaltreatment.
C %hormonalreceptoravailability Calculatestheproportionofpatientswithinvasivelesions ≥95%
(excludingmicroinvasivecancer)whoweresurgicallytreatedand
forwhommeasuringwasprovided,outofthetotalnumberof
patientswithinvasivelesionswhounderwentsurgicaltreatment.
/ %immediatereconstructionafter Calculatestheproportionofpatientswithinvasiveorinsitu /
mastectomy lesionswhohadmastectomiesandimmediatereconstruction,
outofthetotalnumberofpatientstreatedwithmastectomy.
A %radiotherapyinpatientstreated CalculatestheproportionofpatientswhoweretreatedwithBCS ≥95%
withBCS forinvasiveorinsitulesions,andforwhomradiotherapy
followed,outofthetotalnumberofpatientswiththesame
diagnosiswhoweretreatedwithBCS.
A %ofeligiblepatientsthatreceive Calculatestheproportionofpatientswithinvasivelesionsand /
hormonaltherapy positiveestrogenreceptorswhoreceivedhormonaltherapy.
AHRQgradeofunderlyingrecommendation(referencehttp://archive.ahrq.gov/clinic/epcarch.htm),BCS:breastconservativesurgery,pT1:smallprimarytumor,
SLN:sentinellymphnode,CIS:carcinomainsitu.
Sacerdoteetal.BMCHealthServicesResearch2013,13:28 Page5of12
http://www.biomedcentral.com/1472-6963/13/28
Table2Distributionofbreastcancercasesaccordingtovariouspatient,tumorandsurgicalunitcharacteristics,2002
and2004,PiedmontRegion,Italy
Nand%ofpatients Effectofyear
2002 2004 (pvalue)
Samplesize(N) 542(49.6%) 550(50.4%)
Age(years)
50–54 134(24.7%) 124(22.5%) 0.35
55–59 111(20.5%) 122(22.2%)
60–64 166(30.6%) 151(27.5%)
65–69 131(24.2%) 153(27.8%)
Missing 0 0
Educationallevel
Bachelor 69(12.7%) 99(18.0%) 0.11
Secondary 151(27.9%) 146(26.5%)
Professional 222(40.9%) 197(35.8%)
Intermediateandprimary 85(15.7%) 88(16.0%)
Noneandunknown 15(2.8%) 20(3.6%)
PathologicalTstage
InSitu 62(11.4%) 55(10.0%) 0.54
1mic 10(1.8%) 15(2.7%)
1 4(0.7%) 5(0.9%)
1a 17(3.2%) 25(4.5%)
1b 75(13.8%) 85(15.5%)
1c 186(34.3%) 189(34.4%)
2 147(27.2%) 136(24.7%)
3 5(0.9%) 10(1.8%)
4 5(0.9%) 5(0.9%)
4a 3(0.6%) 0(0%)
4b 18(3.3%) 15(2.7%)
4c 1(0.2%) 0(0%)
4d 2(0.4%) 2(0.4%)
X 4(0.7%) 2(0.4%)
Missing 3(0.6%) 6(1.1%)
PathologicalNstage
0 205(43.3%) 158(31.3%) <0.01
Sentinellymphnode 80(16.9%) 167(33.1%)
1 21(4.4%) 37(7.3%)
1a 20(4.2%) 72(14.3%)
1b 7(1.5%) 3(0.6%)
1b1 40(8.4%) 2(0.4%)
1b2 7(1.5%) 3(0.6%)
1b3 56(11.8%) 3(0.6%)
1b4 11(2.3%) 3(0.6%)
2 15(3.2%) 25(5.0%)
3 0 23(4.6%)
X 9(1.9%) 6(1.2%)
Missing 2(0.4%) 2(0.4%)
Sacerdoteetal.BMCHealthServicesResearch2013,13:28 Page6of12
http://www.biomedcentral.com/1472-6963/13/28
Table2Distributionofbreastcancercasesaccordingtovariouspatient,tumorandsurgicalunitcharacteristics,2002
and2004,PiedmontRegion,Italy(Continued)
PathologicalTNMstage
Insitu 65(12.0%) 56(10.2%) 0.04
I 189(34.8%) 233(42.3%)
IIA 127(23.4%) 112(20.4%)
IIB 75(13.8%) 53(9.6%)
IIIormore 47(8.8%) 65(11.9%)
Missing 39(7.2%) 31(5.6%)
Grading(invasiveonly)
Low 58(12.0%) 82(16.5%) 0.09
Intermediate 69(14.4%) 86(17.4%)
High 103(21.5%) 89(18.4%)
Notperformedandmissing 250 238
Diseasedetectedthroughregionalscreeningprogram
Yes 208(38.4) 236(42.9%) 0.001
No,symptomatic 209(38.6%) 171(31.9%)
No,asymptomatic 110(20.3%) 112(20.4%)
Missing 15(2.8%) 31(5.6%)
Surgicalunitannualcaseload
<50 124(22.9%) 126(22.9%) 0.001
50–149 249(45.9%) 250(45.5%)
≥150 155(28.6%) 147(26.8%)
Missing 14(2.6%) 27(4.8%)
confounders (age, educational level, clinical stage, screen- statistically significant difference in the distribution of
ing provenience and surgical unit annual case load). The clinical stages between 2002 and 2004. The proportion of
results are presented as frequencies, adjusted odds ratios cases diagnosed through the regional screening program
(OR ) and 95% confidence intervals (95%CI) estimated increasedfrom38.4%(N=208)in2002to42.9%(N=236)
adj
from logistic models to measure the probability of achie- in2004(p=0.001).OfallpatientsdiagnosedwithCIS,the
ving standards in 2004 compared to 2002. We performed percent diagnosed by the regional screening program
Cox proportional hazards regression analyses to study increased from 45.9% (N = 28) in 2002, to 63.6% (N = 35)
survival.Proportionalhazardassumptionsweretestedwith in2004(datanot shown). The annual case loadofsurgical
theGrambschandTherneautestbeforeanalysis.Statistical unitswasstableoverthetwoperiods(Table2).
analyseswereperformedusingSAS8.2andSTATAv10.
Results Surgicalandmedicaltreatment
Patientcharacteristics In the pre-PGL group 77.1% (N = 368) of patients with
Among the women included in the analyses, the distribu- invasive lesions received BCS as a first-line treatment, and
tion by type of lesion in the pre-PGL group was: 88.0% for62.7%(N=299)ofpatientsinthepre-PGLgroup,BCS
invasive or microinvasive (N = 477) (ICD9-CM code 174), was a definitive surgical treatment. In the post-PGL group
and 12.0% in situ (N = 65) (ICD9-CM code 233.0); in the the numbers were 77.5% (N = 383) and 66.6% (N=329),
post-PGL group it was 89.8% invasive or microinvasive respectively(Figure2).
(N = 494), and 10.2% in situ (N = 56) (non-statistically As for post-surgical medical treatment, following BCS,
significantdifference)(Table1). 87.7% (N = 341) of the pre-PGL group and 87.9%
Table 2 shows the distribution of breast cancer cases by (N = 362) of the post-PGL group received radiotherapy
patient,tumorandcareprovidercharacteristicsforthepre- alone, or in combination with chemotherapy (Figure 3).
PGL and post-PGL groups. Both groups had similar age Post-surgical treatment with adjuvant chemotherapy was
and educational level distribution. In contrast, there was a received by 50.9% (N = 276) of the pre-PGL group (about
Sacerdoteetal.BMCHealthServicesResearch2013,13:28 Page7of12
http://www.biomedcentral.com/1472-6963/13/28
2002 2004
Invasive and microinvasive In situlesions Invasive and microinvasive In situlesions
lesions N=65 lesions N=56
N=477 12.0% N=494 10.2%
88.0% 89.8%
1stsurgery
Mast BCS Other or Mast BCS Other or Mast BCS Other or Mast BCS Other or
N=102 N=368 unknown N=5 N=58 unknown N=108 N=383 unknown N=5 N=51 unknown
21.4% 77.1% N=7 7.7% 89.2% N=2 21.9% 77.5% N=3 8.9% 91.1% N=0
1.5% 3.1% 0.6% 0.0%
No other surgery No other surgery No other surgery No other surgery
N=299 N=43 N=329 N=42
81.3% 74.2% 85.9% 82.3%
2ndsurgery
Other Mast Other Mast Other Mast Other Mast
BCS N=31 BCS N=8 BCS N=18 BCS N=6
N=38 8.4% N=7 13.8% N=36 4.7% N=3 11.9%
10.3% 12.0% 9.4% 5.8%
Mast: mastectomy, BCS: breast conservative surgery.
Figure2Surgicaltreatmentinwomenwithincidentbreastcancerwhounderwentsurgeryduringthefirst6monthsof2002
and2004.
46%followingBCSand67%followingmastectomy)andby nodes, % with no dissection among CIS patients, %
46.2%(N=254)ofthepost-PGLgroup(about39%follow- radiotherapy inpatientstreated withBCS).
ingBCSand69%followingmastectomy)(datanotshown). In the pre/post analysis, studying the probability of
Post-surgical medical treatment of invasive lesions is achieving standards in 2004 compared to 2002, although
shown in Tables 3 and 4. Of all patients with invasive eightofthe14examinedquality-of-careindicatorschanged
lesions at medium-to-high risk of distant metastasis, in the expected direction, only four indicators improved
64.6%receivedchemotherapy(N=520):65.5%inthepre- substantially from 2002 to 2004: percent of malignant
PGL group and 63.6% in the post-PGL group. Among lesionswithcytologicalorhistologicaldiagnosisbeforesur-
113 women with invasive lesions at low risk of distant gery(OR 0.64,95%CI0.49–0.85),percentofBCSinpT1
adj
metastasis14.3%and1.4%receivedchemotherapyin2002 lesions and percent of BCS performed with free margins
and2004,respectively(Table3). (OR 0.41, 95% CI 0.22–0.75 and OR 0.65, 95% CI
adj adj
Hormonal therapy was prescribed to 684 out of 901 0.41–1.01 respectively) and percentage of frozen section in
(75.9%) patients with invasive lesions. Of women who lesions≤10mm(OR 0.32,95%CI0.16–0.65).Theother
adj
were prescribed hormonal therapy, 4.8% (N = 33) had indicatorswerestableinthetwoperiods(Table5).
negative estrogen receptor status. The incorrect pre- The number and percent of women treated with SLN
scription of hormonal therapy in patients with negative technique by surgical unit annual case load is shown in
receptor status decreased from 19.6% (N = 23) in 2002 Table 6. The PGL recommendations said that the SLN
to10.8%(N =10)in2004(p<0.01)(Table 4). technique should be performed, as an alternative to axil-
lary dissection, only by surgical units with a high annual
CompliancewithPGLrecommendations case load. The percent of patients who were treated with
Univariate analysis showed that compliance with PGL SLN technique by a surgical unit with an annual case
recommendations was already high in the pre-PGL load of < 50 (based on the total number of women who
group. Indeed, five of nine quality-of-care indicators did underwent surgery in low-case load units) was 19.1% in
not achieve the proposed standard (% of malignant 2002 (N = 25) and 42.3% in 2004 (N = 55). There was a
lesions with cytological or histological pre-operative big increase from 2002 to 2004 in the use of this tech-
diagnosis, % frozen section in lesions ≤ 10 mm,% of nique across all strata of annual surgical case load, but
patients treated with axillary clearance with > 9 lymph we observed the biggest increase in the lowest category
Sacerdoteetal.BMCHealthServicesResearch2013,13:28 Page8of12
http://www.biomedcentral.com/1472-6963/13/28
2002 2004
Definitive
surgery
BCS Mast BCS Mast
N=389 N=146 N=412 N=137
RT only RT+ CT CT only RT only RT+ CT CT only RT only RT+ CT CT only RT only RT+ CT CT only
N=172 N=169 N=9 N=2 N=30 N=68 N=214 N=148 N=12 N=10 N=36 N=58
44.2% 43.4% 2.3% 1.4% 20.4% 46.6% 51.9% 36.0% 2.9% 7.3% 26.3% 42.3%
Some treatment: N=350 (90%) Some treatment: N=100 (68.5%) Some treatment: N=374 (90.8%) Some treatment: N=104 (75.9%)
No treatment: N=39 (10%) No treatment: N=46*(31.5%) No treatment: N=38 (9.2%) No treatment: N=33 (24.1%)
*One patient refused CT treatment.
BCS: breast conservative surgery, Mast: mastectomy, RT: radiotherapy, CT: chemotherapy.
Figure3Post-surgicalmedicaltreatmentinwomenwithincidentbreastcancerwhounderwentsurgeryduringthefirst6monthsof
2002and2004.
(annual case load < 50 = 55% increase, 50–149 = 40% In a previous report we investigated the distribution,
increase,≥150=41%increase)(Table6). implementation and evaluation of PGL among clinicians
who treat breast cancer in the Piedmont Region. We
found that approximately 90% of surgeons, gynecolo-
Survivalanalyses
gists, oncologists and radiologists working in the field
Between2002and2010,atotalof101deathswereidentified:
(70.2% of those who responded to the questionnaire),
52 women in the pre-PGL and 49 in the post-PGL group
were aware of PGL within 1 year of their release, and
(90.4%and90.1%crude5–yearsurvival,respectively).
generally had a positive attitude to change their practice
We did not find any change in the risk of mortality in
accordingly[3].
the post-PGLversus the pre-PGL group (HR 0.94, 95%CI
In this population-based study we examined clinical
0.56–1.56adjusted for age, clinical stage and surgicalunit
practice patterns before and after the introduction of
annualcaseload).
PGL. We and observed good compliance with PGL
before their release, and a weak increase in the number
Discussion of medical decisions that complied with them after their
This study evaluated compliance with PGL for breast release.
cancer by comparing cases treated before and after their The literature contains several examples of situations
introduction. Patient and tumor characteristics were in which clinical practice guidelines on breast cancer
comparable between pre-PGL and post-PGL patients treatment contributed to an improvement in quality of
and the same methods of ascertainment and data collec- care[6,7,11,12],but very fewof themincluded acompari-
tion wereused. sonofclinicalpracticepriortothereleaseoftheguidelines
[6],orexaminedthispracticeatapopulationlevel[7].
Table3Post-surgicalchemotherapybyriskofdistant
metastasisaccordingtoERandlymphnodestatus,2002
and2004,PiedmontRegion,Italy Table4Post-surgicalhormonaltherapyinpatientswith
LowRisk* Medium-to-highrisk§ p invasivelesionsaccordingtoestrogenreceptor(ER)
Nand% Nand% value status,2002and2004,PiedmontRegion,Italy
2002 6/42 268/409 0.04 Estrogen Estrogen p
receptor-positive receptor-negative value
(14.3%) (65.5%)
Nand% Nand%
2004 1/71 252/396
2002 319/338 23/117 <0.01
(1.4%) (63.6%)
(88.0%) (19.6%)
ER:estrogenreceptor.
*ER-positiveandsize≤2cmandgrading=1withnegativelymphnodestatus. 2004 332/354 10/92
§ALLpatientswithpositivelymphnodestatusorestrogenreceptor-negative (89.8%) (10.8%)
orsize>2cmorgrading≥2regardlessoflymphnodestatus.
Sacerdoteetal.BMCHealthServicesResearch2013,13:28 Page9of12
http://www.biomedcentral.com/1472-6963/13/28
Table5AchievementofPiedmontRegionclinicalpracticeguidelinesquality-of-carestandardsin2002and2004,and
effectsofyear,PiedmontRegion,Italy
Results Effectofyear(adjusted*ORand95%CI)
2002vs2004
2002 2004
Nand% Missing Nand% Missing
%ofmalignantlesionswithcytologicalor 297/513 48 329/495 65 0.64(0.49-0.85)
histologicalpre-operativediagnosis
(57.9%) (8.6%) (66.5%) (11.6%)
%BCSinpT1,unifocal 231/268 0 253/272 0 0.41(0.22-0.75)
(86.2%) (93.0%)
%BCSwithfreemargins(>1mm) 312/368 33 348/387 33 0.65(0.41-1.01)
(84.8%) (8.2%) (89.9%) (7.9%)
%singlesurgeryafterdiagnosis 332/369 4 342/376 8 1.02(0.24-5.91)
(90.0%) (1.1%) (91.0%) (2.1%)
%frozensectioninlesions≤10mm 55/97 1 85/113 0 0.32(0.16-0.65)
(56.7%) (1.0%) (75.2%)
%ofpatientswithinvasivelesiontreated 433/468 0 460/478 1 0.28(0.10-0.85)
withaxillaryclearanceorSLNtechnique
(92.5%) (96.2%) (0.2%)
%ofpatientstreatedwithaxillary 317/336 4 245/268 4 1.71(0.89-3.31)
clearancewith>9lymphnodes
(94.3%) (1.2%) (91.4%) (1.5%)
%withNOdissectionamongCISpatients** 115/123 0 98/104 0 0.60(0.14-2.50)
(93.5) (94.2%)
%correctidentificationofSLN 118/126 30 215/219 54 0.46(0.18-1.05)
(93.6%) (19.2%) (98.2%) (19.8%)
%histopathologicalgradingavailable 455/462 18 466/469 17 0.34(0.08-1.42)
(98.5%) (3.7%) (99.4%) (3.5%)
%hormonalreceptoravailability 455/471 9 446/473 13 1.80(0.94-3.48)
(96.6%) (1.8%) (94.3%) (2.7%)
%immediatereconstructionaftermastectomy 41/154 4 30/137 2 1.16(0.63-2.16)
(26.6%) (2.5%) (21.9%) (1.4%)
%radiotherapyinpatientstreatedwithBCS 341/389 0 362/412 0 1.01(0.60-1.5)
(87.7%) (87.9%)
%ofeligiblepatientsthatreceivehormonaltherapy 319/338 0 332/354 0 1.01(0.69-1.42)
(94.3%) (93.8%)
The2004periodisthereferencecategoryfortheestimationofORs.
*Adjustedforage,educationallevel,clinicalstage,screeningprovenienceandsurgicalunitannualcaseload.
**CalculatedonCISlesionsover-sampledgroup.
OR:oddsratio,BCS:breastconservativesurgery,pT1:smallprimarytumor,CIS:carcinomainsitu,SLN:sentinellymphnode.
In 1997, Ray-Coquard and collaborators [6] conducted 6–month periods were selected and a questionnaire was
a study in France on 200 patients, with a before/after sent to the relevant surgeon about patient characteristics
design, using information from medical records, and and primary treatments. The study showed an improve-
suggested significant changes in the quality of care. ment in quality of care after the introduction of clinical
These changes were probably due to the introduction of practice guidelines. However the study included data
clinical practice guidelines, but their results needed provided directly by surgeons from two different surveys.
confirmationinalarger sampleofcases. Significantly more surgeons completed the questionnaire
White and collaborators [7] performed a study in in the first survey (73%) than in the second one (52%).
Victoria, Australia in 2004 with a similar design using The difference in the response rate between the two
mailed questionnaires. All cases of early breast cancer surveys could have introduced bias, causing a selection of
registered in the Victorian Cancer Registry during two those most interested in the topic, i.e., the surgeons with
Sacerdoteetal.BMCHealthServicesResearch2013,13:28 Page10of12
http://www.biomedcentral.com/1472-6963/13/28
Table6Surgicaltreatmentwithsentinellymphnode centers performed the SLN technique, with an identifi-
techniqueaccordingtosurgicalunitannualcaseload, cation rate thatreachedthe standard.
2002and2004,PiedmontRegion,Italy Finally, looking at the proportion of women who
Surgicalunit Sentinellymphnodetechnique underwentSLNtechniquebytheannualcase load ofthe
annualcaseload 2002 2004 surgical unit, we found that the use of this technique
Nand% Missing Nand% Missing increased by almost 45% from 2002 to 2004, across all
strata of surgical unit annual caseload but in particular
<50 25/131 3 55/130 3
incenters treatinglessthan50breastcancer ayear. This
(19.1%) (2.2%) (42.3%) (2.2%)
finding is clearly in contrast with the PGL recommenda-
50–149 67/244 7 113/250 1
tions that suggest the use of SLN technique only in
(27.4%) (2.8%) (45.2%) (0.4%) specialized centers (surgical unit annual case load > 50).
≥150 55/146 8 95/149 0 The increase of use of SLN technique in low caseload
(37.7%) (5.2%) (63.7%) centers needtobediscouraged.
Between 2002 and 2004 the proportion of women who
Overall 147/521 21 263/529 21
received radiotherapy after breast cancer surgery (87.7%
(28.2%) (3.9%) (49.7%) (3.8%)
in 2002 and 87.9% in 2004) was stable, though still far
from the standard of 95%. In a previous population-
thehighestcaseload[13].Furthermore,answersfurnished based study in Piedmont, the presence of a radiotherapy
bythephysicianscouldhavereflectednotnecessarilywhat unit within the same hospital where the surgical pro-
theydidintheirclinicalpractice,butwhattheyknewthey cedure was performed was associated with a higher
shouldhavedonetocomplywiththeguidelines. probability of receiving radiotherapy after discharge. The
In our study we collected data directly from clinical presence of a radiotherapy unit in the hospital also
records. We found a statistically significant positive trend correlated with the case load and specialization of the
intwooffourquality-of-careindicatorsconcerningdiagno- surgicalunit[14].
sis.Inparticularwefoundanimprovementinthetwoindi- In the analyses of post-surgical medical treatment we
catorsthatwerefarthestfromthestandardinthepre-PGL found a decrease in the percent of patients with invasive
group (% lesions with cytological/histological diagnosis lesions at medium-to-high risk of distant metastasis who
before surgery and% frozen sections in ≤ 10 mm lesions), received chemotherapy after the introduction of PGL.
but not in the two indicators that already have a good Furthermore, we observed a decrease in the percent of
compliancewithPGL(percentof histopathologicalgrading patients with invasive lesions and low risk of distant
availableandpercentofhormonereceptoravailable). metastasis who received inappropriate chemotherapy.
We measured two important quality-of-care indicators The percent of patients who received hormonal therapy
concerning the surgical treatment of breast cancer: per- was stable in the group with positive estrogen receptor
centofBCSinpT1lesionsandpercentofBCSperformed status, and the incorrect prescription of hormonal treat-
with free margins, and the results were positive. We ment in estrogen receptor-negative women dramatically
noticedatrend of improvement in the post-PGL group: decreasedinthe post-PGL group.
93% of patients with pT1 were treated surgically with Apart from the introduction of PGL, the positive trend
BCS in the present study. We found a similar positive insomeofquality-of-careindicatorscanbepartlyattribu-
trend concerning the practice of BCS over a 5–year ted to the increased proportion of breast cancer cases
period (2000–2004) in a previous population-based diagnosedthroughtheregionalscreeningprogram.Infact,
study on women with breast cancer (all ages) carried the patients who were diagnosed in the context of the
out in the Piedmont Region using administrative data screeningprogramwereusuallyreferredtoa surgicalunit
[14]. Nevertheless, the percentage of single surgery withahighannualcaseload.
after diagnosis, which was already good before PGL Underestimation of chemotherapy, radiotherapy and/or
were released, and the percent of reconstruction after hormonal therapy waspossible given that these treatments
mastectomy, which was extremely low, did not show a areadministratedatadifferenthospitaladmissionthanthat
positive trend over time. for the surgical treatment,or even onanambulatorybasis.
The indicators regarding axillary surgery showed an The information we collected about post-surgical medical
increased proportion of patients that were treated with treatmentwastheresultofrecord-linkagesbetweenbreast
axillary clearance with a correct indication. The percent cancer patients and the HDR database, radiotherapy out-
of patients with a clearance of > 9 lymph nodes and the patient record database and pharmaceutical prescription
percent of dissections not performed among CIS record database. Such linkages can generate omissions
patients did not reach the standard and did not improve that are likely to be random, so the resulting bias would
after the release of PGL. Conversely, a higher number of beconservative.