This study evaluated compliance with PGL for breast cancer by comparing cases treated before and after their introduction. Patient and tumor characteristics were comparable between pre-PGL and post-PGL patients and the same methods of ascertainment and data collection were used.
In a previous report we investigated the distribution, implementation and evaluation of PGL among clinicians who treat breast cancer in the Piedmont Region. We found that approximately 90% of surgeons, gynecologists, oncologists and radiologists working in the field (70.2% of those who responded to the questionnaire), were aware of PGL within 1 year of their release, and generally had a positive attitude to change their practice accordingly .
In this population-based study we examined clinical practice patterns before and after the introduction of PGL. We and observed good compliance with PGL before their release, and a weak increase in the number of medical decisions that complied with them after their release.
The literature contains several examples of situations in which clinical practice guidelines on breast cancer treatment contributed to an improvement in quality of care [6, 7, 11, 12], but very few of them included a comparison of clinical practice prior to the release of the guidelines , or examined this practice at a population level .
In 1997, Ray-Coquard and collaborators  conducted a study in France on 200 patients, with a before/after design, using information from medical records, and suggested significant changes in the quality of care. These changes were probably due to the introduction of clinical practice guidelines, but their results needed confirmation in a larger sample of cases.
White and collaborators  performed a study in Victoria, Australia in 2004 with a similar design using mailed questionnaires. All cases of early breast cancer registered in the Victorian Cancer Registry during two 6–month periods were selected and a questionnaire was sent to the relevant surgeon about patient characteristics and primary treatments. The study showed an improvement in quality of care after the introduction of clinical practice guidelines. However the study included data provided directly by surgeons from two different surveys. Significantly more surgeons completed the questionnaire in the first survey (73%) than in the second one (52%). The difference in the response rate between the two surveys could have introduced bias, causing a selection of those most interested in the topic, i.e., the surgeons with the highest case load . Furthermore, answers furnished by the physicians could have reflected not necessarily what they did in their clinical practice, but what they knew they should have done to comply with the guidelines.
In our study we collected data directly from clinical records. We found a statistically significant positive trend in two of four quality-of-care indicators concerning diagnosis. In particular we found an improvement in the two indicators that were farthest from the standard in the pre-PGL group (% lesions with cytological/histological diagnosis before surgery and% frozen sections in ≤ 10 mm lesions), but not in the two indicators that already have a good compliance with PGL (percent of histopathological grading available and percent of hormone receptor available).
We measured two important quality-of-care indicators concerning the surgical treatment of breast cancer: percent of BCS in pT1 lesions and percent of BCS performed with free margins, and the results were positive. We noticed a trend of improvement in the post-PGL group: 93% of patients with pT1 were treated surgically with BCS in the present study. We found a similar positive trend concerning the practice of BCS over a 5–year period (2000–2004) in a previous population-based study on women with breast cancer (all ages) carried out in the Piedmont Region using administrative data . Nevertheless, the percentage of single surgery after diagnosis, which was already good before PGL were released, and the percent of reconstruction after mastectomy, which was extremely low, did not show a positive trend over time.
The indicators regarding axillary surgery showed an increased proportion of patients that were treated with axillary clearance with a correct indication. The percent of patients with a clearance of > 9 lymph nodes and the percent of dissections not performed among CIS patients did not reach the standard and did not improve after the release of PGL. Conversely, a higher number of centers performed the SLN technique, with an identification rate that reached the standard.
Finally, looking at the proportion of women who underwent SLN technique by the annual case load of the surgical unit, we found that the use of this technique increased by almost 45% from 2002 to 2004, across all strata of surgical unit annual caseload but in particular in centers treating less than 50 breast cancer a year. This finding is clearly in contrast with the PGL recommendations that suggest the use of SLN technique only in specialized centers (surgical unit annual case load > 50). The increase of use of SLN technique in low caseload centers need to be discouraged.
Between 2002 and 2004 the proportion of women who received radiotherapy after breast cancer surgery (87.7% in 2002 and 87.9% in 2004) was stable, though still far from the standard of 95%. In a previous population-based study in Piedmont, the presence of a radiotherapy unit within the same hospital where the surgical procedure was performed was associated with a higher probability of receiving radiotherapy after discharge. The presence of a radiotherapy unit in the hospital also correlated with the case load and specialization of the surgical unit .
In the analyses of post-surgical medical treatment we found a decrease in the percent of patients with invasive lesions at medium-to-high risk of distant metastasis who received chemotherapy after the introduction of PGL. Furthermore, we observed a decrease in the percent of patients with invasive lesions and low risk of distant metastasis who received inappropriate chemotherapy. The percent of patients who received hormonal therapy was stable in the group with positive estrogen receptor status, and the incorrect prescription of hormonal treatment in estrogen receptor-negative women dramatically decreased in the post-PGL group.
Apart from the introduction of PGL, the positive trend in some of quality-of-care indicators can be partly attributed to the increased proportion of breast cancer cases diagnosed through the regional screening program. In fact, the patients who were diagnosed in the context of the screening program were usually referred to a surgical unit with a high annual case load.
Underestimation of chemotherapy, radiotherapy and/or hormonal therapy was possible given that these treatments are administrated at a different hospital admission than that for the surgical treatment, or even on an ambulatory basis. The information we collected about post-surgical medical treatment was the result of record-linkages between breast cancer patients and the HDR database, radiotherapy outpatient record database and pharmaceutical prescription record database. Such linkages can generate omissions that are likely to be random, so the resulting bias would be conservative.
Finally, we did not see changes in the survival rate between pre-PGL and post-PGL groups. In fact, the majority of PGL recommendations were oriented to avoid invasive surgery, over-treatment, recurrence, or patient anxiety (i.e., avoid mastectomy in pT1 unifocal, avoid more than one surgery after diagnosis, avoid dissection in CIS patients). Very few recommendations were formulated to improve survival (i.e., to measure hormonal receptor availability, to perform radiotherapy in patients treated with BCS, to perform axillary clearance or SLN technique in patients with invasive lesions). Moreover the indicators more related to survival, with the exception of axillary clearance, showed only a negligible improvement after the introduction of PGL. Finally it is possible that, since breast cancer is generally characterized by long survival, a small improvement in survival will manifest itself only with longer follow-up.
The population-based approach of the study ensures that selection bias was minimal and that the results can be considered representative of the entire Piedmont Region.