Compliance of physicians and patients with a consensus protocol for treatment of advanced breast cancer

In a multicenter study we used a consensus protocol including more than five subsequent therapeutic steps for treatment of patients with advanced breast cancer. A total of 335 evaluable patients from 27 participating hospitals were allocated to a low- or high-risk group, receiving different therapies during the initial phase of treatment. About half of these patients were treated without protocol violations (compliers). The protocol non-compliers were divided into three groups: those receiving more intensive therapy than recommended, those with similarly intensive, and those with less intensive therapy. The reasons for protocol violations were analysed. The intensity of the therapy given actually was correlated with the survival of subgroups. Median survival times were significantly longer in 208 low-risk than in 127 high-risk patients (P less than 0.0001), marginally longer in 165 compliers than in 170 non-compliers (P less than 0.04), significantly longer in low-risk compliers than in low-risk non-compliers (P = 0.002), and significantly shorter in high-risk compliers than in high-risk non-compliers (P = 0.007). Survival of all subgroups of low-risk non-compliers was the same regardless of the actual therapies given. The survival of high-risk patients who received less intensive therapy was significantly longer than that of high-risk compliers (P = 0.015). After six cycles of successful chemotherapy there was no difference, either in time to progression or in survival, between patients who had received either maintenance therapy or no therapy. We postulate that the groups of low-risk and high-risk patients comprised patients with different prognoses. Among low-risk patients, survival of the subgroup with poor prognosis (low-risk non-compliers) was not influenced by therapy. Among high-risk patients, a subgroup with poor prognosis may have been overtreated by using standard chemotherapies as recommended in our consensus protocol.