Forty-seven papers describing 46 studies were considered.
The quality of studies varied. 12 out of 24 (50%) provided information on process. 17 of 32 studies (53%) provided information on outcomes and adjusted the comparison for more than one variable. Only 1 RCT was identified.
Specialization and process of care: 11 observational studies provided information on the impact of specialization for various cancer sites. 5 defined specialisation at the clinician level and 6 at the level of centres. Overall results favoured specialized clinical centres. Only 5 studies adjusted adequately for the case mix between comparison groups. Studies were mostly low-quality and tended to show cancer centres performed specific diagnostic staging procedures more often in breast cancer, childhood cancers and ovarian cancers. Breast conserving surgery (3 studies)was more frequently offered in centres with oncology departments or wards. Mixed results were reported for losses to follow-up.
Proxy definitions of specialization and process of care: 17 studies compared hospital patterns of care according to teaching status (11 studies) and hospital size (5 studies). 13 studies were on breast cancer, 2 on ovarian cancer or included multiple sites. Studies scoring 2 or more on case mix adjustment criteria showed greater reporting of clinical and pathological staging in the notes and greater use of two-stage surgery in larger or teaching centres. Conservative surgical procedures were more commonly used in larger or teaching centres. No difference between non-specialized vs specialized was noted in the use of adjuvant chemotherapy for breast cancer.
Specialization (however defined) and mortality: Generally patients had a lower risk of long-term mortality when treated by specialised centres/clinicians though results from two studies differed.
Specialization (however defined) and mortality for breast cancer (5 studies): all had an adjustment score of 2 or more. Lower 5 year mortality reported when treated in specialist centres or by specialized clinicians OR = 0.82 (95%CI: 0.77, 0.88). Heterogeneity chi-squared = 0.08, P = 0.99. Specialization (however defined) and mortality for haematological cancer (4 studies one of which dealt with 3 types of tumour, giving 6 treatment arms): 5 of the 6 treatment arms showed lower mortality when treated in specialized situations. Specialization (however defined) and mortality for ovarian cancer (7 studies): 6 of 7 studies showed lower mortality when treated in specialized situations. Quality of studies and definition of specialization differed. Heterogeneity chi-squared = 4.5, P = 0.60.
Specialization and mortality for other solid tumours (5 studies): two studies reported statistically significantly lower mortality for colorectal cancer and prostate cancer in teaching vs non- teaching hospitals. Lung cancer (1 study, 2 histological types) results differed according to histology. Testicular cancer (1 study): showed an advantage only for the availability of on-staff urologists and not for oncologist. Few studies focused on types of neurological tumours, sarcomas, or childhood cancers. There was only a limited number of poor quality studies in these fields.
Impact of specialization on outcomes other than long-term mortality. Quality of life in breast cancer (1 RCT): no difference between groups. Studies reporting post-operative/in- hospital mortality in gastrointestinal (1 study), lung (1 study) and ovarian (1 study) showed contradictory results.