Commentary. Are current tumour response criteria relevant for the 21st century?

Monitoring the response of tumours to treatment is an integral and increasingly important function of oncological radiologists. Imaging studies allow an objective method of quantifying tumour response to a variety of physical and pharmaceutical treatments. Current morphological imaging techniques depict tumour response as changes in tumour size and composition [1]. Objective tumour shrinkage has been widely adopted as a standard end-point for selecting new anti-cancer drugs for further study, as a prospective end-point for de®nitive clinical trials designed to estimate the bene®t of treatment in a speci®c group of patients, and is widely used in everyday clinical practice to guide clinical decision-making. Measurement of lesions should whenever possible be bidimensional, by multiplying the largest diameter by its perpendicular to give the product. Criteria for assessing therapeutic response were ®rst proposed in 1981 and, although they have been adapted by various cancer organizations, continue to be used largely unchanged [2]. Recently, the World Heath Organisation has adopted a revised set of size criteria incorporating unidimensional measurements (RECIST criteriaÐResponse Evaluation Criteria in Solid Tumours, June 1999) for assessing tumour response [3]. There are many recognized limitations of size change as a tumour response variable. Size changes for both response and progression are arbitrary and are traditionally expressed as a percentage change from baseline. For example, a minimum of 50% reduction in lesion size (equivalent to a 75% reduction in volume) is needed before documenting a partial response to treatment. When multiple lesions are present, all the products are summed. In contrast, progressive disease is de®ned as an increase of 25% in the size of one or more lesions, or as the appearance of new lesion(s). Measurements performed in this way are laborious and beset by numerous errors. These arise from observer variations of the estimated position of the boundary of lesions. The edges of irregular or in®ltrating lesions are often dif®cult to identify and some tumours are impossible to measure; for example, pulmonary lymphangitis carcinomatosa and ovarian carcinoma seeding into the peritoneal space. Furthermore, there is no consensus on whether necrosis and cystic change should be included or excluded when obtaining tumour measurements. The dif®culty in distinguishing peritumoral ®brosis from tumour spread further confounds attempts at measurement. Measurement errors in estimating the size of small lesions often result in misclassi®cation of tumour response; for example, Lavin and Flowerdew [4] showed that the currently used 25% increase in size of lesions can result in a one in four chance of declaring that progression has occurred when in fact the tumour is unchanged [4]. So serious are these errors that ``independent review panels'' are often employed by pharmaceutical companies to standardize the reporting of tumour response in clinical trials. Independent review panels can disagree with ``home radiologists'' in 50% of cases, with major disagreements occurring in up to 40% [5]. The causes for such disagreements include variations in examination technique for obtaining images, in lesion selection and in siting the edges of target lesions. Bidimensional size criteria do not take full account of tumour volume, while state-of-the-art imaging machines can routinely acquire such information. However, three-dimensional volume measurements are likely to encounter errors similar to bidimensional size measurements. A change in size may also be delayed chronologically, often requiring several treatments before a decision can be made on whether a treatment is effective. Reports from the functional imaging literature, particularly positron emission tomography (PET), suggest that metabolic and physiological changes antecede size change [6]. For example, in patients with lymphoma responding to treatment, changes in energetics can be Received 4 January 2000 and in revised form 27 April 2000, accepted 19 May 2000.