These early stage trials involve small patient numbers and have the principal aims of establishing the range of toxicities of a particular cytotoxic agent. An appropriate dosage range must be established to use in the next phase of testing. The clinician will always be hopeful of a therapeutic benefit, but this is far from certain and often quite uncommon in Phase I studies. Consequently patient selection is generally confined to individuals who have exhausted standard, conventional therapies but have reasonably good remaining health as characterised by:

The typical Phase I trial involves dosage escalation (Perry 1997) such as the modified Fibonacci schema (Perry 1997), wherein cohorts of three patients are treated at each dose level until significant toxicity occurs. A maximum tolerable dose (MTD) is established once a proportion of patients at a particular dose level experience significant toxicity. A dose of 10–25 % less than MTD is generally recommended for the next phase of the testing.

Having established the MTD of a particular drug, it is then important to establish the magnitude of its anti-tumour activity in a range of malignancies. Phase II trials will ideally recruit patients with specific tumour types and again generally after these patients have progressed despite standard therapies. Sufficient patient numbers must be recruited to have the statistical power which rejects the null hypothesis that the drug has no activity. Generally this requires a confidence level of 90–95 % and statistical theory can provide an estimate of the minimum numbers of patients needed to be treated for a specified drug response rate. For example, a drug with a true response rate of 20 % would require a minimum of 14 patients before one might expect to see a single response (Wittes 1988).

The most commonly used accrual plan for Phase II studies was devised by Gehan (1961). It aims to enrol 14 patients and if no responses are seen the trial is terminated. If one or more responses are seen, then a second stage accrual is performed to more accurately estimate the response probability and confidence limits. An additional 20 patients are usually required. Modifications of this design, such as that devised by Fleming (1982) allow the investigators to set percentage response targets. The design provides a sample size large enough to ensure that a drug can be assessed as meaningfully active or inactive. Confidence intervals can then be employed to clarify what interpretations are consistent with the results (Simon 1986).

Patient selection for Phase II trials is usually quite rigorous. These patients must satisfy all the usual criteria of good performance status and metabolic functions, but they must also have measurable disease by standard imaging techniques such as CT scans. Not all malignancies have clearly measurable disease however, and then other, surrogate, markers may be necessary. For instance, in advanced prostate cancer with bone metastases measurement of the bone lesions is not possible. Here surrogate markers might be prostate specific antigen (PSA) levels, skeletal-related events, performance status and pain scores, which are quantifiable.

Treatments demonstrating significant activity in Phase II studies need to be compared to standard treatment programs. These Phase III trials will randomise patients between two or more regimens. The statistical comparison will be between the new program and older more established treatments. It is important that Phase III studies have sufficient statistical power to detect a difference between the treatment groups. Typically, the sample size needs to be quite large to detect meaningful differences of 10 % or greater between the treatment arms. Sub-stratification to allow for subgroup analyses, which consider the influences of variables such as age, sex, organ function, tumour characteristics and other factors on outcomes, dictates progressively larger patient numbers.

Good clinical practice mandates the use of Evidence-Based Medicine. Most practitioners rely on using treatments which are supported by at least level Ia, Ib or IIa data. In other words, data obtained from well-designed, controlled clinical trials, preferably randomised studies. Herein lies the difficulty for most nuclear oncologists, and clinicians working in cancer care, when evaluating the results of radionuclide studies. The vast majority of these studies provide, at best, level III evidence based upon descriptive data and case studies. Common features of most therapeutic nuclear oncology clinical studies include:

Most combination radionuclide–chemotherapy trials have adopted an extended Phase II design; tending to adhere to fixed-dosage protocols, which unfortunately make the assumption that optimal dosage has already been established. These single arm studies, which, at best, make reference to historical control groups, fail to account for the limitations of this approach. Often the historical control group is poorly defined, of limited size, and no justification is given for its comparability to the current Phase II treatment population. If we look at sample size alone, where our historical control group of (for example 50 patients) had an observed Phase II response of 30 %, we would need to treat 70 patients with the new protocol in order to detect a 20 % improvement in response rate with a statistical power of 0.80 (Simon 1989).

We have seen that clinical trials require careful planning to achieve the study objective. Beyond Phase I, most studies will have multiple objectives which must be stipulated prior to trial commencement, because different objectives will only be achievable by adequate sample size, date collection and patient follow-up. Phase II studies principally aim to provide evidence of the activity of treatment in a disease-specific cohort. The ultimate aim will be to decide whether the new treatment shows sufficient promise to warrant testing against standard therapy in a large comparative trial. Phase II studies will set objectives which can be easily, and rapidly, measured as representative of the classical cytotoxic concept of cancer therapy. The most objective outcomes of cytotoxic treatment will be measures of tumour shrinkage, known as objective response rate (ORR). Tumour responses to chemotherapy are defined by the measurement system known as RECIST (response evaluation criteria in solid tumours) (Eisenhauer 2009). Target lesions will show one of several possible responses to therapy:

Generally, response durations are measured from the start of therapy and the timing of CT/MRI/US evaluations as stipulated in the protocol. The response intervals generate a further set of important clinical trial data sets: