Food and Drug Administration commentary on methodological issues in negative symptom trials.

Although the Food and Drug Administration (FDA) has generally discouraged drug development programs focusing narrowly on certain features of a defined psychiatric syndrome when other drugs in the class are considered to be treatments for the broad syndrome, we have made exceptions when there is a legitimate basis for targeting selected aspects of a psychiatric illness. Both “negative symptoms” and “cognitive impairment” associated with schizophrenia are recognized features of schizophrenia that are not adequately treated by available antipsychotic therapies. Patients with acute exacerbations of schizophrenia are treated with either conventional or atypical antipsychotic agents. These drugs provide benefit mostly for positive symptoms of schizophrenia, and patients are typically left with substantial negative symptoms and cognitive impairment as they enter what is referred to as the “residual phase” of this illness. Thus, FDA has endorsed both “negative symptoms”1 and “cognitive impairment”2 as legitimate clinical targets for drug development. The paper by Marder et al, on “Methodological Issues in Negative Symptom Trials” in this issue of the Schizophrenia Bulletin reports on a “Workshop on Methodological Issues in Negative Symptom Trials” held at a meeting of the International Society for CNS Clinical Trials Methodology (ISCTM) in San Diego, California, in September 2009. This workshop, which included representation by FDA, provided an update on a number of issues pertinent to the conduct of clinical trials focusing on negative symptoms of schizophrenia, and FDA either agrees with or has no comment on the summary statements for most of the issues addressed in this paper. We do, however, want to comment on 2 issues addressed in this paper. The first is the issue of whether or not a co-primary assessment of functional improvement would be needed along with a primary measure of negative symptoms in such a trial. The second issue is the optimal study design for a “broad-spectrum” antipsychotic agent, and, importantly, how to interpret the outcome of such a trial. In addition, we would like to comment on another issue not raised in the paper, ie, possible overlap in the constructs “negative symptoms” and “cognitive impairment” in schizophrenia. It is true that FDA has required a co-primary measure of functional capacity in studies targeting cognitive impairment in schizophrenia, for all the reasons given in the paper of Marder et al. They also accurately report that FDA will not be requiring a co-primary measure in studies targeting negative symptoms because the negative symptom construct has obvious face validity and the measures for assessing negative symptoms can be more easily interpreted with regard to clinical relevance than can modest changes in a cognitive battery. Nevertheless, this should not be taken as a discouragement from attempting to assess possible functional changes related to improvements in negative symptoms. Whether or not such assessments would be a regulatory requirement for a negative symptoms trial, it would be of interest to know to what extent improvements in negative symptoms translate into improvements in real world function or at least functional capacity. There is increasing interest in assessing functional improvement in clinical trials generally, and many registration trials now include measures of function or functional capacity along with primary measures of symptomatic change. These additional measures are often included as “key” secondary measures, which means that a replicated positive outcome on such a measure, even though not a requirement to declare a study positive, could be mentioned in labeling for a drug as additional useful information characterizing the benefit of a drug. There is obvious interest in trying to understand the effect on function of modest changes in symptomatic scales even though there is no doubt about the clinical relevance of the symptoms themselves that are included in such instruments. Observing changes in function or even functional capacity is challenging in relatively short trials because it might be unrealistic to expect much movement in these domains in a trial lasting only a few weeks. Given that negative symptom trials will generally be much longer, it might not be unreasonable to expect some movement in functional capacity. As noted by Marder et al, 2 different study designs would be appropriate for studies targeting negative symptoms in schizophrenia. The most straightforward and easiest to interpret would be the add-on trial for an agent that is viewed as a compound specifically targeting negative symptoms in patients whose positive symptoms are controlled with a conventional or atypical antipsychotic agent and who would continue on that agent during the negative symptom trial to maintain control of the positive symptoms. FDA also views this as the optimal study design for this clinical target. The alternative study design would be for a so-called “broad-spectrum” agent, ie, one intended to treat both positive symptoms and negative symptoms. Marder et al propose that the only feasible design for such an agent would be one that involved a run-in period during which patients would be stabilized on a standard antipsychotic agent, either conventional or atypical, and then a double-blind phase during which patients would be randomized to either continue on that standard agent or would be switched to the new “broad-spectrum” agent. We agree that this is the only feasible design for a broad-spectrum agent. We disagree with Marder et al, however, in the interpretation of a finding of a difference in negative symptoms between the two agents in a trial of this design. Marder et al acknowledge the problem of misinterpretation of a difference that is based on what might be called secondary as opposed to primary negative symptoms, ie, related to poorly controlled positive symptoms, extrapyramidal symptoms, or depression caused by the standard agent. They feel that this problem can be addressed by selection of an appropriate standard agent and dose for that agent. We are not confident that this problem can be addressed in this manner, and we have previously discussed this concern.1 In our view, without a placebo arm in such a trial, a finding of an advantage for the new “broad-spectrum” agent over the “standard” agent on negative symptoms cannot be clearly interpreted with regard to an effect on primary negative symptoms. One interpretation, of course, would be that the new agent improves negative symptoms, and the standard agent is having no effect. A second interpretation would be that the new agent in fact has little benefit on negative symptoms but the standard agent is actually causing negative symptoms. A third possibility is that in fact both agents are causing negative symptoms, but the newer agent has less of this negative effect. It's not that we don't feel that such a difference is a benefit but rather that such a difference without a placebo arm for reference cannot be clearly interpreted with regard to a benefit on primary negative symptoms. Of course, if the new agent returns patients to an entirely normal state with no negative symptoms, there would be no problem in interpretation. Such an outcome seems unlikely. Finally, we would like to comment on the issue of possible overlap in the constructs “negative symptoms” and “cognitive impairment” in schizophrenia. These are both primarily residual phase phenomena, or at least the clinical impact of these phenomena is greatest during the residual phase. This is the phase of the illness when patients would like to return to more productive and fulfilling lives, but the residual negative symptoms and cognitive impairment frustrate such a return to function. Although Harvey et al3 have argued that these may be separable domains of this illness, they, nevertheless, acknowledge that more data are needed to fully address this question. From FDA's standpoint, the best course of action at this point is to continue to collect data that can help address this question. Thus, we feel that all future trials, whether targeting primary negative symptoms or cognitive impairment, should collect data on both domains. If it turns out that drugs having a benefit in one domain always have a similar benefit in the other domain, it becomes less attractive from a regulatory standpoint to consider them separate constructs. Perhaps, an alternative terminology could be “residual phase schizophrenia” that would encompass both negative symptoms and cognitive impairment and would, of course, acknowledge changes in both. At this time, this question remains to be answered.