June 15, 2009 (Vol. 29, No. 12)

Jeffrey N. N. Gibbs

A Clinical Validation Shortcut or a Lengthy Detour?

Many marketing applications for in vitro diagnostics (IVDs) need to be supported by clinical data. This is especially likely for new analytes or new intended uses. In some instances, the only means to gather the requisite data is by conducting a prospective study. Sometimes, only freshly collected specimens can be used to generate the necessary data due to the nature of the disease or the assay. Moreover, collecting specimens prospectively gives the applicant the most control over the study design. The applicant, for example, can make sure the clinical specimens are collected in a manner that precisely matches the proposed intended use. 

Prospective studies, however, often suffer the drawback of taking too long to conduct. This is particularly true for predictive or prognostic tests for conditions which manifest themselves, progress, or recur over prolonged periods. These patients may need to be followed for years to assess the IVD’s clinical performance.

There is an obvious shortcut: use banked specimens to conduct a retrospective study.  By testing specimens that were gathered years ago, a company may be able to dramatically accelerate its clinical validation program.

Can retrospective studies ever support a 510(k) clearance or a premarket approval application (PMA)? The answer is yes.  FDA has already cleared or approved a number of applications based on testing banked samples, such as Agendia’s MammaPrint and Pathwork Diagnostics’ Tissue of Origin Test.

Of course, utilizing banked specimens is no panacea. This approach can present multiple obstacles. Before embarking on a retrospective study to support an IVDs marketing, the company must be sure it can overcome these scientific, legal, and regulatory challenges.

In general, an FDA applicant must obtain the written informed consent of subjects before they can participate in a clinical study. For retrospective studies where specimens were obtained years ago, obtaining consent to participate in a post-collection study is often impossible. 

Recognizing this problem, FDA has established a policy permitting the use of remnant samples for IVD studies. Entitled “Guidance on Informed Consent for In Vitro Diagnostic Device Studies Using Leftover Human Specimens that Are Not Individually Identifiable,” this policy allows companies to rely upon stored samples, provided that certain conditions are met. FDA adopted this policy in recognition of it sometimes being “difficult, if not impossible, to locate the donor and obtain consent.”

In order to qualify for the exemption, the study must meet multiple criteria. These include that the specimens are “remnants of specimens collected for routine care or analysis that would have been discarded”; the specimens cannot be individually identified by the sponsor or investigator (meaning that coded specimens can be used); the individuals caring for the patients do not conduct the investigation; and any clinical data accompanying the specimens do not allow the source of the specimen to be identified. This “de-identification” must be irreversible. A study where the IVD results would be reported back to the subject’s healthcare provider would be ineligible under this policy.

One other issue that must be considered is whether the new study is consistent with the informed consent provided by the patients when the sample was collected. It may be that there is no consent form documentation relating to the issue of subsequent use of tissue. The IVD sponsor should be able to use the specimens under this scenario if the banked specimens otherwise meet FDA’s policy. However, sometimes the consent form states that the specimen is to be used only for a specified purpose, e.g., a named study, or the form affirmatively excludes certain uses, e.g., genetic research.

This clear expression of intent generally controls, even if the samples are “de-identified” so that the applicant does not know the patient’s name. The patient’s explicit intent to limit the use of his or her biological materials will ordinarily prevail.

A related issue is whether the patient has any ownership claim to the specimen.  Biological materials may turn out to be valuable, and the patients who provided the specimens may feel that they deserve a share of the proceeds. This is a complex topic, which has been the subject of litigation. In general, a biological specimen collected from a patient is medical waste, and the patient has no legal claim to any intellectual property or other rights stemming from the specimens.

Specific factual circumstances, though, may allow one or more patients to allege that they are entitled to share in the IVD company’s earnings. It is prudent to verify that there is no basis by which patients can plausibly claim that they also have an ownership or other interest in the specimens being used to evaluate the performance of the IVD product, or to address this issue in the Clinical Trial Agreement with the study site.


MammaPrint identifies patients with early metastasis, providing doctors with a clear rationale to assess the benefit of chemotherapy in addition to other clinical information and pathology tests, according to Agendia.

FDA Marketing Application Issues

Using banked specimens can yield significant savings in time and money.  Nevertheless, they can also present FDA pitfalls. The crux of the problem is that the applicant has no control over how the specimens were collected. Because the proposed intended use must be consistent with the specimens’ clinical history, this can severely circumscribe the intended uses that may be claimed (See “Adroit Crafting of Intended Use Critical,” GEN, November 15, 2008). Furthermore, it may be impossible to fill in data gaps in the clinical records. In its remnant specimen policy FDA cautions that sponsors “accept the risk that they may not be able to provide sufficient information to satisfy FDA’s premarket review needs.”

The applicant also needs to be sure that the specimen-collection process did not introduce a biasing effect. The doctors must have collected the specimens while treating their patients, and not while conducting a study. This could introduce some bias, or render the data set nonrepresentative of how the test would actually be used. Are potential confounding variables accounted for, such as the heterogeneous use of treatments in an oncology patient set? 

Companies also need to be certain that if the study endpoint is clinical truth, the patient files adequately document and verify the clinical outcome. For instance, if the IVD is intended to give prognostic information regarding the likelihood of cancer recurrence, do the patient records demonstrate whether the cancer recurred? How was recurrence—or its absence—determined? FDA routinely audits study sites for premarket approvals. Will these sites holding the clinical data pass an FDA bioresearch monitoring inspection?  

FDA may be skeptical of certain kinds of claims supported by retrospective data.  There seems to be some indication that FDA is more likely to allow a retrospective prognostic claim than a predictive one relating to selection of a treatment.

The complexities of using banked specimens to support a companion diagnostic were illustrated by the December 2008 FDA panel to consider K-Ras testing before certain drugs could be prescribed. FDA speakers raised a number of concerns such as possible data dredging and the criteria that should be applied to these studies. A senior official from the Office of In Vitro Diagnostics commented, “The extent to which revision of the drug’s use and clinical validation of the diagnostic test can be based on retrospective analyses of retained specimens requires scrutiny.”

While less costly than a new prospective study, retrospective studies can represent a significant investment of time and money.  The cost of doing the study can be substantial. The cost in time and money will be even greater if FDA finds the study results unacceptable, denies the application, and the company essentially must start all over again. Therefore, before initiating a large-scale or novel study using banked specimens, the applicant should hold a pre-IDE meeting with FDA (See “Make the Most of Pre-IDE Meetings,” GEN, March 1, 2009). While this meeting cannot ensure FDA acceptance of the study data, this can help minimize the chances of conducting a study that won’t pass FDA muster.

The use of banked specimens can offer dramatic savings in time and money. These studies, however, are not risk free. IVD companies should be sure that they can gain access to an appropriate set of specimens that are backed by adequate data, present no informed consent or related legal issues, comply with applicable regulatory requirements, are consistent with the applicant’s proposed intended use, and are backed by adequate clinical records.

Jeffrey Gibbs ([email protected]) is director of Hyman, Phelps & McNamara. Web: www.hpm.com.

Previous articleCustomized Cancer Vaccines Finally (Maybe) Arrive
Next articleLinkMed to Distribute Qiagen’s Genomic HLA Testing Products in Transplantation Market