In April, the updates come out for those handy little books published by Barnett Educational Services.  I like the perfect bound versions but you can get the spiral bound books too. 

The pricetag is reasonable ($11.95 or $13.95) but remember this same information is free on the www.fda.gov website.  Unfortunately, the problem with the info on the web is you have to use the computer to look up the information you want and sometimes it is just easier to leaf through the little book to the page you want to read again.  I’ve used these handbooks over and over and I keep them on my bookshelf even though I know I can find the same info on the web.  These little beauties are easy to grab and they contain just the right information to help you answer questions on the fly.

The first of these books in the “2009 CFR/ICH GCP Reference Guide” which is updated annually in April (so hang on for the 2010 version in just a few weeks).  This book contains the US FDA 21 CFR parts 11, 50, 54, 56, 312 and 314 for GCP and drug applications, part 58 for GLP, the ICH E6 and E2A guidelines for GCP and Clinical Safety Data Management along with the EU clinical trials directive and the EU GCP directive.  If you’re working in the pharmaceutical industry and did not know you need to read and understand this information, you should buy this book to provide a basic understanding of the regulations related to drug development and clinical trials.

A separate book is dedicated to Medical Devices, the “2009 Code of Federal Regulations Reference Guide for Medical Devices” which was updated November 5, 2009.  This book contains the same US FDA 21 CFR parts 11, 50, 54 and 56 information on GCP added to information on Medical Devices and Quality Systems in parts 801, 803, 806, 807, 812, 814, 820, 822 and the ICH E6 guidelines for GCP along with some information about product jurisdiction and combination products. 

This same group (Barnett) publishes two other, more expensive resources called the “Good Clinical Practice: A question & answer reference Guide, 2009” and “State-by-state clinical trial requirements reference guide 2006” (each is less than $50).  I like these two because most of us don’t have all the answers to those tough questions and I think it’s difficult to stay on top of the ever changing state laws.  The Q&A book is great for training and offers about 700 of the most common questions along with 2009 interviews with senior FDA and EMEA GCP compliance officials.  The state-by-state book summarizes the key requirements in each state including areas like age of consent, state licensing authorities, HIV testing and rules for genetic testing or cancer research.

More information at:www.barnettinternational.com/EducationalServices/Publications.aspx

Dr. Frestedt is president and CEO for Frestedt, Incorporated and has over 30 years of scientific, clinical and regulatory experience in the pharmaceutical and medical device industries. Dr. Frestedt has a Ph.D. in Pathobiology from the University of Minnesota Medical School and she currently serves as Chair of the Ethics Committee for the Regulatory Affairs Professionals Society.

Back to Top

In our fantasies, clinical trials are run with perfect adherence.  Patients never drop out, visits to clinical sites are never missed and required evaluations always take place.

In the real world, questionnaires are left unanswered, samples spoil in transit to the lab and dates of study visits are forgotten.  When these things happen, the consequence is missing data.  Once upon a time, the most practical way to address the problem was simply to try to limit the amount of missing data (it goes without saying that that's always a good idea).

But recent years have seen advances in statistical methods and a corresponding rise in interest from FDA, manuscript reviewers and the scientific community at large in missing data.  In short, better tools are now available, and we in industry are expected to use them.

Some kinds of missing data are relatively harmless.  A patient who forgets to fill out a questionnaire produces missing data, but that sort of missingness doesn't bias the results of the study.  Statisticians call this "missing completely at random"; it's as if you went through a database with your eyes closed and deleted some of the numbers, and it only causes a problem if there's a lot if it.

For another kind of missing data, suppose a study endpoint wasn't implemented until partway through the trial, so earlier patients don't have it.  Now there's a pattern to the missing data; statisticians call this "missing at random" -- not "random" meaning truly unpredictable, but meaning that the reason the data are missing has nothing to do with the data values themselves.  There are good statistical methods available here.

The worst kind of missing data is called "missing not at random" or "nonignorable."  For instance, suppose one of your trial endpoints involves exercise, like a treadmill test to evaluate cardiac function.  If your study includes patients with cardiac problems -- and it probably does -- some of them will be unable or unwilling to perform the test.

The reason for this kind of missingness is tied up in the datapoints themselves; patients who didn't take the treadmill test would probably have done poorly on it.  There are sometimes no statistically credible ways to fix this problem, and it should be avoided at all costs -- easier said than done, of course.

(Credit where it's due: the terminology above and a lot of important statistical work in missing data are due to Professors Don Rubin and Rod Little, whose 1987 book "Statistical Analysis with Missing Data" should be required reading for any statistician who works with missing data -- meaning almost all of us.)

Now that we know how data can come up missing, let's see what we can do about it.  Here are a few popular methods.

1. Do nothing -- just analyze the data you have.  This is fine when the data are missing completely at random, and in that situation it's conservative and obviously easy to implement.  In other situations it's less credible, especially in the "nonignorable" case.

2. Substitute the mean value -- fill in the missing datapoints with the average of the other patients' data.  This has fallen out of favor because it tacitly makes the assumption that the missing datapoint comes from an “average” patient (it probably doesn't); it also acts as if we know the exact value of a missing datapoint, which is not a good thing.

3. Regression-based imputation.  This is a more sophisticated method that takes into account other variables; for instance, if older subjects are more likely to miss clinic visits due to health or transportation issues, a regression model that includes age will do a better job of "guessing" the missing values than just inserting the average.  Still, like mean substitution, this method acts as if we know the exact value of the missing datapoint.

4. Multiple imputation.  One of the best choices statistically, it uses methods that introduce random error into the guess, removing one of the problems with regression-based imputation.  It can also handle pretty much any missing data except for the nonignorable kind.  Unfortunately, it can be hard to explain to potential critics and requires a bit of work (and usually specialized software).

Naturally, there are dozens of variations on these methods and dozens of other approaches I haven't even mentioned.  To find out which is right for you, contact your friendly neighborhood statistician for advice.

Scott Brown is a Principal Statistical Consultant with The Integra Group, providing statistical expertise to the medical device and biotechnology industry. Contact him at 763-951-7461.

Back to Top

Despite the continued uncertainty surrounding the economy, many Minnesota manufacturing firms indicate that employee pay raises are planned for 2010. One year ago, the average wage/salary adjustment anticipated for 2009 was only 0.5 percent.

Some firms considered and implemented pay reductions.  Survey administrator Heather Lintner, Principal at Denarius Human Resources, Inc, noted that salary reductions during 2009 were common, not just with manufacturers, but across industry groups.

Those actions are about to be reversed. In the soon to be released 2010 Manufacturing Compensation & Benefits Survey, the annual pay survey conducted by Manufacturers Alliance and MPMA, firms indicate average adjustments of 1.8% are planned. According to Kirby Sneen, Sales and Marketing Manager at Manufacturers Alliance “This reaffirms our belief that progressive manufacturers stay on top of employee compensation developments and use pay and benefits, in part, to retain their best people.” Now is an excellent time to gage your competitive position. And, the compensation survey is just the tool you need to do it.

Copies of the 2010 Manufacturing Compensation & Benefits Survey will be shipped (PDF format) on April 5, 2010. Bound copies will follow by April 19, 2010. If you haven’t ordered your copy yet, contact Manufacturers Alliance at 763-533-8239 to reserve your copy.

The Medical Device Alliance includes press releases from local medical device companies every month. Please feel free to send all PR to kirbys@mfrall.com for potential future inclusion.

Back to Top

Greatbatch gets FDA approval on medical device

Greatbatch Inc. said Wednesday it received Food and Drug Administration clearance to sell its OptiSeal device, which is used in inserting leads and catheters into veins.

The FDA approved the device on Monday, Greatbatch said. The device is also approved in Canada, the company said.

Its Greatbatch Medical division, which helped develop OptiSeal, focuses on technologies used in medical devices for the cardiac rhythm management, neuromodulation, vascular access and the orthopaedic markets.

EnteroMedics Announces Submission of IDE Application for Maestro RC System

ST. PAUL, MN -- (MARKET WIRE) -- 03/15/10 -- EnteroMedics Inc. (NASDAQ: ETRM), the developer of medical devices using neuroblocking technology to treat obesity and other gastrointestinal disorders, announced today that it has submitted an Investigational Device Exemption (IDE) application with the US Food and Drug Administration (FDA) for the Company's next-generation Maestro(R) RC System in the treatment of morbid obesity using VBLOC(R) vagal blocking therapy. The submission is the first step in the iterative IDE review process in support of a possible Premarket Approval (PMA) application.

"This IDE submission marks an important next step in advancing our Maestro RC System toward a pivotal study in obesity," said President and CEO Mark B. Knudson, Ph.D. "We look forward to working through the IDE review process with the FDA. We expect to be in a position to provide an update on our strategy after this process has reached a conclusion."

CVRx lands on top VC-backed companies list

CVRx Inc., a med-tech startup that’s raised more than $200 million in financing, is one of the hottest venture-backed startups nationwide, according to a recent ranking by VentureSource.

Brooklyn Park-based CVRx, which is developing a device to treat high blood pressure, ranked 31 on VentureSource’s Top 50 list. Companies were judged based on amount of equity raised, its executives’ and board members’ track records, and change in valuation for the year ending Nov. 30, 2009. CVRx ranked highest in the first category, with a rating of 3.

It was the only Minnesota company to make the list.

“This is a very defining year for CVRx, one of the most exciting years of our existence,” company founder Rob Kieval told VentureSource. “While many therapies in use today offer only symptomatic relief...our technology is intended to address the source of these problems and improve cardiovascular function, as well as the symptoms.” He also said a “public offering is a possibility.”

CVRx declined to disclose its valuation to VentureSource. However, a report in VentureBeat pegged it at about $424 million in July of 2008.

The Medical Device Alliance includes press releases from local medical device companies every month. Please feel free to send all PR to kirbys@mfrall.com for potential future inclusion.

Back to Top