MONMOUTH JUNCTION, NJ ( TheStreet) -- There are some significant red flags flying from Insmed's ( INSM) phase III study of the antibiotic Arikace in cystic fibrosis patients. Insmed claims the phase III study demonstrated the still-experimental Arikace was just as effective as currently approved Novartis' ( NVS) TOBI in treating bacterial lung infections afflicting cystic fibrosis patients. The company issued a press release declaring Arikace's success last July and promised to present actual data from the phase III study at a later date. Those Arikace data were finally published this weekend at the North American Cystic Fibrosis conference but they don't exactly support Insmed's claims -- unless you're willing to accept some dodgy analysis and ignore missing data. Insmed was unable to discuss the Arikace phase III results because the company was in a self-imposed quiet period ahead of third-quarter financial results on Nov. 5, according to a company spokesperson. Why the company chose to publicize the Arikace phase III results during its quiet period but not answer questions about the data was not explained. Let's dig into the Arikace study results. You can download a poster presentation of the data from Insmed's web site. First, take a look at this diagram pulled from the poster. It shows how patients were enrolled and randomized into the study. Of the 302 cystic fibrosis patients randomized, 148 were treated with LAI, which is the scientific name for Arikace. The other 146 patients were treated with TOBI. This comprises the modified intent-to-treat patient population (mITT) -- defined as all patients who received at least one dose of study drug. The set of boxes underneath discloses the "per protocol" patient population, which is not defined explicitly anywhere on the poster but seems to exclude patients who discontinued the study for some reason. There are 129 Arikace and 137 TOBI patients in the per protocol group, or 28 fewer patients than in the mITT group. 19 Arikace and 9 TOBI patients missing. Generally speaking, efficacy analyses conducted on ITT or mITT patient populations are more conservative, more stringent and more reliable than analyses done with per protocol patient populations. When you omit patients from a study analysis, bias can creep into results. This next chart graphs the primary endpoint of the phase III study, which compares the relative change in FEV1 from baseline to week 24 (Day 168) between Arikace and TOBI. FEV1 is a commonl measure of lung function. Insmed says the study achieved the primary endpoint because Arikace demonstrated statistical non-inferiority to TOBI. The lower bound of the 95% confidence interval was -4.95 -- greater than -5%, which was set prospectively as the non-inferiority margin necessary for the study to be successful. Look at the chart again and notice that Insmed relies on the per protocol patient population for the analysis of the study's primary endpoint. Why not use the more stringent mITT patient population? Insmed doesn't explain. The per protocol patient population excludes 28 patients. If you look closely at the chart, you'll see that the number of patients remaining in the study at Day 168 is even smaller. Another four patients are missing. What would an analysis of the FEV1 non-inferiority primary endpoint look like if the mITT patient population was analyzed? As disclosed, the per protocol analysis is just barely positive. If FDA decides to analyze the study more conservatively using the mITT patient population, this study is at risk of failing.