Three separate endpoints were analysed; detection of K65R, detection of M184V and ZD1839 mw detection of either K65R or M184V. Person-time was calculated from the start date of the regimen to detection of the mutation(s) being analysed. Follow-up was censored at the earliest of the stop-date of the regimen and the last visit date. The number of events (detection of mutation(s)) was divided by the person-time to calculate the rate of an event according to whether the regimen contained 3TC or FTC. Rates were also stratified by demographic variables (age, sex,

exposure and ethnicity), current CD4 count, most recent viral load (VL), VL at entry and year of starting regimen. Associations between these variables and the event of interest were determined using Poisson regression, allowing for multiple observations from each patient. Univariable and multivariable analyses were performed; multivariable analyses were adjusted for variables mentioned above. In sensitivity analyses, only first treatment episodes of either the 3TC or the FTC regimen were analysed. In a further

sensitivity analysis we restricted to those who had experienced virological failure (1 viral load >400 copies/ml) whilst receiving the regimen and had a resistance test available. Logistic find more regression was used to determine whether there were any significant associations between 3TC and FTC containing regimens and detection of resistance mutations. In total, 5455 patients received either (or both) 3TC, TDF and EFV or FTC, TDF and EFV through the course of follow up, contributing a total of 6465 episodes over 9962 person-years. Forty-seven of these episodes were preceded by a resistance test showing evidence of the K65R (n = 4) or M184V (n = 43) mutations and were hence excluded from the analyses. Table 1 shows the baseline (at start of regimen) characteristics of the remaining 6418 episodes, contributed by 5414 patients. The majority of episodes consisted of FTC- (n = 5190) rather than 3TC- (n = 1228) MRIP based regimens. Age, sex, ethnicity and exposure were similarly distributed between the two groups. FTC-based episodes

were associated with higher median CD4 counts at the start of regimen (297 vs. 276 cells) compared to 3TC-based episodes (p = 0.01), though the median viral load at start of the regimen was lower in the 3TC-based episodes (53 vs. 312 copies/ml) (p = 0.27). Two hundred and thirty nine of 5140 patients receiving FTC (4.6%)had resistance tests performed compared to 65/1228 patients receiving 3TC (5.3%) (p = 0.31). A higher number of patients failing on 3TC containing regimens had resistance tests performed at the time of failure with 53/277 (19.1%) patients failing on 3TC having resistance tests performed, compared to 148/1060 (14%) of patients receiving FTC (p = 0.03). Over the course of follow up 21 cases of K65R were detected, giving a K65R event rate of 0.21 (95% CI: 0.12, 0.31)/100 person-years follow up (PYFU).