The design and conduct of the 2-year studies have been fully described, including the selection of doses, by Stammberger et al.¹⁰ A total of 50 animals per sex and per group were used. Three dose levels of insulin glargine were used (2, 5, and 12.5 IU/kg) along with saline control, vehicle control, and neutral protamine Hagedorn (NPH) insulin (12.5 in mice or 5 IU/kg in rats) groups. The expected in-life parameters (body weight, food consumption, survival, and behavior) were regularly monitored. Animals were palpated for nodules monthly until 6 months of age and then every 2 weeks to study end. Rats found dead were autopsied the same day.

Blood samples for the determination of insulin were collected from the retrobulbar venous plexus from 3 female, nonstarved rats each at 1, 2, 3, 4, 7, and 24 hours after dose 27, 188, and 370. The vehicle control group served as the control group for both the insulin glargine and NPH insulin groups. The blood was centrifuged and the serum concentration of insulin was determined by radioimmunoassay using a commercial human insulin RIA kit (RIA-gnost Insulin; Behringwerke, Marburg, Germany). For the insulin glargine-treated groups, a total of 100 µL of standard/sample was incubated with 200 µL of ¹²⁵I-insulin tracer and 200 µL of anti-insulin serum for 21 to 23 hours at room temperature. Antibody-bound and free radiolabeled ligand were separated by adding 1 mL of a 17.5% polyethylene glycol solution to each tube and vortexing until a homogenous solution was achieved. After centrifugation at ∼1500g for 15 minutes at room temperature, the supernatant was decanted and radioactivity in the precipitate counted. An insulin glargine standard curve was prepared by serial dilutions of a stock solution in human insulin-free serum. The limit of quantification (LOQ) was 0.5 ng/mL and the measuring range was 0.5 to 100 ng/mL. Samples >50 ng/mL were diluted. The serum concentration of insulin in the NPH insulin-treated and vehicle control groups was determined using the commercial human insulin RIA kit (RIA-gnost Insulin; Behringwerke) as described by the manufacturer. The LOQ was 7.5 µIU/mL. Samples >165 µIU/mL were diluted. A total of 16.8 µIU/mL corresponded to 1 ng/mL of human insulin.

For toxicokinetic analysis, the maximum concentration (C_max) of insulin glargine was obtained directly from measured data. Area under the serum insulin glargine concentration–time curve (area under the curve [AUC]_{0-24 h}) was calculated using the linear trapezoidal rule where values below the LOQ were entered as 0.25 ng/mL. The 2 pharmacokinetic parameters were summarized using descriptive statistics.

For each type of tumor, statistical analyses were performed using a modified Peto lifetime-adjusted analysis¹⁴ and the Bieler-Williams Poly-3 test,^15,16 2 of the recommended methods.¹³ In the modified Peto lifetime adjustment, time strata were defined, in weeks, as 0 to 50, 51 to 80, 81 to terminal sacrifice, and terminal sacrifice. When less than 20 findings were present, exact permutation tests were used.¹⁷ Three separate tests were performed, a 1-tailed test for increasing monotonic trend in tumor rate for the saline control group versus the treated groups, a 1-tailed test for increasing monotonic trend in tumor rate for the vehicle control group 2 versus the treated groups, and a 1-tailed test for increasing monotonic trend in tumor rate for the pooled control groups versus the treated groups. A test for differences between the control groups was also performed at the 5% level. Pairwise comparisons between the controls (separately or pooled) and the high-dose group were performed.

A Bieler-Williams Poly-3 test approach does not depend on the classification of tumors (fatal or incidental). It is a survival-adjusted quantal-response procedure that modifies the denominator in the quantal estimate of lesion incidence of the Cochran-Armitage linear trend test to approximate more closely the total number of animal-years at risk. The thresholds considered were those given in the FDA guidance¹³ for trend tests; 0.025 for rare tumors and 0.005 for common tumors and, for control-high pairwise comparisons, 0.05 for rare tumors and 0.01 for common tumors. All tumor analyses were performed using the MULTTEST and LIFETEST procedures in version 9.1 of the SAS system on Windows XP.

The current guidelines call for 2-year carcinogenicity studies in rats and mice with 50 animals per sex and per group, 3 dose levels of experimental drug plus saline and vehicle control groups and, if appropriate, a comparator group. The design and conduct of both studies meet the practice guidelines of the present time.

More than 20 female mice were still alive at week 80 in all groups but not at week 90 (Table 1). In the saline control group, 23 mice were alive at the start of week 86; in the vehicle control group, 20 were alive at the start of week 87; in the glargine low-dose group, 21 were alive at the start of week 84, while 21 animals remained at week 90 in the middle-dose group and 27 in the high-dose group. There were 21 animals alive at week 90 in the NPH group.

For female rats, more than 20 animals were still alive at week 80 in all groups and at week 90 in all but the high-dose group (Table 1). Only in the high-dose group the number of live animals was below 20 at week 90; and in this group, 21 animals were still alive at the start of week 89. Thus, comparing these survival rates with the FDA “rule of thumb” that 20 to 30 animals should still be alive between weeks 80 and 90, it is considered that there were enough animals living long enough to provide adequate exposure to the drug and to be at risk of forming late-developing tumors in both studies.

Malignant tumors were found in mammary glands from 2 female mice in the saline control group and in 2 from the glargine high-dose group. Analysis of tumor incidence revealed no statistical significant difference between the control groups and the glargine high-dose group using either the Peto analysis at the 5% level (Table 2) or the Bieler-Williams Poly-3 test at the 2.5% level. In the latter analysis, P values for treated versus saline control, vehicle control, or dual control groups were 0.5934, 0.0334, and 0.3738, respectively. Similar results were obtained with rats using either the Peto analysis of individual tumor (Table 2) or combined tumor (Table 3) incidences, or using the Bieler-Williams Poly-3 test (Table 4). Raw mammary tumor data are available in Supplemental Tables 1 and 2.

The toxicokinetic parameters of insulin glargine in female rats are summarized in Table 5. Both C_max and AUC_{0-24 h} increased with repeated dosing and with increasing dosing for insulin glargine and with repeated dosing for the single dose of NPH insulin. Mean values of both C_max and AUC_{0-24 h} at 5 U/kg insulin glargine were similar to or greater than those with 5 U/kg NPH insulin and were approximately 2- to 3-fold greater at 12.5 U/kg insulin glargine compared with 5 U/kg NPH insulin.