Prior event rate ratio adjustment produced estimates consistent with randomised trial: a diabetes case study