If you google "obesity paradox" you'll see some discussion on this.
I think this meta-review is good:
Lancet. 2006 Aug 19;368(9536):666-78. Association of bodyweight with total mortality and with cardiovascular events in coronary artery disease: a systematic review of cohort studies. Romero-Corral A, Montori VM, Somers VK, Korinek J, Thomas RJ, Allison TG, Mookadam F, Lopez-Jimenez F. Source Division of Cardiovascular Diseases, Mayo Clinic College of Medicine, Mayo Foundation, Rochester, MN 55905, USA. Abstract BACKGROUND: Studies of the association between obesity, and total mortality and cardiovascular events in patients with coronary artery disease (CAD) have shown contradictory results. We undertook a systematic review to determine the extent and nature of this association. METHODS: We selected cohort studies that provided risk estimates for total mortality, with or without cardiovascular events, on the basis of bodyweight or obesity measures in patients with CAD, and with at least 6 months' follow-up. CAD was defined as history of percutaneous coronary intervention, coronary artery bypass graft, or myocardial infarction. We obtained risk estimates for five predetermined bodyweight groups: low, normal weight (reference), overweight, obese, and severely obese. FINDINGS: We found 40 studies with 250,152 patients that had a mean follow-up of 3.8 years. Patients with a low body-mass index (BMI) (ie, <20) had an increased relative risk (RR) for total mortality (RR=1.37 [95% CI 1.32-1.43), and cardiovascular mortality (1.45 [1.16-1.81]), overweight (BMI 25-29.9) had the lowest risk for total mortality (0.87 [0.81-0.94]) and cardiovascular mortality (0.88 [0.75-1.02]) compared with those for people with a normal BMI. Obese patients (BMI 30-35) had no increased risk for total mortality (0.93 [0.85-1.03]) or cardiovascular mortality (0.97 [0.82-1.15]). Patients with severe obesity (> or =35) did not have increased total mortality (1.10 [0.87-1.41]) but they had the highest risk for cardiovascular mortality (1.88 [1.05-3.34]). INTERPRETATION: The better outcomes for cardiovascular and total mortality seen in the overweight and mildly obese groups could not be explained by adjustment for confounding factors. These findings could be explained by the lack of discriminatory power of BMI to differentiate between body fat and lean mass.
Here's a very recent study showing the same in diabetics:
Atherosclerosis. 2012 Sep 21. pii: S0021-9150(12)00617-X. doi: 10.1016/j.atherosclerosis.2012.09.004. [Epub ahead of print] Obesity paradox: Differential effects on cancer and noncancer mortality in patients with type 2 diabetes mellitus. Tseng CH. Source Department of Internal Medicine, National Taiwan University College of Medicine, Taipei, Taiwan; Division of Endocrinology and Metabolism, Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan. Electronic address: ccktsh@ms6.hinet.net. Abstract OBJECTIVE: To investigate associations between body mass index (BMI) and cause-specific mortality in patients with type 2 diabetes mellitus (T2DM). METHODS: Prospective follow-up of a nationally representative cohort of 89,056 Taiwanese patients with T2DM recruited since 1995. Vital status was matched with the National Death Certificate Database until the end of 2006. Self-reported body weight and height were used to calculate BMI, which was treated either as a continuous or categorical variable (underweight, <18.5; normal, 18.5-22.9; overweight, 23.0-24.9; obesity I, 25.0-29.9; and obesity II, =30.0 kg/m(2)). Causes of death were classified as all-cause, cancer, diabetes complications (macrovascular and microvascular), and other. Cox regression was used to estimate the hazard ratios. RESULTS: A total of 26,951 patients (30.3% of the cohort) died during follow-up (cancer 5.4%, diabetes complications 17.4%, and other causes 7.5%). As a continuous variable, BMI was inversely associated with mortality from all-cause, cancer, diabetes complications, and other causes, with respective adjusted hazard ratios (95% confidence intervals) of 0.942 (0.939-0.946), 0.966 (0.958-0.975), 0.935 (0.930-0.939), and 0.942 (0.935-0.949). Compared to normal weight, underweight was significantly predictive for any of causes of death, while overweight, obesity I, and obesity II were all significantly associated with mortality in an inverse pattern. After excluding patients with a follow-up duration <2 years, most BMI categories were not significantly predictive of mortality from cancer, suggesting a potential bias of cancer-induced weight loss. CONCLUSIONS: The obesity paradox, mainly observed in noncancer mortality, exists in patients with T2DM, suggesting a survival advantage in obese diabetic patients.