The overall study results showed that after one year follow-up in the "Oslo Adiposity Intervention Study" there were a statistically significant reduction in BMI z-score and Δiso-BMI30 in the total group. We found that even a stable/modest reduction in BMI z-score (≥ 0.00-< 0.10) was associated with improvement in several cardiovascular risk factors. An increase in BMI z-score was associated with worsening of some of the risk factors studied.
The participants in our study had a mean reduction in BMI z-score of 0.13. In comparison Oude et al reported reductions in BMI z-score after one year of lifestyle interventions between 0.17 to 0.24 in children younger than 12 years and from 0.08 to 0.21 in children older than 12 years . The group with the greatest reduction in BMI z-score in our study tended to be the youngest, but there were no significant difference between the four groups. Other studies have shown that age might be an important factor for success, with younger children achieving larger reduction in BMI z-score than older ones [43–45]. The reduction in our study is of the same magnitude as two Swedish studies [37, 46]. Reinehr et al found a mean reduction in BMI-SDS of 0.36 after one year among children and adolescents attending their obesity intervention (Obeldick) . This reduction is larger than in our study, but only children and adolescents motivated for lifestyle intervention were included in their study  in contrast to the present study, where motivation was not assessed.
The group in our study with the greatest reduction in BMI z-score was the group with the lowest baseline BMI z-score. This is in agreement with other investigators . The group in our study with the lowest BMI z-score initially also tended to have the lowest HOMA-IR and insulin values at the beginning of the intervention even though the difference was not statistically significant. Sabin et al  saw a trend towards greater improvement in BMI SDS over one year in those with initially lower HOMA-IR, although the differences between the groups were not statistically significant.
We found that a very small reduction in BMI z-score (≥ 0.00-< 0.10) improved insulin and insulin resistance. This is an important finding since insulin resistance may underlie future risk of diabetes and cardiovascular disease . In comparison, Reinehr et al found a significant improvement in insulin sensitivity expressed as ISI-HOMA with reduction in SDS-BMI ≥0.5 after one year follow-up . The authors suggested that insulin sensitivity may improve significantly at a lower level of weight reduction in larger cohorts, compared to their small cohort (n = 57) . They found the same results in another study using HOMA as measurement for insulin resistance . Ford et al  found an improvement in HOMA-IR with a reduction in BMI z-score ≥0.25 among 88 obese adolescents, but greater benefits occurred with a reduction ≥0.5. Unlike Reinehr et al we did not find an increase in insulin resistance in the group with an increase in BMI z-score [25, 26], though C-peptide concentrations increased, indicating increased insulin production and future risk of diabetes. None of the groups showed significant improvements in glucose concentrations. This is in agreement with other studies [14, 25]. Reinehr et al found that glucose concentration did not improve in the group with a BMI z-score reduction as large as ≥0.5 . No change in glucose and a simultaneous lowering of insulin indicates that the insulin resistance is improved, and less insulin is needed to maintain the same glucose concentration.
We found a small but significant improvement in total cholesterol, total cholesterol/HDL cholesterol ratio and LDL cholesterol concentrations in the total intervention group. Also other studies evaluating intervention programs for childhood obesity have found improvements in lipid profile [14, 22, 26, 27]. In a study investigating changes in the atherogenic risk factor profile according to degree of weight loss, Reinehr et al found that reduction in SDS-BMI between 0.25 and 0.5 was associated with a significant improvement in LDL cholesterol, but not in triglycerides and HDL cholesterol . When the reduction in SDS-BMI was ≥0.5 they found significant improvements in all lipid fractions studied . Another study found improvements in LDL cholesterol, total cholesterol/HDL cholesterol and triglycerides with reduction in SDS-BMI ≥0.25 . None of the groups in our study showed significant improvements in HDL cholesterol and triglycerides, but total cholesterol, LDL cholesterol and total cholesterol/HDL cholesterol improved even in the group with modestly improved or stable BMI z-score (≥ 0.00-< 0.10). If maintained, these improvements may impact the atherosclerotic process which begins in childhood and is causally linked to blood cholesterol levels . The group with increase in BMI z-score also experienced an increase in total cholesterol/HDL cholesterol ratio.
When comparing changes in metabolic parameters between the three groups with stable or reduced BMI z-score and the group with increased BMI z-score adjustment was made for improvement in aerobic fitness where the data was available, since cardiovascular fitness and CVD risk factors might be independently related in obese children . A previously published study found that low cardiovascular fitness was strongly associated with the clustering of CVD risk factors in children independent of country, age and sex . Adjusting for fitness in our study population had little influence on changes in HOMA-IR, insulin and total cholesterol. Though total/HDL cholesterol and LDL cholesterol effects were somewhat explained by changes in fitness this may be due to the smaller sample size. Only 102 subjects (44%) had data on aerobic fitness, and when repeating the analyzes in the 102 subject only adjusting for gender, baseline BMI z-score and waist circumference we got the same results as when aerobic fitness were included in the model. Former research has demonstrated divergent results regarding the relationship between insulin resistance and physical fitness in obese children and adolescents, and the effect of fitness on insulin may be mediated through a direct pathway and indirectly through changes in body composition [49, 51]. A newly published study among 6th grade youths concluded that both fatness and fitness are associated with cardiometabolic risk factors, but that fatness has a stronger association than fitness . This was also the conclusion in a review from Eisenmann where he states that the association between fatness and metabolic syndrome is stronger than for fitness in children and adolescents .
Earlier published studies evaluating different treatment approaches for childhood obesity have focused on either hospital based or primary care based treatment. To our best knowledge this is the first study combining hospital and public health nurse approaches. Advantages with involving the public health nurse in the treatment are the opportunity for families to get follow-up in their local environment, and that the public health nurses have good knowledge about preventive work among children and adolescents. A limitation of our study is that we do not have data regarding the frequency of contact between the public health nurse and each subject and their families. One of the strengths of our study is the lack of strict selection criteria. All children and adolescents who were referred and met our inclusion criteria were included in the project regardless of motivational level. Also, we had a good completion rate as > 80% of participants completed the one year follow-up. The participants who did not complete the one year follow-up were older than the completers, but did not differ from completers in regard to BMI z-score or ethnicity. Summerbell et al reported that drop out rates vary from 7-43% in studies of 12 months duration . A limitation of the study is that we do not have a control group of overweight and obese children receiving no treatment. Other studies have, however, shown that overweight and obese children receiving no lifestyle intervention are more likely to increase their overweight [13, 27]. Another limitation is that the follow-up in our study was relatively short and maintenance of the treatment effect was not studied beyond one year. Furthermore, we do not have data of good quality regarding dietary habits and physical activity. We measured change in aerobic fitness as ml·kg-1·min-1. One may argue that the improvements could be due to change in mass (kg) and not "real" aerobic fitness. To avoid this it would have been useful to express aerobic fitness also as mean change in treadmill time to exhaustion or expressed as ml pr unit fat free mass. Unfortunately we did not collect data on time to exhaustion or fat free mass.