In this large prospective and outcome-wide investigation of associations between circulating IGF-I and a range of non-cancer outcomes, we found that, after accounting for multiple testing and regression dilution bias, higher IGF-I concentration was positively associated with risk of carpal tunnel syndrome and inversely associated with incident varicose veins, cataracts, diabetes, and iron deficiency anaemia. The associations for carpal tunnel syndrome, varicose veins, and iron deficiency anaemia did not vary by follow-up time at diagnosis. For cataracts and diabetes though, the associations were closer to null in those diagnosed after five or more years of follow-up, suggesting that these associations might have been affected by reverse causality (whereby IGF-I levels change as a result of early pathophysiological processes).
To our knowledge, this is the first study of IGF-I and risk of carpal tunnel syndrome in a general, population-based cohort. In line with our results, one case-control study found that 34 adults with acromegaly had larger peripheral nerves, and that biochemical control of IGF-I concentrations over a one-year follow-up resulted in reduced nerve size [34]. The positive association between circulating IGF-I and carpal tunnel syndrome could be plausible due to IGF-I’s involvement in nerve growth and formation [35]; in adults with acromegaly, carpal tunnel syndrome has been attributed to median nerve enlargement, which was correlated with circulating IGF-I [36, 37].
We found a small inverse association between IGF-I and risk of varicose veins. This is a novel epidemiological finding, although an in-vitro study found that IGF-I was associated with the proliferation of smooth muscle cells of saphenous veins [38]. IGF-I’s role in promoting the growth and survival of smooth muscle cells would suggest that the inverse association we observed is plausible.
We also found an inverse association between IGF-I concentration and iron deficiency anaemia. We are not aware of any published prospective evidence, but previous cross-sectional studies have found that low IGF-I concentrations were associated with lower haemoglobin concentration and higher prevalence of anaemia in middle-aged and elderly adults [39–42]. IGF-I might play an important role in erythropoiesis [43], suggesting that the observed inverse association is plausible. However, iron deficiency anaemia might have a longer lag time from disease onset or initial diagnosis to hospital admission [44], therefore reverse causality cannot be ruled out and longer follow-up is needed.
We found a novel inverse association between IGF-I and risk of cataracts. However, in analyses stratified by follow-up time at diagnosis this appeared to be due to reverse causality, possibly related to a shared pathophysiology with insulin resistance. In support of our findings, in adults with acromegaly, visual disturbances appear to relate to the effects of space-occupying lesions of pituitary adenomas rather than to circulating IGF-I levels [45]. Also, some evidence from in-vitro rat models has shown that IGF-I might decrease the amount of α-crystallin (a lens protein) made in the lens fibre cells[46], therefore this might be expected to lead to a positive association of IGF-I with risk of cataracts since higher levels of α-crystallin have been associated with a lower risk of cataract formation [46].
Recent prospective and genetic evidence has suggested that there could be a positive association between circulating IGF-I concentration and type 2 diabetes risk [9, 8, 16], possibly due to its involvement in glucose homeostasis [47]. This is in contrast to the inverse association we observed for incident diabetes (which is likely to be mostly type 2 diabetes due to the older age structure of this cohort). However, our findings may be due to reverse causality. Diabetes may go undiagnosed for years [48], therefore a long follow-up period is needed to avoid picking-up prevalent cases or pre-clinical disease.
We also found evidence for positive associations between IGF-I and IHD, haemorrhoids, colon polyps, osteoarthritis, kidney stones, and uterine fibroids, and an inverse association with pneumonia in participants diagnosed after five or more years of follow-up. The findings for IHD [9], osteoarthritis [12], colon polyps [15, 14], and uterine fibroids[49] are in line with some of the available prospective or genetic evidence from population-based studies, and the results for kidney stones are supported by studies in adults with acromegaly [5]. It is possible that some of these associations were masked by reverse causality in the first five years of follow-up, and follow-up with larger numbers is needed to clarify whether IGF-I might associate with these outcomes.
This is the first study to adopt an outcome-wide approach to the investigation of IGF-I and risk of 25 common conditions (other than cancer); this comprehensive approach allowed us to assess and compare the effect sizes of multiple outcomes within the UK Biobank and reduce outcome-selection bias. Additional strengths of our study include its population-based design, the use of national record linkage to ascertain information on disease incidence, which eliminates misclassification and reduces attrition bias at follow-up, and its large size; this is the largest prospective study of IGF-I and most of these 25 common conditions.
Nevertheless, our study is not without its limitations. Some measurement error could have occurred when measuring IGF-I at baseline, but we reduced the potential impact of regression dilution bias by correcting baseline measures in all participants with a repeat IGF-I measure from a subsample of participants. Furthermore, we cannot rule-out that multiple testing could have led to some chance findings, though we addressed this using Bonferroni correction. However, this is a strict approach and it is possible that some of the associations with a P < 0.05 do not reflect a chance finding. Additionally, because we used an outcome-wide approach, it is possible that we did not fully adjust for confounders that might have affected some of the individual conditions. However, many of these associations are exploratory and therefore not all of the confounders are known. Moreover, some conditions might go undiagnosed for some time and only require hospital care at later stages, and therefore might reflect prevalent or preclinical disease and/or more severe disease (such as in the case of diabetes). Furthermore, IGF-I–related proteins such as IGF-II and IGF-binding proteins (IGFBP), which play a role in the regulation of IGF-I bioavailability and signalling [50], were not measured in this study. Therefore, the observed associations could partially reflect other aspects of the IGF signalling pathway. Finally, the UK Biobank is predominantly made up of white Europeans, so the generalisability of our findings might be limited.