This is the first study in a large primary population of older women with increased fracture risk that has quantified the prevalence of abnormalities during routine laboratory examination and the associations of these abnormalities with risk factors for fractures, recent fractures, low BMD and prevalent vertebral fractures. While the prevalence of low vitamin D is high, the prevalence of other laboratory abnormalities is low. Although we have observed an association between all laboratory abnormalities and the presence of both clinical risk factors and low BMD, we have to conclude that the associations are weak. The only exception is the association with severe renal failure. In this population, recent fractures and prevalent vertebral fractures do not increase the probability of finding laboratory abnormalities. About half of the laboratory abnormalities were previously reported in the medical file. Furthermore, 82% the new laboratory abnormalities did not lead to a new diagnosis or treatment. New diagnoses were as often found in participants with or without low BMD, recent fractures and prevalent vertebral fractures.

If we weigh this limited yield against the costs, the discomfort of examination and unnecessary further examination in false positives, the conclusion might be that there is not enough ground for a routine laboratory panel in fracture risk assessment or screening in primary care. The only exception might be the assessment of vitamin D and renal function. An alternative solution to the assessment of vitamin D would be to advise vitamin D supplementation in all women at risk for fractures since deficiency is very frequent. This is currently the policy in the Netherlands [18]. Another alternative could be to select those with risk factors for vitamin D deficiency for supplementation or measurement [19]. Since severe chronic kidney failure is a contra indication for bisphosphonates, we suggest that the kidney function is routinely verified or assessed before start of treatment.

Leaving out most of the routine laboratory examinations will sound contra-intuitive to many physicians. When you have a patient with osteoporosis you do not want to miss that one patient with hyperparathyroidism or a malignancy. On the other hand, we have to conclude that low BMD, recent fractures and prevalent vertebral fractures turn out to be no strong signals for these conditions in a primary care population. Primary care seems to have particularly primary osteoporosis. Performing additional laboratory examinations should therefore be performed on clinical indication only.

Compared to other studies the number of laboratory abnormalities is low in our study. There are several explanations. In the first place, some studies use a more extensive laboratory panel. Secondly, the threshold in our study for abnormal renal function is lower than in some other studies (some studies include renal failure stage 3, 30–45 ml/min). And in the third place, our study population differs from other studies. The population in studies in a fracture liaison service or secondary care might be a selection of less healthy persons.

Another difference with the previous studies is that of all laboratory abnormalities the proportion of new abnormalities is lower in our population. We think that this is caused by not only collecting information from the participants but also including information from the GP medical file in our study.

Five studies looked at the association between osteoporosis and laboratory abnormalities or newly found secondary osteoporosis, three in a fracture liaison, one in a population of healthy older women and one in an unselected cohort of older men [20,21,22,23,24]. Only one study in the fracture liaison found an association between osteoporosis and newly found secondary causes of osteoporosis. The weak associations we have found in our study population is in concordance with the inconsistence between the studies.

The primary strength of this study is the strictly primary care setting, since information on the prevalence of laboratory abnormalities in women at risk for fractures in primary care is not available. A second strength of the study is the sample size. This opened the opportunity to identify not only associations between risk factors and the complete laboratory panel, but also with the single measurements. The third strength of this study is the follow-up by checking the medical files at least 1 year later. The benefit of the approach is that we could estimate the yield of the routine laboratory examination and get insight in what are true findings and false positives. We think that this approach is important to support solid recommendations in guidelines. The fourth strength in this study is that not only the association with low BMD was taken into account but also with prevalent vertebral fractures and recent fractures. These are important pillars of fracture risk assessment besides the diagnosis low BMD based on bone densitometry alone.

A limitation in this study is that only about 50% of the people invited to the SOS participated in the trial. Furthermore, all women with recent bone densitometry or earlier use of osteoporosis medication were excluded; this comprised 6.6% of the population. This could have led to selection bias. Indeed, we saw a lower participation in higher age in the SOS, indicating a possible selection of the healthier individuals. In a previous pilot study of SOS women of 50 years or older, which was performed during a clinical improvement program without requesting extensive informed consent of the participants, the risk of selection was smaller with a response rate of 83.4% [14]. In the women of 65 years and older (n = 324) with risk factors for fractures, the prevalence of vitamin D deficiency was 44.9% and other laboratory abnormalities was 4.1% (unpublished). Since these percentages are similar to our study, this indicates that the effect of selection bias is probably limited.

In our study, we have focused on laboratory examinations that are easily accessible in primary care and which cover most prevalent and relevant secondary causes. Another limitation of this study is, therefore, that we did not measure PTH, resulting in the inability to identify secondary hyperparathyroidism. However, since we have measured vitamin 25(OH)D and creatinine clearance, the main causes of secondary hyperparathyroidism are covered. With our panel, we have probably also missed some rare causes of secondary osteoporosis, but that is unlikely to have much effect on the results, since the prevalence of the more common causes of secondary osteoporosis are already very low. Furthermore, there are two restrictions regarding generalizability. As this study included only women in primary care, the results cannot be extrapolated to men nor to a fracture liaison or secondary care setting.