**3. Measurement of prolactin and definition of hyperprolactinaemia**

Units of measurement have the potential to cause some confusion as US and EU data are often presented in ng/ml whereas most UK data are in mIU/L. Conversion rates from ng/ml to mIU/L are not standardised and vary between 21.2 and 36 dependent on the assay employed (Bushe et al, 2008). Furthermore, clinical reports do not always report either the normal range utilised or sometimes the units of measurement (McEvoy et al, 2007).

Definitions of hyperprolactinaemia vary depending upon the upper limit of normal (ULN) for the local assay. Normal ranges for females tend often to be around 30% higher than males, with some laboratories also reporting separate ranges for premenopausal and postmenopausal females. In the psychiatric literature, some of the highest ULNs for females are around 700 mIU/L and, for males, 500 mIU/L (Bushe and Shaw 2007), with lowest ULN at 300 mIU/L for females (Meaney et al, 2004). The Maudsley guidelines 10th edition (Taylor et al, 2009) gives fairly specific advice on blood sampling (1 hour after waking or eating) and cites normal ranges in both ng/ml and mIU/L. In their view, the ULN for females is <530 mIU/L and for males is <424 mIU/L; re-testing is advised if the prolactin level is between 530-2120 mIU/L.

There is currently also no specific definition for an elevated prolactin level that may be regarded as clinically non-significant and when we published our original data set there was no specific guidance to either diagnose or grade level of severity of hyperprolactinaemia (Bushe/Shaw 2007). Thus, we created three specific grades of hyperprolactinaemia: slightly elevated (<1000 mIU/L), significant elevation (1001-2000 mIU/L) and severe elevation (>2000mIU/L). This was based on empirical judgement and not with relation to specifically defined outcomes. In general terms, prolactin levels <2000 mIU/L may be due to a medication effect but other causes can include microprolactinoma, pituitary stalk compression, renal failure or hypothyroidism (Holt, 2008). The literature currently reports that macroprolactinomas are the most common cause of prolactin levels >2120 mIU/L in the general population (Bushe et al, 2010) although other authors propose higher levels (3180 mIU/L) at which hyperprolactinaemia can be assumed to be caused by a macroprolactinoma (Holt, 2008).

When evaluating hyperprolactinaemia it is also critical to understand the incidence or prevalence of hyperprolactinaemia from the patient perspective as opposed to a mean level from a cohort. Recent data are now tending to more commonly include both variables (Mackin et al, 2011) whereas in our 2008 review of this topic we reported that though 60% of studies reporting prolactin data included some degree of categorical analysis, this was seen mainly in the naturalistic studies (88%) rather than the randomised controlled trials (42%) (Bushe et al, 2008).

### **4. Consequences of hyperprolactinaemia**

Many of the longer term definitive outcomes associated with elevated prolactin remain unknown. Recent findings of prolactin receptors in atherosclerotic plaques in coronary arteries of healthy subjects indicate a possible role of prolactin even in coronary artery disease (Reuwer et al, 2009). There are, however, three areas of pathology that would seem to be closely linked to elevated prolactin, sexual function, bone loss and cancer and these can be considered as short- and longer term potential adverse events.

Prolactin and Schizophrenia, an Evolving Relationship 437

the finding that deterioration can be measured over a single year and essentially cannot be prevented (Meaney and O'Keane, 2007; O'Keane, 2008). A second set of key epidemiological studies evaluating fractures in large UK cohorts was published suggesting that hip and other bone fractures are a sequelae of mental illness and its treatment. Howard reported that hyperprolactinaemia and prolactin-elevating antipsychotics have been associated with a doubling of the risk of hip fracture in schizophrenia patients in a large UK study (OR 2.6, CI 2.43-2.78) (Howard et al, 2007). A second study also using the UK General Practice Research Database (GPRD) reported that in women the highest relative risk of fracture in a mentally ill population were in the youngest cohorts, whereas in males the greatest risks were seen in older age (Abel et al, 2008). The results showed that the relative risk (RR) of any fracture was increased more than double in females with psychotic disorders (RR 2.5: CI 1.5-4.3) but that even greater risk was measured in the cohort aged 45-74 years with psychotic disorders, with a relative risk in women of RR 5.1 (CI 2.7-9.6) and in males RR 6.4 (CI 2.6-16.1) when looking specifically at hip fractures (Abel et al, 2008). This risk may be seen to an even greater extent in males than females (Howard et al, 2007) and is present after adjusting for the other risk factors for osteoporosis highly prevalent in a cohort of patients with severe mental illness (poor diet, low exercise rates, increased alcohol consumption and decreased sunlight exposure). Other data however are needed for other fracture sites ( radius and vertebrae) together with some indication as to whether it is the cumulative length of hyperprolactinaemia that is crucial (a sort of area under the curve measurement) or the effect of a critical peak level of prolactin. Recent data in non-schizophrenic males with prolactinoma reported that using DEXA scanning of the lumbar spine vertebral fractures were diagnosed in 37.5% of patients compared with 7.8% of controls (p<0.001) (Mazziotti et

al, 2011) and that these developed independently of hypogonadism.

A recent systematic review concluded that breast cancer is significantly increased in females with schizophrenia but the data have simply not been published to establish the degree of the putative role of prolactin in this increased risk (Bushe et al, 2009). A number of epidemiological studies have reported data over the last 25 years but it is only in the last few years that clarity has emerged. The importance of systematic review in addressing a clinical question is clear. In this case, when studies with adequate powering and follow up undertaken in an age group where cancer developed (>50 yrs for breast cancer predominantly) are considered, the results were clear. The specific relevance of breast cancer is that it is the most common cancer in women in the UK, it accounts for 23% of all female cancer cases worldwide, there is a lifetime risk of 1 in 9 in the general population and this risk is increasing (Bushe et al, 2010). A recent meta-analysis that included fewer studies than our systematic review (Catts et al, 2008) reported a 12% increased risk (Standardised Incidence Ratio [SIR] 1.12, 95% CI 1.02-1.23) with a more recent UK study reporting an increased risk of 52% in schizophrenia adjusting for recognised confounders such as poverty (Hippisley-Cox et al, 2007). One can only speculate over the role of prolactin and mammary carcinogenesis, however in animal toxicity and molecular studies, it has been recognised over many years (Harvey 2008) that there is a very strong association. The US Nurses' Health Study evaluated prolactin samples from 32,826 patients with normal prolactin levels during the period 1989 to 1990 and these subjects have been extensively followed over 20 years, providing conclusive evidence linking prolactin and breast cancer in the general

**4.2.2 Possible association with cancer** 
