The high level of heterogeneity precluded a
The high level of heterogeneity precluded a meta-analysis and means it PFK-158 is difficult to draw any firm conclusions about the diagnostic potential of these biomarkers for detecting recurrence. We therefore summarise the findings of the eligible studies narratively. Numerical results regarding diagnostic accuracy estimates are available in Table 1 and Appendix 2. Three studies reported solely on patients for whom the primary tumour was seminoma [23,25,26]. One found low sensitivity (50%) and moderate specificity (73%) for LDH in the surveillance of a group of 55 patients with Stage I seminoma . Those with elevated AFP were excluded from the cohort. HCG and serum placental alkaline phosphatase (SLAP) were measured pre-operatively but not reported post-operatively. The authors concluded that LDH was not an adequate marker of recurrence in this patient group. Another study reported HCG for a group of 151 patients with Stage I–IV seminoma and found comparatively low sensitivity (57%) but high specificity (99%), concluding that HCG may have a useful role in monitoring irrespective of the level of this biomarker pre-operatively . Unusually, but in common with De Bruijn et al. , this paper also presented longitudinal trends in the monitored HCG level for a subset of patients in the post-operative period, but used a fixed 10 u/l cut-off to define biomarker positivity. HCG was elevated in 7 of the 11 recurrences. Marker-positive recurrences occurred in patients with stage I–IV disease and in patients both with and without initial marker rise at time of diagnosis. The sample size (11 recurrences among 7 patients) may have been insufficient to incorporate the time trend into a decision strategy. The final study in this subgroup provided results on a group with Stages I-III seminoma that could not be used for comparative purposes as they were reported at a per-sample level rather than a per-patient level, with an average of approximately four follow-up biomarker measurement visits per patient . HCG and LDH were the markers most frequently used in follow-up. For HCG, they reported very high specificity (98%) and low sensitivity (48%) with 90 positive results among 2790 samples taken, 19 of which were in patients found to have disease recurrence. Results were similar for LDH with 15 recurrences among the 77 elevated results from a total of 2033 samples. Reported specificity was again very high (97%) and sensitivity low (46%). Of 611 placental alkaline phosphatase (PLAP) samples taken, there were 92 positive results but only 6 of these patients had recurrent disease. All 6 patients with disease recurrence in whom PLAP was measured had a positive result. The authors concluded Crossing-over the combination of these three tumour markers should be used in monitoring for recurrence of seminoma to increase sensitivity, but highlight the high false positive rate, particularly amongst PLAP values in smokers. Three studies reported results for a mixed group of seminoma and NSGCTs as the primary tumour [18,22,24]. The first of these is a short paper dating from 1978 which, despite showing an apparently promising level of diagnostic accuracy, suffers from several methodological limitations (Fig. 2), including lack of specification of the patient population, index test and reference test, which limit its applicability . Two more recent studies demonstrate much lower sensitivity levels, of 59% and 40%, for HCG and LDH respectively [22,24]. Specificity remains above 90% in both studies. However, at least one of these studies appears likely to suffer from incorporation bias as biomarker elevation explicitly formed part of the criteria for “active disease” , while in the other the reference standard was not clearly stated . Misclassification of the timing of recurrence therefore appears possible in both studies. Of the studies to report solely on patients with NSGCTs, two estimate 100% specificity for AFP, although these use different cut-offs for marker positivity and one combines Stage I-III tumours , while the other considers Stage I tumours only . One of these studies estimates sensitivity at 36%  and the other at 100% , although the second of these has a wide confidence interval owing to its small sample size (1 recurrence in 5 patients) and suffers from methodological concerns relating to the specification of the reference standard. Sensitivity is increased to 86% in De Bruijn et al. , with little adverse effect on specificity, by considering a combined biomarker of AFP and HCG. However, when a similar combination was considered in a purely Stage I group, the sensitivity was much lower (36%) .