A combined biomarker panel shows improved sensitivity for the early detection of ovarian cancer allowing the identification of the most aggressive type II tumours

Abstract

Background: There is an urgent need for biomarkers for the early detection of ovarian cancer (OC). The purpose of this study was to assess whether changes in serum levels of lecithin-cholesterol acyltransferase (LCAT), sex hormone-binding globulin (SHBG), glucose-regulated protein, 78 kDa (GRP78), calprotectin and insulin-like growth factor-binding protein 2 (IGFBP2) are observed before clinical presentation and to assess the performance of these markers alone and in combination with CA125 for early detection.

Methods: This nested case-control study used samples from the United Kingdom Collaborative Trial of Ovarian Cancer Screening trial. The sample set consisted of 482 serum samples from 49 OC subjects and 31 controls, with serial samples spanning up to 7 years pre-diagnosis. The set was divided into the following: (I) a discovery set, which included all women with only two samples from each woman, the first at<14 months and the second at >32 months to diagnosis; and (ii) a corroboration set, which included all the serial samples from the same women spanning the 7-year period. Lecithin-cholesterol acyltransferase, SHBG, GRP78, calprotectin and IGFBP2 were measured using ELISA. The performance of the markers to detect cancers pre-diagnosis was assessed.

Results: A combined threshold model IGFBP2 >78.5 ng ml^-1 : LCAT <8.831 μg ml^-1 : CA125 >35 U ml^-1 outperformed CA125 alone for the earlier detection of OC. The threshold model was able to identify the most aggressive Type II cancers. In addition, it increased the lead time by 5-6 months and identified 26% of Type I subjects and 13% of Type II subjects that were not identified by CA125 alone.

Conclusions: Combined biomarker panels (IGFBP2, LCAT and CA125) outperformed CA125 up to 3 years pre-diagnosis, identifying cancers missed by CA125, providing increased diagnostic lead times for Type I and Type II OC. The model identified more aggressive Type II cancers, with women crossing the threshold dying earlier, indicating that these markers can improve on the sensitivity of CA125 alone for the early detection of OC.

Figure 1

Box plot showing putative biomarker expression in the discovery sample set at >32 months tDx and <14 months tDx. The whisker limits represent the 5th and 95th percentiles; the box limits represent interquartile range; the closest point in the notches (><) of the box plot represents the median and the span from the bottom to the top of the notch is 95% confidence interval (for significant values between cases and controls it can be seen that these do not overlap). Significant P-values are indicated on the plot. (For this initial triage the value for IGFBP2 is shown as it is close to the cutoff value).

Figure 2

Comparison of putative biomarker expression in pre-diagnosis sets with no division by tDx. The whisker limits represent the 5th and 95th percentiles; the box limits represent interquartile range; the closest point in the notches (><) of the box plot represents the median and the span from the bottom to the top of the notch is 95% confidence interval (for significant values between cases and controls it can be seen that these do not overlap). Significant P-values are indicated on the plot.

Figure 3

Comparison of putative biomarker expression in pre-diagnosis sets, divided into yearly intervals. The whisker limits represent the 5th and 95th percentiles; the box limits represent interquartile range; the closest point in the notches (><) of the box plot represents the median and the span from the bottom to the top of the notch is 95% confidence interval (for significant values between cases and controls it can be seen that these do not overlap). Significant P-values are indicated on the plot.

Figure 4

Graph showing the increase in sensitivity, over time, of the combined threshold models versus CA125 alone for Type I OC, Type II OC and Pan OC. Significant P-values are indicated on the plot.

Figure 5

Survival curves. (A) for the combined threshold model (IGFBP2 : LCAT : CA125), (B) for CA125, using time to death post diagnosis. The dotted black line represents Type II patients who did not breach the threshold. The solid black line represents Type II patients with samples that breached the threshold.