Usefulness of serum mass spectrometry to identify women diagnosed with higher grades of cervical intraepithelial neoplasia may differ by race

Figures & data

Figure 1 Association between the intensity of peak m/z = 4459 and odds ratio for CIN 2+^a.

Note: ^aReference value for odds ratio shown is 0.58 (logistic model predicted odds for CIN 2+ at median intensity of the peak).
Abbreviation: CIN, cervical intraepithelial neoplasia.

Figure 2 Association between the intensity of peak m/z = 4154 and odds ratio for CIN 2+ for Caucasian Americans^a and African Americans.^b

Notes: ^aReference value for odds ratio shown is 0.75 for Caucasian Americans (logistic model predicted odds for CIN 2+ among Caucasian Americans at median intensity of the peak); ^bReference value for odds ratio shown is 0.56 for African Americans (logistic model predicted odds for CIN 2+ among African Americans at median intensity of the peak).
Abbreviation: CIN, cervical intraepithelial neoplasia.

Figure 3 Receiver operating characteristic curve for all cases (CIN 2+) and all controls (≤CIN 1).

Abbreviation: CIN, cervical intraepithelial neoplasia.

Table 1 Cross-validated screening test measures for the final multivariate model and individual models with the two components of the final model predicting cervical intraepithelial neoplasia (CIN) 2+ vs ≤CIN 1 (n = 127)

Model: Predictor(s)	Sensitivity	Specificity	Cutoff prob.a
Final model: [m/z = 4459], Race*[m/z = 4154], Race, [m/z 4154]	79%	52%	0.32
Model with Component 1 only: [m/z = 4459]	69%	58%	0.34
Model with Component 2 only: Race*[m/z = 4154], Race, [m/z = 4154]	58%	58%	0.38

a Note: Optimal cutoff probabilities were determined by receiver operating characteristic curves.

Figure 4 Receiver operating characteristic curve for CIN 3 and controls diagnosed with <CIN 1.

Abbreviation: CIN, cervical intraepithelial neoplasia.

Table 2 Cross-validated screening test measures for the final multivariate model and individual models with the two components of the final model predicting cervical intraepithelial neoplasia (CIN) 3 vs <CIN 1 (n = 47)

Model: Predictor(s)	Sensitivity	Specificity	Cutoff prob.a
Final model: [m/z 4459], Race	70%	87%	0.60
Model with Component 1 only: [m/z 4459]	61%	75%	0.50
Model with Component 2 only: Race	74%	71%	0.72

a Note: Optimal cutoff probabilities were determined by receiver operating characteristic curves.

Table 3 Cross-validated screening test measures by race for a model predicting cervical intraepithelial neoplasia (CIN) 2+ vs ≤CIN 1, using peak [m/z = 4459] as the only predictora

Race	n	Sensitivity	Specificity
African American	58	58%	64%
Caucasian	69	76%	53%
Combined	127	69%	58%

a Note: Cutoff probability was 0.34.

Table 4 Cross-validated screening test measures by race for a model predicting cervical intraepithelial neoplasia (CIN) 2+ vs ≤CIN1, using peak [m/z = 4154] as the only predictor

Race	n	Sensitivity	Specificity	Cutoff prob.
African American	58	63%	64%	0.32
Caucasian	69	62%	53%	0.40

Table 5 Cross-validated screening test measures by race for a model predicting cervical intraepithelial neoplasia (CIN) 3 vs <CIN 1, using peak [m/z = 4459] as the only predictor (n = 47)a

Race	n	Sensitivity	Specificity
African American	23	50%	76%
Caucasian	24	65%	71%
Combined	47	61%	75%

a Note: Cutoff probability was 0.5.