The current study included 44 patients with breast cancer treated in the National Medical Research Center for Obstetrics, Gynecology and Perinatology named after Academician V.I. Kulakov of the Ministry of Healthcare of Russian Federation, Moscow. The exclusion criteria were neoadjuvant therapy and the presence of malignant neoplasms of other localization prior to the diagnosis of breast cancer. All experimental protocols and methods are approved by the Ethical Committee of the National Medical Research Center for Obstetrics, Gynecology and Perinatology named after Academician V.I. Kulakov of the Ministry of Healthcare of Russian Federation, Moscow. All clinical investigations are conducted according to the principles expressed in the Declaration of Helsinki. All the patients signed informed consent.
More than half of the patients (55%) had metastases to at least one lymph node. In patients with metastases to regional lymph nodes stage pT1N1M0 was diagnosed in 5 (21%) patients, stage pT2N1-3M0 was diagnosed in 18 (75%) patients, and stage pT3N3M0 was diagnosed in 1 (4%) patient. In the group of patients without metastases to regional lymph nodes, one half had the pT1N0M0 stage, and the other half had the pT2N0M0 stage.
Two samples of breast tissue were collected from each patient: a tumor site and a site of normal tissue away from the tumor. Histological verification was performed for each sample. The analysis of lipid composition of the tissue was carried out by HPLC-MS according to the previously developed protocol 9–13. The dried lipid extract was redissolved in acetonitrile / isopropanol mixture (1/1) and separated on a Dionex UltiMate 3000 chromatograph (Thermo Scientific, Germany) with detection on a Maxis Impact qTOF mass spectrometer (Bruker Daltonics, Germany) both in the positive and negative ion detection modes. To verify chemical identification, tandem MS analysis with a scanning window of 5 Da was additionally carried out.
The resulting .d files were converted into ms2 files, which contained information on the ion fragmentation spectra at each time point (those .d files that contained tandem MS data were transformed), and MzXml, which contained full-MS data at each time point of chromatographic analysis. The free software msConvert (Proteowizard, 3.0.9987) was used for file conversion. The MxXml were then processed in MzMine to isolate the ion peaks and normalize them to the total ion current. Tandem MS files were used to identify lipids by means of LipidMatch scripts. Times and masses of ions from a table generated by MzMine were correlated with the tandem MS data of corresponding ions at a given time point. To evaluate the relevane of the ion fragmentation spectrum to the lipid fragmentation spectrum, a library of characteristic fragments included in the package 24 was used. Lipid nomenclature is consistent with LipidMaps 25.
Statistical analysis was done using scripts in the R language (3.3.3) in the RStudio (1.383 GNU) environment 26,27. The clinical data of the patients and the histological characteristics of tissues related to the numerical characteristics were verified for normality using the Shapiro-Wilk test (p > 0.05). The presence of statistically significant differences for normally distributed variables was determined using the Student's t-test with an accepted critical value of p < 0.05. Values outside the normal distribution were tested by the nonparametric Mann-Whitney test for statistically significant differences with an accepted critical value of p < 0.05. To assess the differences in factorial histological characteristics of tissues in patients with and without metastasis, the Pearson chi-square criterion of agreement was used with the accepted critical value p < 0.05. The identified lipids were tested for significant differences in the level in the presence and absence of metastases separately for tumor tissues and for tissues of normal mammary gland by the nonparametric Mann-Whitney test with the accepted critical value p < 0.05.
Categorical data were described using the absolute number and percentages of the total number of patients in the group. Quantitative normally distributed data were described using the arithmetic mean value (M) and standard deviation (SD) as M ± SD. Quantitative data with a distribution other than normal were presented as the median (Me) and quartiles Q1 and Q3 as Me (Q1; Q3).
Lipids with levels that statistically significantly changed in the group were used to create a diagnostic model based on logistic regression. The optimization of logistic regression was carried out by the stepwise addition of variables and verification of the Akaike information criterion 28. Lipid levels in tissues were used as variables. The diagnosis for the presence / absence of metastases was used as response variables. To assess the quality of a potential diagnostic model based on logistic regression, N logistic regressions were constructed based on N different samples containing (N – 1) object, followed by a test on an object not participating in the construction of regression, where N is the number of all objects in a pair of clinical groups. Sensitivity and specificity were evaluated as the number of true positives / total number of patients with metastases and the number of true negative results / total number of patients without metastases, respectively. The predictive value of positive and negative results was evaluated as the number of true positives / the number of positives and true negative results / the number of negative results.