This study was conducted in accordance with regulations of the Helsinki Declaration (1964) and conforms to the Regional Ethics Committee, Gothenburg.
Two dogs were used, Hanna, a 10-year-old black Giant Schnauzer (chip no. 967000000389928), and Lotti, a 6-year-old black Giant Schnauzer (chip no. 098100311386). The owner and handler is GH. The dogs live as family pets with the owner and his family. There is free access to fresh water all day, and feeding three times a day. The dogs spend several hours each day in a garden or on walks. Health checks are made at Värmdö Animal Clinic, Gustavsberg.
The training method has been described in detail elsewhere [5, 6]. For 2 years prior to the present study, only once-a-week maintenance training was used. Each of the training sessions included 4–10 boxes , 0–3 of which contained cancer tissue or blood from patients with ovarian carcinoma. The setup was randomly selected before each session. This program allowed the dogs to be confronted with a different problem to solve in each training occasion. The dogs were rewarded only for a correct identification. Blood samples with >500 U/ml CA-125 values as an indicator of ovarian carcinoma  were used for training. One drop of the training sample was placed in a small plastic dish inside each box. The blood samples used during the training period were not used in the tests. The dogs were used in the experiment with the permission of the Regional Ethical Review Board in Gothenburg, license number: S-220-08. The dogs had free access to fresh water during training and testing hours. After 2 hours of work, the dogs were walked or had 20 minutes of free time.
Patients were selected for inclusion in this retrospective study from the clinic and bio-bank databases. The latter contains blood and cancer tissue from patients with ovarian cancer. Material was collected after obtaining permission from the individual patient and was regulated by the treatment program for ovarian cancer in West Sweden. Patient selection first was made from the clinic database, then the results were correlated with the bio-bank database to obtain blood samples.
The major selection criterion applied to all study patients was clinical complete remission (CR) before the sixth (final) scheduled chemotherapy course. Patients selected for Series I were divided into 3 groups. Group A, included patients with 3 years of relapse-free survival, Group B included patients who had relapsed within 6 months after the last treatment session; and for Group C included patients who had relapsed between 1 and 2 years after treatment. A total of 66 patients with CR in the years 2001–2007 were selected in Series 1. All were from the Gothenburg area and were also in the bio-bank database. Forty-two patients had samples in a biobank corresponding to one of the three secondary selection criteria for inclusion in Group A, B or C. Selection for Series II was made using the biobank database only.
Blood samples were collected from patients living in the Gothenburg area of West Sweden. The population is about 600,000. The treatment program for ovarian cancer in requires CA-125 analysis on two occasions. First, prior to or directly after primary surgery, and the second before the sixth course of chemotherapy treatment. However, as usual in the treatment program for ovarian carcinoma, CA-125 level was not included in the follow-up, although doctors have the option to check it. Blood samples with >500 U/ml CA-125 values were used for dog training, with one drop being placed in a small plastic dish inside each box. Blood samples used during the training period were not used in the tests.
Reference blood samples
Material for the reference group was selected from the bio-bank database. Clinicopathological variables were not considered because our previous study results showed that they did not affect the dogs’ sensitivity of detection [5, 6]. A total of 62 samples (42 for Series I and 20 for Series II), from different individuals, with CA-125 values >200 U/ml were randomly selected as reference material in Series I and II.
Test blood samples in Series I
Forty-two samples were collected and used as test material in Series I. These blood samples were taken before the sixth course of chemotherapy. Patients were divided into three groups. Group A consisted of 13 patients who had a CR, 3-year relapse-free survival, and normal (<35 U/ml) CA-125values before their last treatment course. Group B consisted of 12 patients who had a CR, relapsed within 6 months, and all except 2 (61 U/ml; <200 U/ml) had CA-125 <35UI. Group C consisted of 17 patients with a CR, recurrence between 1 and 2 years, and normal CA-125 values. Tumor histopathology, stage and grade varied within the groups.
Blood samples with >500 U/ml CA-125 values were used for training, with one drop being placed in a small plastic dish inside each box. Blood samples used during the training period were not used in the tests.
Test blood samples in Series II
Samples were taken 3 and 6 months after the sixth, final chemotherapy course. Unfortunately, we could not follow patients in Series I, Group A because none of them had blood in the blood bank. We collected blood from 10 other patients who were followed regularly. Median donor age was 65 years (range, 38–78 years).
Control blood samples
Control samples were collected during the 2 years between 2007 and 2009, mostly from female medical staff volunteers in Gothenburg. Inclusion criteria were that the patients felt healthy, were not pregnant and were free of gynecological disease. Control and test materials were not age-matched. Younger persons were consciously chosen as the source of control samples to reduce the risk of the presence of asymptomatic of ovarian cancer. Thus, both the control and test groups had samples from pre-and postmenopausal women.
Blood samples were collected in EDTA tubes, and then centrifuged at 3000 rpm for 10 min with plasma pots over the small plastic tubes. After centrifugation, the plasma was divided into two parts, one for CA-125 analysis and the other for subsequent experiments. The latter part was kept at −80°C in the tumor bank (Ethical Committee license number: S-220-08, Regional Ethical Review Board in Gothenburg). Control plasma samples were processed and stored identically to the targets. However, tubes with control blood were stored separately. Median donor age was 45 years (range, 29–65 years).
Tests were carried out in a double-blind fashion as previously described . To summarize, both test leader and handler were blinded to the location of the target samples, and were present in the test location only when the dogs were working. The dogs were tested in two series. Series I covered 4 days (2 days per occasion), while Series II covered 2 days. Ten runs were performed on each day, except for one day in Series I when 11 runs were carried out. Each run included seven boxes, placed in a circle about 2 m apart from each other. Each box [5, 6] contained a drop of plasma; five contained control materials, one contained a test sample, and one contained a reference sample. Reference materials were taken before, or shortly after the primary operation, and thus had a high concentration of odor molecules. The placement of the target and reference boxes was changed by an assistant between each run. Each box was cleaned with alcohol between runs. The tests were documented on paper and DVD by the test leader and one assistant .
A positive response was defined as indicating the target box by scratching with the foreleg, lying down and sniffing it (and not indicating the control samples). A negative response was defined as sniffing and indicating a control box and not indicating the target. An uncertain response was defined as stopping at the box, smelling it, scratching at it, and possibly barking, but going straight on and not lying down.
Treatment of ovarian carcinomas
In line with the standard treatment program in West Sweden, patients were treated by total hysterectomy, bilateral salpingo-oophorectomy, omentectomy, multiple peritoneal biopsies, and peritoneal washings with cytology. Approximately 4 weeks after primary surgery the first carboplatin, cyclophosphamide, and epirubicin combination was given. A total of six courses were administered at 4-week intervals.
The raw data were summarized as sensitivity (the conditional probability of the dog indicating cancer when the condition was present) and test specificity (the conditional probability of the dog ignoring a sample from a healthy donor). Sensitivity and specificity give insight into the general classification ability of the dogs.
The positive and negative predictive probability that the test would give the correct diagnosis were also calculated. Point estimates were calculated with 95% confidence intervals . Both sensitivity and specificity were expressed as proportions, thus standard techniques for proportions could be applied for statistical inference. Confidence intervals were based on the normal approximation,
where p is the estimated sensitivity, respective specificity and n is the number of test runs.
The test runs are best described as having a hypergeometric distribution. The probability of a perfect test run (i.e., finding the test sample and ignoring the controls) by chance was 1/6, and the probability of performing all the runs without making any errors follows the binomial distribution. Reference samples were not included in the statistical analyses.