The most common chromosomal abnormality
due to non-obstructive azoospermia (NOA)
is Klinefelter syndrome (KS) which occurs in
1-1.72 out of 500-1000 male infants (
It is important to evaluate the chances for
sperm retrieval among KS patients and compare
them to azoospermia with normal karyotype by
using a model which provides more efficient
results for researchers. When researchers have
more efficient results they can choose the best
method to handle this problem and patients may
not need to undergo invasive surgeries (
This cohort study enrolled 134 patients with KS and 537 patients without KS. Patients were referred to Royan Institute, a referral infertility clinic in Tehran, Iran between 2009 and 2011. We individually matched patients and controls according to disease duration, time of surgery, and surgeon in order to avoid confounding factors such as surgical skills and age.
The 47,XXY karyotype was confirmed by Gbanding of more than 30 peripheral blood lymphocytes. Semen analysis was performed according to World Health Organization (WHO) guidelines to evaluate sperm parameters. At least two analyses confirmed azoospermia in each patient. Blood samples were taken in the morning to measure FSH, LH, and testosterone levels.
Microsurgical testicular sperm extraction
(Micro-TESE) was performed on the patients under
general anesthesia as described by Schlegel (
The binary logistic regression model, as previously
where: F(.) is acumulative distribution function. If F is the distribution function of normal and logistic, then the above model will be alogistic and probit regression model, respectively.
Assume that a binary outcome “Yit" (t=1,
where b_i is a random effect with normal distribution in this formula. Adding this term into model 1 enables the observations to be considered independently. In model 2, which is similar to the independent model, symmetrical logistic or normal distribution functions are assumed for F.
Asymmetric links are used for regression models to enable better data fit. In this study, the dependent regression model as previously introduced by Chen et al. (
Data were analyzed by R software (version 3.2.2) using the BB package and GEE pack for asymmetric and symmetric models, respectively. The logistic regression model with the symmetric and asymmetric link was fit to the data and we compared the goodness of fit of the two models by quasi Akaike’s information criteria (QAIC) (
The mean age ± SD of patients with KS was 32.64 ± 0.64 years (range: 22-49) and the mean age of participants without the KS was 34.11 ± 0.27 years. The mean serum level for FSH was 34.5 mlU/m Lin KS patients and 22.6 mlU/mL in those without KS. Patients with KS had a mean serum testosterone level of 2.65 ng/mL; those without KS had a mean testosterone level of 4.04 ng/mL. Of 134 patients with KS, 38 patients had sperm retrieved; out of 537 participants, 119 had sperm retrieved. The sperm retrieval rate was 28.4 in KS patients and 22.2 in patients without KS (
We performed a marginal effect logistic regression analysis with serum FSH, LH, testosterone, age, and patient groups to determine an association between the probability of sperm retrieval and the covariates during micro-TESE. Adjusted association from the model showed that probability of retrieving sperm during micro-TESE did not differ between the two groups (control and KS) after adjusting for the covariates and could not be predicted by any of the variables. On the other hand, the results obtained from skewed logistic regression showed that the probability of retrieving sperm was not the same for patients with and without KS (P=0.01). QAIC for logistic regression with the symmetric link was equal to 85, whereas for an asymmetric link it was 74. The results of comparing the regression model are shown in Table 2. The proportion of sperm retrieval was almost equal between patients with and without KS.
Clinical characteristics of patients and controls subdivided according to Micro-TESE outcome
|Age||32.64a ± 0.48||30.0b ± 0.65||33.68 ± 0.6||34.11 ± 0.27||34.6 ± 0.55||33.93 ± 32|
|FSH (mIU/ml)||34.52a ± 1.4||34.69 ± 2.52||34.44 ± 1.68||22.6 ± 0.83||23.54 ± 1.65||22.22 ± 0.96|
|LH (mIU/ml)||17.89a ± 1.34||17.0 ± 1.94||18.28 ± 1.24||8.83 ± 0.41||9.22 ± 0.85||8.67 ± 0.47|
|T (ng/ml)||2.65 ± 0.22||3.4b ± 0.48||2.33 ± 0.23||4.04 ± 0.51||3.52 ± 0.29||4.2 ± 0.67|
FSH; Follicle-stimulating hormone, LH; Luteinizing hormone, T; Testosterone, a; Significant compared to the control group, and b; Significant compared with failures in the Klinefelter syndrome (KS) group. Data are presented as mean ± SD.
Results of logistic regression and skewed regression comparing sperm retrieval between KS and control groups after adjusting for LH, FSH, age and testosterone
|Parameter||Logistic regressiona (95% CI)||P value||Skewed regressiona (95% CI)||P value|
|Klinefelter||0.35 (-0.92 _ 0.21)||0.22||-0.49 (-0.79 _ -0.19)||0.01|
|LH (mIU/mL)||-0.01 (-0.04 _ 0.02)||0.51||-0.02 (-0.03 _ 0.01)||0.26|
|FSH (mIU/mL)||0.01 (-0.01 _ 0.03)||0.34||0.03 (-0.01 _ 0.07)||0.28|
|Testosterone (ng/mL)||-0.001 (-0.03 _ 0.03)||0.95||-0.002 (-0.03 _ 0.01)||0.48|
|Age||-0.01 (-0.05 _ 0.03)||0.66||-0.01 (-0.04 _ -0.02)||0. 4|
FSH; Follicle-stimulating hormone, LH; Luteinizing hormone, CI; Confidence interval, a; Significant compared to the control group, and b; Significant compared with failures in the Klinefelter syndrome (KS) group.
The present study has compared sperm retrieval in azoospermia patients with and without KS by two regression models. Sperm from KS patients is typically retrieved by conventional TESE and micro-TESE techniques. Although both techniques have successfully retrieved sperm in 44% of patients, a comparison between conventional TESE and micro-TESE showed that micro-TESE had a higher success rate (55%) compared to conventional TESE (44%) (
In some studies it was found that the sperm recovery rate in men with KS reduced with increased age, however there were no effects on serum FSH, LH, or testosterone levels (
We used logistic regression with symmetric and asymmetric links for data analysis. To the best of our knowledge, this was the first study that applied a regression model with asymmetric link in reproduction research and in this group of infertile patients. The incorrect use of asymmetric link instead of an asymmetric link could lead to a poor fit and result in biased estimates of the regression coefficients. Chen’s model that has been used in this study was more flexible than common models. Therefore, the QAIC value was lower and has shown a better fit than logistic regression with a symmetric link (
Our findings have revealed that the logistic regression model with an asymmetric link is more flexible and a better fit than the conventional regression model. Since this is the first time a skewed regression has modeled data from infertility studies, we recommend that additional studies and analyses be conducted to evaluate how well this model fits a set of observations. It is also should be notice, sperm could be retrieve of non-mosaic KS as well as without KS patients.