Serum leptin in obesity is related to gender and body fat topography but does not predict successful weight loss

European Journal of Endocrinology (1997) 137 61 67 ISSN 0804-4643 Serum leptin in obesity is related to gender and body fat topography but does not predict successful weight loss Leo K Niskanen 1,2, Steven Haffner 3, Leila J Karhunen 1, Anu K Turpeinen 1,2, Heikki Miettinen 2,3 and Matti I J Uusitupa 1 Departments of 1 Clinical Nutrition and 2 Medicine, University of Kuopio, Kuopio, Finland and 3 Department of Medicine/Division of Clinical Epidemiology, The University of Texas, San Antonio, Texas, USA (Correspondence should be addressed to L Niskanen, Department of Medicine, University of Kuopio and Kuopio University Hospital, P O Box FIN- 70210 Kuopio, Finland) Abstract Objective: Leptin is the product of the ob gene shown to regulate body fat and appetite in mice. It is produced by human adipose tissue also, but its physiological functions in man are poorly known. Study design and methods: We studied serum leptin concentrations in ten obese men and 35 obese women (age and body mass index 42 7 years and 35:1 3:6 kg/m 2 respectively) before (baseline) and at 17 and 57 weeks during weight loss of 10:9% of the initial weight. Results: Serum leptin concentrations at baseline were 55% higher in women than in men (after adjustment for age and body fat mass, P ¼ 0:002) and remained so during the follow-up. At baseline, serum leptin correlated with fat mass (r ¼ 0:60, P < 0:001) estimated by bioelectrical impedance, and the changes in leptin concentrations from baseline to week 17 correlated with the changes in fat mass (r ¼ 0:73, P < 0:001), but baseline leptin levels were not predictive of the successful weight loss. Leptin concentrations correlated with hip circumference (r ¼ 0:49, P < 0:001 at baseline adjusted for age and sex), but the correlation with waist circumference became evident only during the weight loss (at week 57, r ¼ 0:63, P < 0:001). Conclusions: Serum leptin concentrations are higher in obese women than in obese men before and during weight loss, but the topography of fat tissue influences serum leptin concentrations. Serum leptin concentrations do not predict the response to weight reduction. European Journal of Endocrinology 137 61 67 Introduction Serum leptin is a 16 kda product of the ob gene shown to regulate food intake and reduce body fat in mice (1 4). This ob gene product is also expressed in human white adipose tissue (5 7). Curiously, however, increased expression of the ob gene in adipocytes of obese subjects has been reported (5, 6, 8). Furthermore, serum leptin assays have shown that concentrations are in fact increased in obese subjects, correlate with body weight and decline with weight loss (9, 10). These findings are suggestive of resistance to the biological actions of leptin in obese subjects. Leptin receptor (OB-R) from mice choroid plexus and its human analog has recently been cloned (11). It is likely that the central nervous system is the main target for circulating leptin as systemic administration of leptin in ob/ob mice reduced hypothalamic neuropeptide Y gene expression (12). Neuropeptide Y, in turn, may increase energy intake and reduce energy expenditure (13). Although peripheral leptin levels associate closely with body weight, heterogeneity in leptin concentration among individuals at each body mass index (BMI) has still been observed; however, the role of body fat distribution and the effects of sex are not well known. Furthermore, the physiological role of leptin in human obesity and in the regulation of body weight remains unclear. An intriguing possibility is that serum leptin level could serve as a prognostic marker for therapeutic interventions aiming at successful weight loss. These aspects are described in this study, in which serum leptin levels were analysed in a group of wellcharacterized obese subjects before and during a weight reduction programme. Subjects and material Subjects A total of 45 obese (initial inclusion criterion, BMI 28:0-43:0 kg/m 2 ) middle-aged subjects (10 men, 35 women), whose aim was to participate in a weight reduction programme, were recruited mainly from the primary or occupational health care system in Kuopio 1997 Society of the European Journal of Endocrinology

62 L K Niskanen and others EUROPEAN JOURNAL OF ENDOCRINOLOGY (1997) 137 (14). Only those whose overweight status was stable were accepted for the study. The mean S.D. (range) age of the participants was 42 7 (24 55) years and the mean BMI and waist/hip ratios were 35:1 3:6 (30:4 43:3) kg/m 2 and 0:92 0:09 (0:76 1:14) respectively. Five participants (11.1%) were current smokers. Table 1 summarizes the baseline characteristics by sex. The main criteria for exclusion from the study were body weight loss >4 kg in the 3 months before screening, previously known or newly diagnosed diabetes (by repeated measurements of fasting plasma glucose), significant thyroid, liver or kidney disease, eating disorders assessed by standard questionnaires, and markedly increased blood pressure (diastolic blood pressure >105 mmhg). Ethics The study plan was approved by the joint Ethics Committee of the University of Kuopio and Kuopio University Hospital. Methods All the measurements were made after an overnight fasting, using standardized methods. Waist circumference was measured at the level midway between the lateral lower rib margin and the iliac crest. Hip circumference was measured at the levels of the major trochanters through the pubic symphysis. Body composition was determined by a bioelectrical impedance method (RJL systems, Detroit, USA). Bioimpedance was validated as a measure of body composition from a separate cohort of obese subjects (n ¼ 56) with DEXA (Lunar DPX, Lunar Radiation Corp, Madison, WI, USA) and the methods showed acceptable agreement (r ¼ 0:973 for lean body mass, r ¼ 0:921 for fat mass) (P Rissanen, P Hämäläinen, E Vanninen, M Tenhunen-Eskelinen & M Uusitupa, unpublished observations). Table 1 Baseline clinical characteristics by sex. Values are means S.D. Variable Men Women P value Number of subjects 10 (22%) 35 (78%) Age (years) 45 9 41 6 0:16 Body weight (kg) 109:4 14:4 92:2 11:3 <0.001 Body mass index (kg/m 2 ) 36:3 4:0 34:8 3:4 0:24 Waist circumference (cm) 116 9 101 8 <0.001 Hip circumference (cm) 112 7 113 7 0:52 Waist/hip ratio 1:04 0:05 0:89 0:06 <0.001 Body fat mass (%) 31:9 3:9 42:1 3:4 <0:001 Body fat mass (kg) 35:2 7:9 39:0 7:2 0:16 Lean body mass (kg) 74:1 7:6 53:1 5:5 <0:001 Serum glucose (mmol/l) 6:3 0:8 5:7 0:9 0:07 Serum insulin (pmol/l) 142 76 85 38 0:04 Serum leptin (mg/l) 19:8 6:9 35:7 14:4 0:002 Serum insulin was analysed by an RIA technique with double antibody polyethylene glycol (CIS Bio International, B P 32, F-91192 Gif-sur-Yvette Cedex, France). Blood glucose was analysed by kinetic photometry with glucose dehydrogenase (15). Serum samples for leptin determinations were stored at ¹70 C for an average duration of 3:5 years. The leptin assay was performed in the laboratory of Professor Steven Haffner. Leptin was measured by a commercial RIA (Linco Research Inc., St Louis, MO, USA) (16). Weight reduction programme At the beginning of the study, the patient was counselled by a dietitian on a reduced energy intake. Energy intake prescribed to the patients was based upon an estimate of their initial maintenance energy needs minus 2510 kj/day and corrected for daily activities. The diet was a nutritionally balanced, hypocaloric diet designed to cause a weight loss of 0:25 0:5 kg/week, with 30% of energy intake as fat (optimally as 10% saturated, 10% monounsaturated and 10% polyunsaturated), 50% as carbohydrate, 20% as protein, and a maximum of 300 mg/day cholesterol. Alcohol consumption was limited to no more than 150 g of alcohol per week. The diet was distributed in three main meals and, if desired, a low-fat snack. The patient remained on the diet prescribed for the entire duration of the study, with adjustment for actual body weight on day 169 that is, prescribed energy intake minus 840 kj/day. Food intake was measured by four-day food records (two weekdays and two weekend days). The subjects were instructed on the completion of the food records by an academic dietitian, and were given food scales for weighing the food items. Calculations for energy and nutrient intake were made using the Nutrica Software package for nutrient intake analysis (Social Insurance Institution, Helsinki, Finland). The measurements described above were performed at baseline and at weeks 17 and 57. In addition, the patient s body weight was recorded regularly during the treatment period. Statistical methods The normal distribution of variables was checked with the Kolmogorow Smirnow test. Serum leptin concentrations in this population of obese subjects showed a normal distribution. Student s t-test or analysis of variance (ANOVA) were used for comparison of variables among the groups and continued with an analysis of covariance (ANCOVA) in order to adjust for possible confounders. Time-related changes were analysed with repeated measures analysis of variance (MANOVA). Pearson and adjusted (partial) correlation coefficients were calculated for selected variables. Two-sided P values less than 0:05 were considered

EUROPEAN JOURNAL OF ENDOCRINOLOGY (1997) 137 Serum leptin in obesity effect of weight loss 63 statistically significant. All the analyses were performed by SPSS for Windows (SPSS Inc., Chicago, USA). Results The percentage of body fat and serum leptin concentrations were greater in women, whereas the lean body mass and serum insulin were greater in men (Table 1). The difference in baseline serum leptin concentrations between the sexes remained statistically significant after adjustment for age and body fat mass, women having 55% higher values (F ¼ 10:6, P ¼ 0:002). Serum leptin concentrations remained greater in women than in men throughout the follow-up (week 17: 24:6 12:7 compared with 12:7 5:6ng/ml, F¼ 11:8, P ¼ 0:001 by ANCOVA after adjustment for age and fat mass; week 57: 23:3 14:8 compared with 13:1 6:7ng/l, F¼ 6:96, P ¼ 0:012; adjusted means are shown in Fig. 1). As leptin levels correlated with hip circumference (see later), we made further adjustment for hip circumference, but this did not alter the results (data not shown). During the 17-week treatment period, the mean body weight decreased by 7:9% (from 96:0 13:9 to 88:4 13:9 kg), and from 17 week to week 57 it decreased by 3:3% (to 85:5 15:5 kg; P < 0:001, MANOVA). Serum leptin concentrations paralleled the changes in body fat mass (Fig. 2). Serum insulin also declined with the weight loss (from 97 36 to 76 43 pmol/l at week 57, P ¼ 0:01). The reductions in fat mass, waist and hip circumferences and lean body mass were all statistically significant (Fig. 2, P < 0:001 MANOVA), as were fasting glucose and insulin concentrations (reduced from 5:8 0:9 to 5:4 0:9 mmol/l (P < 0:001) and from 97 36 to 76 43 pmol/l (P ¼ 0:01) respectively at week 57). A linear correlation (r ¼ 0:60, P < 0:001) was observed between the body fat mass and serum leptin concentrations in the whole study population. The change in leptin concentrations correlated with the change in body weight (r ¼ 0:63, P < 0:001) and even more strongly, with the reduction in body fat mass (r ¼ 0:73, P < 0:001). Serum leptin concentrations correlated with BMI throughout the follow-up, but after adjustment for sex, age and, most noticeably, for body fat mass, the correlations disappeared (Table 2). We subdivided the subjects on the basis of the median of the weight loss (¼ ¹9:6 kg; Fig. 3). Interestingly, there was no difference in the mean baseline serum leptin concentrations (31:2 11:7and 33:2 17:5 mg/l, fat mass adjusted means 30:3 and 34:2mg/l respectively; P ¼ 0:28) between those whose weight loss was >9:6kg and in those with a smaller weight loss. At baseline, there was a negative correlation of serum leptin with lean body mass, but this relationship vanished after adjustment for age, sex and body fat mass. At baseline, serum leptin associated linearly with hip circumference but not with waist circumference (Table 2). However, during weight loss, serum leptin concentrations correlated with waist circumference and the respective correlation with hip circumference became, if anything, stronger. To illustrate this association further, we divided hip and waist circumferences at baseline and at 17 weeks into tertiles. Figures 4 and 5 show the leptin concentrations in relation to the tertiles of hip and waist circumferences at baseline and at week 17 respectively. There was a linear association (P ¼ 0:009 and 0:002 for trend by ANOVA respectively) of increasing hip circumference with increasing leptin concentrations at baseline and at 17 weeks, but no such association was observed regarding waist circumference. Discussion In this study we were able to confirm in a large number of subjects that there is a close interrelationship Figure 1 Serum leptin concentrations adjusted for body fat mass in relation to sex at baseline and weeks 17 and 57. Sex differences are statistically significant: P < 0:001 0.012.

64 L K Niskanen and others EUROPEAN JOURNAL OF ENDOCRINOLOGY (1997) 137 Figure 2 Serum leptin concentrations, body fat mass and waist and hip circumferences (mean S.E.) at baseline and during the weight reduction programme. between body fat and serum leptin concentration. Furthermore, we extended these previous findings by demonstrating that the changes in leptin levels correlated with the changes in body fat mass, even more than with the changes in body weight. Serum leptin concentrations were constantly greater in women, even after the adjustment for body fat mass. A novel finding was that serum leptin concentrations did not predict the weight loss. Interestingly, serum leptin levels were more closely associated with hip circumference than with waist circumference, even after adjustment for the effects of sex. However, during weight loss, waist circumference also showed a statistically significant correlation with serum leptin. Our study participants were recruited mostly from the primary or occupational health care system and therefore they mostly had uncomplicated obesity (14); subjects with eating disorders were also excluded. The findings of the present study may not therefore be directly applicable to other groups with more complicated obesity, or to other ethnic groups. As a whole, the weight loss programme can be considered to have

EUROPEAN JOURNAL OF ENDOCRINOLOGY (1997) 137 Serum leptin in obesity effect of weight loss 65 Table 2 Unadjusted and adjusted correlation coefficients between serum leptin concentrations and variables reflecting body composition at baseline, at week 17 and at week 57 (body mass index and lean body mass values adjusted for age, sex and body fat mass, and the remainder adjusted for age and sex). Baseline Week 17 Week 57 Variable Unadjusted Adjusted Unadjusted Adjusted Unadjusted Adjusted Body mass index 0:302* ¹0:046 0:449** 0:023 0:622*** 0:074 Lean body mass ¹0:311* ¹0:275 ¹0:238 ¹0:284 ¹0:053 ¹0:190 Waist circumference ¹0:121 0:220 0:053 0:372* 0:408** 0:632*** Hip circumference 0:450** 0:485** 0:573** 0:609** 0:693*** 0:716*** Waist/hip ratio ¹0:405** ¹0:159 ¹0:306* ¹0:103 ¹0:180 0:017 * P < 0:05, ** P < 0:01, *** P < 0:001. been rather successful, leading to well-known beneficial metabolic effects. It has been found that female mice have relatively higher leptin concentrations than male mice, despite the fat content of the body (17). In humans, the study of Considine et al. (10) indicated that serum leptin concentrations were significantly greater in normalweight and obese women than in their male counterparts. However, the sex differences in that study were not statistically significant when men and women with equivalent percentages of body fat (estimated by bioelectrical impedance) were compared. In the present study, the sex difference in leptin concentration persisted after adjustment for the weight of body fat. However, because of the limited number of men studied, these results must be interpreted with caution. One possible explanation for this apparent sex difference with respect to serum leptin concentrations could be a relative abundance of peripheral fat tissue in women (reflected as large hip circumferences) compared with men. Previously, heterogeneity in leptin concentration among individuals in each BMI category has been noted (9). An interesting and somewhat controversial finding of the present study was that serum leptin concentrations were more closely associated with hip circumference than with waist circumference, irrespective of sex. This contrasts with the recent findings from a study of Polynesian men and women (18), which found waist circumference to be an independent determinant of serum leptin concentration. Although these differences could be due to methodological problems or to unknown confounders, population-specific determinants of leptin expression cannot be ruled out. In line with this population-specific interpretation, it has recently been demonstrated that African-American postmenopausal women have 20% lower plasma leptin concentrations than white women (19). Both our study and that by Zimmet et al. (18) used the robust measurement of waist circumference, which is not able to differentiate intra-abdominal fat from subcutaneous fat. However, the results suggest that the expression of the ob gene in obese humans could be influenced by body fat distribution. This concept is in agreement with the findings of Masuzaki et al. (7), who showed that ob mrna levels in the adipose tissues varied from region to region even within the same individual, but were greater in subcutaneous adipose Figure 3 Serum leptin concentrations (mean S.E.) at baseline in obese subjects in relation to the median value of the weight loss achieved at week 57 (¹9:6 kg). No significant differences.

66 L K Niskanen and others EUROPEAN JOURNAL OF ENDOCRINOLOGY (1997) 137 Figure 4 Serum leptin concentrations (mean S.E.) in relation to the tertiles of hip circumference of the obese study population at baseline (A) (tertile limits: 110 and 116.0 cm) and at week 17 (B) (tertile limits 106 and 113 cm). P values are for trend by ANOVA. tissue than in omental, retroperitoneal and mesenteric adipose tissues. It should be noted that, at the end of the follow-up period of our study, the correlation between waist circumference and serum leptin emerged as statistically significant, suggesting a relative increase in leptin signalling from truncal subcutaneous or intraabdominal fat. Weight changes seem to affect ob gene expression differently in different types of fat tissue; in VMH (ventromedial hypothalamus)-lesioned rats, gaining weight was associated with enhanced ob mrna expression in mesenteric fat compared with subcutaneous fat (20). Furthermore, Hamilton et al. (5) showed that, within the same individual, omental adipocyte ob protein expression varied depending on the fat cell size: smaller fat cell size implied relatively lower protein expression. The mechanism(s) by which fat cell size influences serum leptin concentrations and the physiological significance of the regional differences in fat mass distribution are currently unknown, but may be modulated by genetic, environmental or hormonal factors. Although peripheral leptin concentrations are related to body fat depots, it is not known whether leptin contributes to the regulation of body weight in humans. Therefore a novel finding in this study was that the response to the weight loss programme could not be predicted from serum leptin concentrations. Thus, although it has been speculated that obese subjects with relatively low leptin levels would have difficulty in restricting their food intake (9), this did not seem to be the case. Furthermore, in a study of obese women, the serum leptin concentrations were not associated with feelings of hunger or desire to eat, thus suggesting that leptin concentrations are not involved in the regulation of eating, at least in the short term (21). However, whereas leptin concentrations in obese white subjects do not predict the response to weight reduction therapy, conversely, a recent study of Pima Indians, a population prone to obesity, suggested that low plasma levels of leptin may precede weight gain (22). The role of leptin as a regulator of body weight in man thus remains rather unclear. Figure 5 Serum leptin concentrations (mean S.E.) in relation to the tertiles of waist circumference of the obese study population at baseline (A) (tertile limits 99 and 108 cm) and at 17 weeks (B) (tertile limits 95 and 102 cm). P values are for trend by ANOVA.

EUROPEAN JOURNAL OF ENDOCRINOLOGY (1997) 137 Serum leptin in obesity effect of weight loss 67 In this study, serum leptin levels were estimated by radioimmunoassay. However, recent data indicate that, in rodents and in humans, the majority of leptin circulates bound to proteins in plasma and this binding is modified by obesity. In hyperleptinaemic human obesity, the free unbound leptin increases with body weight (23, 24). At present, the biological significance of this phenomenon is unknown, but it may potentially alter the bioactivity, transport and clearance of leptin. To conclude, serum leptin levels in obese subjects correlate with body fat and fat tissue topography. Furthermore, the changes in fat mass and leptin during weight loss are parallel, but the concentrations remain constantly greater in women. However, leptin levels do not seem to predict the response to weight reduction therapy. Acknowledgements This study was supported by grants from F Hoffman La Roche Ltd to the University of Kuopio and the Academy of Finland, Research Council for Health. References 1 Zhang YR, Proenca R, Maffel M, Barone M, Leopold L & Friedman JM. Positional cloning of the mouse obese gene and its human homologue. Nature 1994 372 425 432. 2 Campfield LA, Smith FJ, Guisez Y, Devos R & Burn P. Recombinant mouse OB protein: evidence for a peripheral signal linking adiposity and central neural networks. Science 1995 269 546 549. 3 Halaas JL, Gajiwala KS, Mafei M, Cohen SL, Chait BT, Rabinowitz D etal. Weight-reducing effects of the plasma protein encoded by the obese gene. Science 1995 269 543 546. 4 Pelleymounter MA, Cullen MJ, Baker MB, Hecht R, Winters D, Boone T et al. Effects of the obese gene product on body weight regulation in ob/ob mice. Science 1995 269 540 543. 5 Hamilton BS, Paglia D, Kwan AYM & Deitel M. Increased obese mrna expression in omental fat cells from massively obese humans. Nature Medicine 1995 1 953 956. 6 Lönnqvist F, Arner P, Nordfors L & Schalling M. Overexpression of obese (ob) gene in adipose tissue of human obese subjects. Nature Medicine 1995 1 950 953. 7 Masuzaki H, Ogawa Y, Isse N, Satoh N, Okazaki T, Shigemoto M et al. Human obese gene expression. Adipocyte-specific expression and regional differences in the adipose tissue. Diabetes 1995 44 855 858. 8 Considine RV, Considine EL, Williams CJ, Nyce MR, Magosin SA, Baver TL et al. Evidence against either a premature stop codon or the absence of obese gene mrna in human obesity. Journal of Clinical Investigation 1995 95 2986 2988. 9 Maffei MJ, Halaas J, Ravussin E, Pratley RE, Lee GH, Zhang Yet al. Leptin levels in human and rodent: measurement of plasma leptin and ob RNA in obese and weight-reduced subjects. Nature Medicine 1995 1 1155 1161. 10 Considine RV, Sinha MK, Heiman AL, Kriauciunas A, Stephens TW, Nyce MR et al. Serum immunoreactive-leptin concentrations in normal weight and obese humans. New England Journal of Medicine 1996 334 292 295. 11 Tartaglia LA, Dembski M, Weng X, Deng N, Culpepper J, Devos R et al. Identification and expression cloning of a leptin receptor, OB-R. Cell 1995 83 1263 1271. 12 Schwartz MW, Baskin DG, Bukowski TR, Kuijper JL, Foster D, Lasser G et al. Specificity of leptin action on elevated blood glucose levels and hypothalamic neuropeptide Y gene expression in ob/ob mice. Diabetes 1996 45 531 535. 13 Dryden S & Williams G. The role of hypothalamic peptides in the control of energy balance and body weight. Current Opinion in Endocrinology and Diabetes 1996 3 51 58. 14 Uusitupa M, Karhunen L, Rissanen A, Franssila-Kallunki A, Niskanen L, Kervinen K et al. Apolipoprotein E phenotype modifies metabolic and hemodynamic abnormalities related to central obesity in women. American Journal of Clinical Nutrition 1996 64 131 136. 15 Rindfrey H, Helger R & Lang H. Kinetic determination of glucose concentrations with glucose dehydrogenase. Journal of Clinical Chemistry and Clinical Biochemistry 1977 15 217-220. 16 Ma ZA, Gingerich RL, Santiago SV, Klein S, Smith HC & Landt M. Analysis of human plasma leptin by radioimmunoassay. Clinical Chemistry 1996 42 942 946. 17 Frederich RC, Hamann A, Anderson S, Löllman B, Lowell BB & Flier JS. Leptin levels reflect body lipid content in mice: evidence for diet-induced resistance to leptin action. Nature Medicine 1995 1 1311 1314. 18 Zimmet P, Hodge A, Nicolson M, Staten M, de Courten M, Moore J et al. Serum leptin concentration, obesity, and insulin resistance in Western Samoans: cross-sectional study. British Medical Journal 1996 313 965 996. 19 Nicklas BJ, Toth MJ, Goldberg AP & Poehlman ET. Racial differences in plasma leptin concentrations in obese postmenopausal women. Journal of Clinical Endocrinology and Metabolism 1997 82 315 317. 20 Funahashi TI, Shimomura I, Hiraoka H, Arai T, Takakashi M, Nakamura T et al. Enhanced expression of rat obese (ob) gene in adipose tissue of ventromedial hypothalamus (VMH)-lesioned rats. Biochemical and Biophysical Research Communications 1955 211 469 475. 21 Karhunen LJ, Haffner S, Lappalainen RI, Turpeinen AK, Miettinen H & Uusitupa MIJ. Serum leptin and short-term regulation of eating in obese women. Clinical Science 1997 92 573 578. 22 Ravussin E, Pratley RE, Maffei M, Wang H, Friedman JM, Bennett PH et al. Relatively low plasma leptin concentrations precede weight gain in Pima Indians. Nature Medicine 1997 3 238 240. 23 Sinha MK, Opentanova I, Ohannesian JP, Kolaizynski JW, Heiman ML, Hale J et al. Evidence of free and bound leptin in human circulation. Studies in lean and obese subjects and during short-term fasting. Journal of Clinical Investigation 1996 98 1277 1282. 24 Houseknecht KL, Mantzoros CS, Kuliawat R, Hadro E, Flier JS & Kahn BB. Evidence for leptin binding to proteins in serum of rodents and humans: modulation with obesity. Diabetes 1996 45 1638 1643. Received 23 December 1996 Accepted 17 March 1997